Assembly line fluid filling system and method

[0001] A computer program listings appendix is contained on a compact disc submitted herewith hereby incorporated by reference. One compact disc and one duplicate are submitted according to 37 C.F.R. 1.52(e) and each contains the following files:

	FILE NAME	SIZE IN BYTES	DATE OF CREATION
	Cal.rsp	1 KB	May 21, 1995
	Cal_eval.c	12 KB	Jul. 15, 1995
	Cal_main.c	72 KB	Sep. 12, 1995
	Cal_main.h	10 KB	Jul. 15, 1995
	Cal_tran.c	14 KB	Oct. 22, 1995
	Cal_util.c	8 KB	May 14, 1995
	Cc_cd.dat	9 KB	Oct. 28, 1995
	Cc_cpcc.dat	1 KB	Oct. 28, 1995
	Ccal.c	4 KB	Apr. 1, 1995
	Ccal.log	4 KB	Oct. 28, 1995
	Ccal.run	1 KB	Apr. 2, 1995
	Ccal_fo.c	4 KB	Apr. 1, 1995
	Changes.txt	6 KB	Oct. 28, 1995
	Csyn.c	7 KB	May 21, 1995
	Dpmpar.c	7 KB	Apr. 1, 1995
	Dpmpar.f	6 KB	Feb. 15, 1995
	Ecal.c	4 KB	May 17, 1995
	Ecal.log	3 KB	Oct. 28, 1995
	Ecal.run	1 KB	Oct. 28, 1995
	Ecalmain.c	23 KB	Oct. 15, 1995
	Ecccpcc.dat	1 KB	Oct. 28, 1995
	Encccpcc.dat	1 KB	Oct. 28, 1995
	Enorm.c	4 KB	Apr. 1, 1995
	Enorm.f	4 KB	Mar. 25, 1994
	F2c.h	5 KB	Feb. 25, 1992
	F2c.ps	138 KB	Oct. 20, 1995
	Faq.txt	12 KB	Oct. 28, 1995
	Fdjac2.c	5 KB	Apr. 1, 1995
	Fdjac2.f	4 KB	Mar. 25, 1994
	Gasdev.c	2 KB	Jul. 17, 1995
	Ic2wc.c	3 KB	Apr. 1, 1995
	Index.txt	3 KB	Oct. 28, 1995
	Lmdif.c	17 KB	Apr. 1, 1995
	Lmdif.f	16 KB	Mar. 25, 1994
	Lmpar.c	10 KB	Apr. 1, 1995
	Lmpar.f	9 KB	Mar. 25, 1994
	Makefile.bor	3 KB	Oct. 23, 1995
	Makefile.unx	3 KB	Oct. 20, 1995
	Matrix.c	12 KB	Jul. 15, 1995
	Matrix.h	1 KB	Jul. 15, 1995
	Minpack.rsp	1 KB	May 14, 1995
	Ncc_cd.dat	27 KB	Oct. 28, 1995
	Ncc_cpcc.dat	1 KB	Oct. 28, 1995
	Nccal.c	4 KB	May 20, 1995
	Nccal.log	4 KB	Oct. 28, 1995
	Nccal.run	1 KB	Apr. 2, 1995
	Nccal_fo.c	4 KB	May 20, 1995
	Ncsyn.c	7 KB	May 21, 1995
	Notes.txt	6 KB	Oct. 28, 1995
	Qrfac.c	7 KB	Apr. 1, 1995
	Qrfac.f	6 KB	Mar. 25, 1994
	Qrsolv.c	8 KB	Apr. 1, 1995
	Qrsolv.f	7 KB	Mar. 25, 1994
	Wc2ic.c	3 KB	Apr. 1, 1995
	Xfd2xfu.c	4 KB	Jul. 15, 1995

[0002] 1. Technical Field

[0003] This invention relates to a robotic assembly line filling system.

[0004] 2. Description of the Related Art

[0005] Referring to FIG. 1, a known system 10 for fueling an automotive vehicle 12 is shown. The vehicle 12 is moved along an assembly line via a conveyer line 14. When the vehicle 12 progresses into a vehicle fueling area an operator 16 inserts a fuel nozzle 18 into a fuel stem (not shown) to fuel the vehicle 12. Because the conveyer line 14 and the vehicle 12 are moving, a gantry 20 and a carriage 22 are utilized to move the fuel nozzle 18 along with the vehicle 12. After insertion of the nozzle 18, a predetermined amount of fuel is pumped into the fuel stem of the vehicle 12. Thereafter, the operator 16 may remove the nozzle 18 from the fuel stem or the nozzle 18 may be automatically removed from the fuel stem. The known fueling system 10, however, has a drawback in that the operator 16 must manually insert the fuel nozzle 18 into the fuel stem. Labor and associated manufacturing costs are increased.

[0006] U.S. Pat. No. 4,708,175 issued to Janashak et al. discloses a robot that fills a container mounted on a vehicle with a fluid. The vehicle is mounted on a conveyer line that moves into a work cell where the robot is located. The robot utilizes a vision system to determine the position of the inlet of the container. The robot then moves a robotic arm to the position of the inlet to fill the container with fluid, using gauge holes as visual target points. While Janashak et al. appear to disclose that the system may be used to fill a moving container provided the robot is capable of tracking the moving vehicle, no description of how this can be accomplished is provided. In particular, Janashak et al. do not disclose how to compensate for motions of the vehicle relative to the assembly line conveyor that cannot be detected by an encoder connected to an assembly line conveyor, do not disclose how to compensate for vehicle to vehicle variations in the location of the fuel stem relative to other parts of the vehicle, nor do Janashak et al. disclose how to accomplish such filling of a container while the vehicle moves without first touching the vehicle or attaching anything to the vehicle in order to facilitate detection of the container inlet or, finally, how to coordinate such motion to fill the container without constraining the motion of the vehicle in an uncharacteristic fashion (i.e., without stopping them, stabilizing them, etc. where such motion is only a requirement of the fueling system and not typical of the conveyor itself). Applicants assume that, Janashak et al. teach no more than that the conveyor line therefore must stop in order to fill the container. Stopping a conveyer line results in an increased time to complete the vehicle build, which results in increased manufacturing and labor costs.

[0007] There is thus a need for a fluid filling system and method that reduces and/or minimizes one or more of the above-identified deficiencies.

[0008] One advantage of the present invention is that the vehicles on the assembly line need not be stopped to have a fluid container, such as a fuel tank, filled.

[0009] The present invention, broadly, provides a system and method for filling a container with a fluid in a vehicle moving along an assembly line. The fluid can be, for example only, fuel, coolant, windshield washer fluid, and the like. The method includes the steps of determining a position of an inlet of the container (as the vehicle is moving) using a machine vision system. The next step involves moving a fluid fill outlet to the container inlet using a robot. Finally, the container is filled via the outlet while the vehicle continues to move.

[0010] In a preferred embodiment, a fueling system in accordance with the present invention includes a gantry having a carriage configured to move generally parallel to a first axis (i.e., the conveyor line on which the vehicle moves). The fueling system further includes a robotic arm attached to the carriage that moves with the carriage. The robotic arm has an end effector configured to mate with a fuel stem on the vehicle and to supply fuel to the fuel stem. The fueling system further includes a vision system including first and second cameras for iteratively determining a three-dimensional position of the fuel stem relative to a predetermined coordinate system. The position is preferably a center point of the fuel stem lying on a plane defined by the sealing surface (typically an outer edge) of the fuel stem. Finally, the fueling system includes a robot controller configured to command the carriage to move proximate to the position of the fuel stem and to move the end effector to the three-dimensional position to mate with the fuel stem. The robot controller is further configured to move the robotic arm relative to the first axis at a speed substantially equal to a speed of the conveyer line.

[0011] A method is also provided for fueling an automotive vehicle on a conveyer line. The method utilizes a robotic arm with an end effector for fueling a fuel stem of the vehicle. The method includes determining a first position of the fuel stem along a first axis. The method further includes moving the robotic arm to position the end effector proximate to the first position. The method further includes determining a three-dimensional position of the fuel stem utilizing a vision system. The method further includes moving the end effector proximate to the three-dimensional position to enable the end effector to mate with the fuel stem and moving the robotic arm relative to the first axis at a speed substantially equal to a speed of the conveyer line. Finally, the method includes fueling the fuel stem with the end effector.

[0012] A further method for providing position data to a controller is also provided. The method provides position data to the controller to enable a robotic arm controlled by the controller to move to a position of an object. The method utilizes first and second cameras disposed at first and second camera coordinate systems, respectively. The method in a preferred embodiment includes simultaneously acquiring a first digital image and a second digital image of a workspace including the object utilizing the first camera and the second camera, respectively. The method further includes searching the first digital image with a first fuel stem template to determine a location in the first image where a first correlation score between that portion of the image being searched and the template is greater than a predetermined threshold. The method further includes searching the second digital image with a second fuel stem template to determine a location in the second image where a second correlation score between that portion of the second image being searched and the template is greater than a predetermined threshold. The method further includes calculating a three-dimensional position of the object with respect to a predetermined coordinate system. The position is calculated responsive to the locations in the first and second digital images where the correlation scores were above the threshold. The method further includes calculating a triangulation error of the position. In a more preferred embodiment, the method further includes transferring position data indicative of the calculated position to the controller when the fuel stem matches are found in both images and the triangulation error is less than a threshold error value.

[0013] The fueling system for fueling an automotive vehicle on a conveyer line and the method related thereto represent a significant improvement over conventional fueling systems and methods. In particular, the inventive fueling system allows the automatic fueling of a fuel stem without the need for an operator, resulting in labor savings. Further, the inventive fueling system can track the position of the fuel stem on a moving vehicle (on a conveyer line) and move an end effector to mate with the fuel stem. Thus, the vehicle can be fueled without stopping the conveyer line resulting in decreased manufacturing costs and increased manufacturing efficiency.

[0014] These and other features and advantages of this invention will become apparent to one skilled in the art from the following detailed description and the accompanying drawings illustrating features of this invention by way of example.

[0015] FIG. 1 is a schematic of a known fueling system for an automotive vehicle on a conveyer line.

[0016] FIG. 2 is a schematic of a fueling system in accordance with the present invention.

[0017] FIG. 3 is a schematic of a gantry and a robotic arm shown in FIG. 2.

[0018] FIG. 4 is a side view of the gantry and the robotic arm shown in FIG. 3.

[0019] FIG. 5 is an enlarged perspective view of the robotic arm shown in FIG. 3.

[0020] FIG. 6 is a side view of the robotic arm in its kinematic zero position (all joint variables equal to zero).

[0021] FIG. 7 is a schematic showing the coordinate systems utilized to control the position of the robotic arm.

[0022] FIG. 8 is an exploded schematic of an end effector of the robotic arm shown in FIG. 5.

[0023] FIG. 9 is block diagram of a robotic control system utilized by the inventive fueling system.

[0024] FIGS. 10A-10C are schematics illustrating the tracking variables utilized for positioning the robotic arm.

[0025] FIG. 10D is a tracking equation utilized to position the robotic arm along the J1 axis.

[0026] FIG. 11 is a schematic illustrating a triangulation technique utilized by the vision system to determine the three dimensional position of the fuel stem.

[0027] FIG. 12 is a schematic illustrating a two-dimensional digital image coordinate system.

[0028] FIG. 13 is a schematic illustrating a two-dimensional camera sensor coordinate system.

[0029] FIG. 14 is a schematic illustrating a three-dimensional camera coordinate system.

[0030] FIG. 15 is a schematic illustrating equations that define the intrinsic camera model.

[0031] FIG. 16 is a schematic illustrating a force vector applied to the end effector.

[0032] FIG. 17 is a schematic illustrating a Jacobian Transpose relationship utilized for force feedback control of the robotic arm.

[0033] FIG. 18 is a schematic illustrating an inverse kinematics model utilized by the robot control system to position and orient an end effector.

[0034] FIG. 19 is a schematic of a fuel stem.

[0035] FIG. 20 is a schematic of a docking path of an end effector with a fuel stem.

[0036] FIGS. 21A-21G are flowcharts illustrating the modes of operation for the fueling system in accordance with the present invention.

[0037] Referring now to the drawings wherein like reference numerals are used to identify identical components in the various views, FIG. 2 illustrates a fueling system 24 for fueling an automotive vehicle 26 on a conveyer line 28. As shown, the vehicle 26 is on a carriage 30 moving on the conveyer line 28. It should be understood, however, that the inventive fueling system 24 may be configured to operate with any known type of conveyer line, including for example, overhead conveyer lines (see FIG. 1) and floor mounted conveyer lines. An advantage of the fueling system 24 is that the vehicle 26 may be fueled without stopping the vehicle 26 on the conveyer line 28. The inventive fueling system 24 includes a gantry 32, a robotic arm 34, and a robot control system 36.

[0038] Referring to FIG. 3, the gantry 32 is provided to move the robotic arm 34 substantially parallel to the conveyer line 28. The gantry 32 includes a frame 38, a carriage 40, and a motor 42. The frame 38 may be supported by legs (not shown) or may be mounted to ceiling supports. The motor 42 is mounted on the gantry 32 and is operatively connected to a drive belt (not shown) that is connected to the carriage 40. Thus, rotational movement of a rotor (not shown) of the motor 42 causes the drive belt to move the carriage 40 along the J1 axis in either a forward direction (to the right in FIG. 2) or a backward direction (to the left in FIG. 2). The motor 42 is electrically connected to a motor driver 44 (see FIG. 9) that is controlled by the robot controller 46. Thus, the controller 46 can selectively control the position of the carriage 40 and the robotic arm 34 along a J1 axis. The motor 42 further includes an internal encoder 48 (not shown in FIG. 1). The encoder 48 generates a position value J1_VAL indicative of the position of the carriage 40 that is received by the robot controller 46.

[0039] Referring to FIG. 3, the robotic arm 34 is provided to fuel the automotive vehicle 26. As shown, the robotic arm 34 is connected to the carriage 40 and moves with the carriage 40 along the J1 axis. Prior to fueling the vehicle 26, the robotic arm 34 is rotated to a desired angle about the J2 axis which matches the orientation of a fuel stem 90 in the plane of J2 (see FIG. 10A) of the vehicle 26. In a preferred embodiment, the robotic arm 34 is automatically rotated to the desired J2 angle. In an alternate embodiment, the J2 angle could be manually fixed. Referring to FIG. 5, the robotic arm 34 includes a manipulator arm 50, a pitch arm 52, an end effector 54, a camera arm 56, and a light 58.

[0040] The manipulator arm 50 is provided to support the pitch arm 52 and the end effector 54. Referring to FIGS. 4 and 5, the manipulator arm 50 includes a frame 60, a motor 62, a linear actuator 64, and a joint 66. As shown, the frame 60 is connected to a first end of the pitch arm 52 via the joint 66. The motor 62 is provided to drive the linear actuator 64 to thereby move the pitch arm 52 about the J3 axis. In particular, the linear actuator 64 converts rotary movement of a rotor (not shown) of the motor 62 into linear movement of the upper end of a rod 68. The lower end of the rod 68 is connected to the pitch arm 52 via a pin joint 70. Thus, linear actuation of the upper end of the rod 68 causes rotational motion of the pitch arm 52 about the J3 axis.

[0041] Referring to FIGS. 5 and 9, the motor 62 is electrically connected to a motor driver 72 that is controlled by the robot controller 46. Thus, the controller 46 can selectively control the position of the linear actuator 64 to move the pitch arm 52 to a desired rotational angle about the J3 axis. Further, an encoder 74 is operatively mounted on the joint 66. The encoder 74 generates a position value 3_VAL indicative of a rotational angle of the pitch arm 52 about the J3 axis that is received by the robot controller 46. Still further, an encoder 76 is operatively mounted between the carriage 40 and the manipulator arm 50. The encoder 76 generates a position value 2_VAL indicative of a rotational angle of the manipulator arm 50 about the J2 axis that is received by the robot controller 46.

[0042] Referring to FIG. 4, the pitch arm 52 is provided to support the end effector 54. The pitch arm 52 includes an L-shaped frame 78, a pneumatic cylinder 80, and a joint 82. As illustrated, a second end of the pitch arm 52 is connected to the end effector 54 via the joint 82. The pneumatic cylinder 80 is mounted on the pitch arm 52 and has a cylinder rod 84 connected to the end effector 54. Further, the cylinder 80 is connected to a pneumatic control valve 86 (see FIG. 9) which is selectively controlled by the robot controller 46. The valve 86 may be a dual solenoid three-position open or closed center valve having an extend solenoid (not shown) and a retract solenoid (not shown). When the extend solenoid is energized (and the retract solenoid is de-energized), the cylinder 80 extends the rod 84 to rotate the end effector 54 to a desired rotational angle about the J4 axis as defined by an adjustable hard stop. The hard stop connected to the cylinder rod is adjusted prior to moving the actuator (which only makes full motion movements). This may be done manually, but is preferably done automatically. This method of controlling the joint position takes advantage of the speed, power, and safety (i.e., in proximity to gasoline) of a pneumatic actuator while avoiding the difficulty in mid-positioning pneumatic actuators, and particularly avoiding the lack of stiffness exhibited by a mid-positioned pneumatic actuator. The hard stop is preferably moved by a low power electric or pneumatic motor using a self-locking mechanism such as a worm gear drive or a self-locking lead screw so that the power required to move the stop is less than the power required to move the joint. When the retract solenoid is energized (and the extend solenoid is de-energized) the cylinder 80 retracts the rod 84 to rotate the end effector 54 back to a home position about the J4 axis. Referring to FIGS. 5 and 9, an encoder 88 is operatively mounted on the joint 82. The encoder 88 generates a position value 4_VAL indicative of a rotational angle of the pitch arm 52 about the J4 axis that is received by the robot controller 46.

[0043] The end effector 54 is provided to mate with the fuel stem 90 of the automotive vehicle 26. Referring to FIG. 8, the end effector 54 includes a rodless cylinder 92, a mounting bracket 94, a guide 96, a manifold 98, a fuel hose 100, a boot 102, a front housing 104, load cells 106, 108, 110, 112, pneumatic cylinder 114, fuel valves 116, 118, elbows 120, 122, 124, 126, 128, 130, pipe tees 132, 134, 136, 138, check valves 140, 142, vapor fittings 144, 146, 148, 150, and a position potentiometer 152.

[0044] The rodless cylinder 92 is provided to move the mounting bracket 94 and the remaining components of the end effector 54 along the J5 axis. In particular, the cylinder 92 is provided to move the boot 102 against the fuel stem 90. The cylinder 92 has a slidable plate 154 that may be extended or retracted along the J5 axis. Referring to FIGS. 8 and 9, the cylinder 92 may be operatively connected to a pneumatic control valve 156 that is controlled by the robot controller 46. As shown, the plate 154 is attached to the mounting bracket 94.

[0045] Referring to FIG. 8, the mounting bracket 94 is provided to support the remaining components of the end effector 54 (excluding the rodless cylinder 92). As shown, the guide 96 is attached to bracket 94 to allow axial movement of the manifold 98 and the fuel hose 100 relative to the bracket 94 and the boot 102.

[0046] The manifold 98 is provided to direct fuel from either fuel valve 116 or fuel valve 118 through the fuel hose 100. Further, the pneumatic cylinder 114 is provided to selectively move the manifold 98 (discussed in greater detail below) along the guide 96 relative to the bracket 94.

[0047] The fuel valve 116 may receive fuel (e.g., gasoline) from the elbow 122 which is connected to a first fuel line (not shown). When the valve 116 is open, fuel is supplied to the manifold 98 and to the fuel hose 100. When the fuel valve 116 is closed, the fuel may be recirculated through elbow 120 for cooling by cooling equipment (not shown). The valve 116 is selectively controlled by the robot controller 46.

[0048] The fuel valve 118 may receive fuel (e.g., diesel fuel) from the elbow 126 which is connected to a second fuel line (not shown). When the valve 118 is open, fuel is supplied to the manifold 98 and to the fuel hose 100. When the fuel valve 118 is closed, no fuel is supplied to the manifold 98. The valve 118 is also selectively controlled by the robot controller 46.

[0049] The boot 102 is provided to mate with the fuel stem 90 during fueling of the stem 90. The boot 102 may be constructed of a resilient material such as rubber or plastic and includes a hollow body portion 103 and a boot seal 101. The boot seal 101 is configured to seal against a top edge of the fuel stem 90 to prevent fuel and fuel vapor from escaping from the stem 90 during fueling. As shown, the boot 102 is attached to the front housing 104. The boot is also electrically connected to the front housing, and provides a conductive path which grounds the robot to the vehicle upon insertion.

[0050] The fuel hose 100 is provided to be inserted into the fuel stem 90 beneath the no-lead insert (not shown) after the boot 102 has mated with the stem 90. The fuel hose 100 is constructed of a flexible plastic and extends from an outlet (not shown) on the manifold 98 and through the housing 104 and the boot 102. As shown, the pneumatic cylinder 114 is provided to move the manifold 98 and the fuel hose 100 along the J5 axis to insert the hose 100 into the stem 90. Referring to FIGS. 8 and 9, the cylinder 114 is operatively connected to a pneumatic control valve 158 that is selectively controlled by the robot controller 46.

[0051] The check valves 140, 142 are provided to purge manifold 98 and fuel hose 100 of fuel. When both fuel valves 116, 118 are closed, an air supply (not shown) may apply air through both of the check valves 140, 142 to force any fuel remaining in the manifold 98 and the fuel hose 100 into the fuel stem 90.

[0052] The vapor recovery fittings 144, 146, 148, 150, pipe tee 138, and elbow 130 may be connected together for fuel vapor recovery as known by those skilled in the art.

[0053] The load cells 106, 108, 110, 112 are provided to measure a force applied to the end effector 54 by the fuel stem 90 during fueling. The load cells 106, 108, 110, 112 are conventional in the art and are mounted between the front housing 104 and the mounting bracket 94. Thus, the front housing 104 transmits force exerted on the boot 102 to the load cells 106, 108, 110, 112. Referring to FIG. 9, the load cells 106, 108, 110, 112 are electrically connected to the robot controller 46. The controller 46 may receive signals generated by the load cells 106, 108, 110, 112 and calculate a force vector for force feedback control of the robotic arm 34 during fueling.

[0054] Referring to FIGS. 4 and 8, during the process of mating the boot 102 with the fuel stem 90, the end effector 54 is rotated about the joint 82 to a calculated rotation angle ₄. The rodless cylinder 92 then extends the remaining components of the end effector 54 along the J5 axis to mate the boot 102 with the fuel stem 90. Finally, the pneumatic cylinder 114 extends the fuel hose 100 along the J5 axis to enter the fuel stem 90 beneath the no-lead insert (not shown).

[0055] Referring to FIG. 5, the camera arm 56 is provided to house cameras 182, 184. As shown, the camera arm 56 is attached to a side of the manipulator arm 50.

[0056] The light 58 is provided to illuminate the fuel stem 90 on the vehicle 26 to allow the vision system 164 to recognize the fuel stem 90 regardless of ambient lighting conditions. The light 58 may comprise a fluorescent metal halide fixture (e.g., a Class I Div. 2 rated fixture in a constructed embodiment). Referring to FIG. 7, the coordinate systems utilized by the robot controller 46 to position the robotic arm 34 (and the end effector 54) are shown. In particular, a coordinate system CS0 represents a home position of the carriage 40. The origin of the coordinate system CS0 lies on the J1 axis. A coordinate system CS1 is located on the carriage 40 and accordingly moves with the carriage 40. The coordinate system CS1 is utilized extensively for tracking the fuel stem 90 along the J1 axis, as discussed in greater detail below. The coordinate system CS2 is utilized for positioning the robotic arm 34 about the J2 axis. Coordinate system CS2 is also the reference coordinate system for the vision system, which uses it to locate and orient the camera coordinate systems and to report object position and orientation vectors. The coordinate system CS3 is utilized for positioning the pitch arm 52 about the J3 axis. Further, the coordinate system CS4 is utilized for positioning the end effector 54 about the J4 axis. Finally, the coordinate system CS5 is utilized for positioning the boot 102 along the J5 axis.

[0057] Referring to FIG. 9, a robot control system 36 is provided. The robot control system 36 includes a light sensor 160, a conveyer line encoder 162, a robot controller 46, a vision system 164, encoders 48, 74, 76, 88, the position potentiometer 52, load cells 106, 108, 110, 112, motor drivers 44, 72, control valves 86, 156, fuel valves 116, 118, and the fuel hose extend valve 158.

[0058] Referring to FIG. 2, the light sensor 160 is provided to detect when the vehicle 26 enters a vehicle fueling area on an assembly line. The light sensor 160 is conventional in the art and includes a light transmitter and a light receiver in package 166 to be used with a corresponding reflector 168. The package 166 is oriented on a first side of the conveyer line 28 to project a light beam, in one embodiment, across the conveyer line 28. In the illustrated embodiment, the reflector 168 is oriented and positioned on an opposite side of the conveyer line 28 to receive the light beam. When the vehicle 26 enters the fueling area, the vehicle 26 blocks the light beam from being reflected by the reflector 168. In response, the receiver in package 168 generates a trigger signal V_TRIG that is received by the robot controller 46. In the case where the beam is reflected by the vehicle, it is the presence (not absence) of a reflected beam that generates the trigger signal. In any event, the position of the light sensor 160 is designated as the light sensor trigger position hereinafter (see FIG. 10A).

[0059] In an alternate embodiment V_TRIG could be generated by any other sufficiently repeatable mechanism (e.g., mechanical switch, ultrasonic sensor, bar code reader, etc.) to reliably indicate the presence of the next vehicle to the robot controller.

[0060] Referring to FIGS. 2 and 9, the conveyer line encoder 162 is provided to generate an encoder count CL_VAL indicative of a current encoder count with respect to the conveyer line 28. The robot controller 46 utilizes two encoder counts to determine a gross displacement distance along the axis 170 of the vehicle 26 (and the fuel stem 90) from the light sensor trigger position.

[0061] Referring to FIG. 9, the robot controller 46 is provided to control the carriage 40 and the robotic arm 34 for fueling the automotive vehicle 26. The robot controller 46 has a communication bus 172 for communicating with the vision controller 174. Further, robot controller 46 also has a communication bus 176 for communicating with a supervisory PLC 178 which may control the operation of the fuel pumping and metering equipment and various safety devices. The PLC 178 is not a component of the robot control system 36 and is shown only for illustrative purposes. As shown, the robot controller 46 is electrically connected to the conveyer line encoder 162, the light sensor 160, the load cells 106, 108, 110, 112, and the vision controller 174. The robot controller 46 receives encoder counts (from the conveyer line encoder 162), and the signals V_TRIG, V_L1, V_L2, V_L3, V_L4, and a three-dimensional position P_e of the fuel stem 90 from the vision controller 174to control the position of the robotic arm 34. The method for determining the position of the fuel stem 90 will be explained in greater detail below. The controller 46 is also electrically connected to encoders 48, 74, 76, 88 and the position potentiometer 52. Thus, the controller 46 receives the joint position values J1_VAL, 2_VAL, 3_VAL, 4_VAL, J5_VAL and is able to calculate a current position of the end effector 54. The controller 46 is also electrically connected to motor drivers 44, 72, control valves 86, 156, 158, fuel valves 116, 118, and generates control signals to control the foregoing devices. The controller 46 further includes a programmable memory for implementing the various modes of operation of the fueling system 24 which will be described in greater detail below.

[0062] Referring to FIG. 18, the robot controller 46 utilizes an inverse kinematic model of the robotic arm 34 to position the arm 34. In particular, the inverse kinematic model is a set of equations that allow the joint variables (i.e., 3 and J5 of the robotic arm 34 to be calculated in order to place the end effector 54 in a desired position and orientation. Inverse kinematic models are utilized extensively in robotic applications and can readily be determined by one skilled in the art. Accordingly, the underlying inverse kinematic model equations will not be discussed in any further detail. As shown, one set of inputs x, y, z represent the desired position P_e of the end of boot 102 with respect to a predetermined coordinate system, such as coordinate system CS1. The inputs x, y, z (i.e., point P_e) are determined by the vision system 164 while tracking the fuel stem 90 and may be stored in the variables CARRIAGE_TO_STEMX, CARRIAGE_TO_STEMY, CARRIAGE_TO_STEMZ in the robot controller 46. The input a_x, a_y, a_z, is a unit vector representing the orientation of the fuel stem 90 (see FIG. 20) with respect to the coordinate system CS 1. Because the orientation of the fuel stem 90 does not vary a large amount when the vehicle 26 moves on the conveyer line 28, the orientation of the fuel stem 90 may be a stored value. In particular, unit vector a_x, a_y, a_z, may be stored as VID data in non-volatile memory (e.g., hard disk) of the vision controller and is readily determined by one skilled in the art for a specific fuel stem 90. The DH constants (see FIGS. 6 and 7) are geometric dimensions relating to the robotic arm 34 and are stored in the non-volatile memory (e.g., hard disk) of the robot controller 46. Finally, the angles 2 and 4 represent desired positions of the robotic arm 34 about the J2 axis and J4 axis, respectively, for fueling the fuel stem 90. The angles 2 and 4 may also be stored in the non-volatile memory (e.g., hard disk) of the robot controller 46 or may be calculated from the unit vector a_x, a_y, a_z.

[0063] Referring to FIG. 9, the vision system 164 is provided to generate a three-dimensional position value (i.e., point P_e) of a fuel stem 90 with respect to a predetermined coordinate system (e.g., CS1). The vision system 164 includes the vision controller 174, the frame grabber 180, inner camera 182, and outer camera 184.

[0064] The vision controller 174 is provided to calculate a three-dimensional position P_e of the fuel stem 90. In particular, the robot controller 46 may request a position of the fuel stem 90 from the controller 174 via the bus 172. In response, the controller 174 may calculate the position P_e and return the position P_e to the controller 46. The vision controller 174 may calculate the position of the fuel stem 90 responsive to two digital images generated by the cameras 182, 184which will be explained in greater detail below.

[0065] The robot controller 46 may also request vehicle identification data (VID) from the vision controller 174. The VID data may be stored in non-volatile memory (e.g., hard disk) of the vision controller 174. In particular, a user or an assembly line controller may input a VID number identifying the particular vehicle type, via an input device or a serial bus (not shown), to the robot controller 46. The robot controller 46 may transmit the VID number to the vision controller 174, which retrieves a VID record from its hard disk. The VID record may be transmitted to the robot controller 46 and maintained in a random access memory (RAM) of the vision controller 174. The VID record contains vehicle dependent information used by the robot controller 46 to track the position of the fuel stem 90. In particular, the VID record may include the following information:

[0066] a_x, a_y, a_z, unit vector defining orientation of the fuel stem 90 with respect to the coordinate system CS1;

[0067] HOME_TO_STEMdistance from a home position of the carriage 40 to a position P₀ on the J1 axis projection of fuel stem center onto z0) when the vehicle 26 crosses the light sensor trigger position (see FIG. 10A);

[0068] DESIRED_CARRIAGE_TO_STEMYdesired distance measured with respect to y1 from the carriage 40 to a position P₁ on the J1 axis for the end effector 54 to mate with the fuel stem 90;

[0069] DIGITAL_IMAGE_TEMPLATE1digital image template of the fuel stem 90 for the inner camera 182;

[0070] DIGITAL_IMAGE_TEMPLATE2digital image template of the fuel stem 90 for the outer camera 184;

[0071] APPROACH_ANGLEthe angular offset in the vertical plane between z5 and the stem axis 243.

[0072] The information contained in the VID record may be determined by one skilled in the art. In a preferred embodiment, the information is measured and stored by the vision controller 174 and the robot controller 46. In an alternate embodiment, the information could be determined using conventional survey equipment.

[0073] The vision controller 174 may utilize commercially available vision software for template matching to determine if cameras 182 and 184 are viewing the fuel stem 90. In a constructed embodiment, the commercially available vision software comprised Matrox Imaging Library, Version 6.0 sold by Matrox Electronic Systems Ltd. of Dorval, Quebec, Canada. In particular, the conventional software searches the first digital image of the workspace, acquired by camera 182 and frame grabber 180, for a location in which the correlation score with the first digital image template (i.e., DIGITAL_IMAGE_TEMPLATE1) is higher than a pre-determined acceptance level. If such an image location is found then there is a fuel stem match in the first digital image. The conventional software searches the second digital image of the workspace, acquired by camera 184 and frame grabber 180, for a location in which the correlation score with the second digital image template (i.e., DIGITAL_IMAGE_TEMPLATE2) is higher than a pre-determined acceptance level. If such an image location is found then there is a fuel stem match in the second digital image. If there is a fuel stem match in both digital images, i.e., if both cameras 182, 184 are viewing the fuel stem 90, then the vision controller 174 further proceeds to calculate a three-dimensional point P_e corresponding to the center of the fuel stem. The software methodology for calculating the point P_e will be explained in greater detail below.

[0074] The frame grabber 180 is provided to simultaneously trigger the cameras 182, 184 and to simultaneously digitize the first and second analog images acquired by the cameras 182, 184, respectively, and to transmit the first and second digital images to the vision controller 174. The frame grabber 180 is conventional in the art and may comprise a multi-channel frame grabber capable of simultaneously triggering and digitizing images coming from at least two monochrome progressive scan analog cameras. In a constructed embodiment, frame grabber 180 comprises a Matrox Meteor-II/MC Frame Grabber manufactured by Matrox Electronics Ltd. of Dorval, Quebec, Canada. The frame grabber 180 may receive a retrieve signal V_TR from vision controller 174. In response, the frame grabber 180 may generate retrieve signals V_TRI and V_TRO to simultaneously trigger the cameras 182, 184 to acquire the first and second analog images, respectively, of the workspace 186. The cameras 182, 184 may simultaneously begin to transfer the first and second analog images to the frame grabber 180, which may simultaneously digitize them and may further transfer the first and the second digital images to the vision controller 174.

[0075] The first and second cameras 182, 184 are used to acquire the first and the second analog images, respectively, of the workspace 186. The cameras 182, 184 are conventional in the art and may comprise monochrome CCD cameras having a progressive scan capability and a trigger shutter mode capability. In a constructed embodiment, each of cameras 182, 184 comprised a XC-55 Progressive Scan Camera Module manufactured by Sony Electronics Inc. of Itasca, Ill. As previously discussed, the cameras 182, 184 may acquire first and second analog images of the workspace responsive to the trigger signals V_TRI and V_TRO, respectively.

[0076] Referring to FIG. 5, the cameras 182, 184 are mounted within the camera arm 56. The camera arm 56 is provided to serve a number of functions. It acts as an enclosure which protects the cameras 182, 184 and their lenses (not shown) from dust, drips, and tampering. The camera arm is mounted to the manipulator arm 34 using precision shoulder screws (not shown) to guarantee precise and repeatable alignment. This allows camera arms 56 to be factory calibrated but interchangeable in the field. Finally, the camera arm is provided to position the cameras 182, 184 such that their fields of view cover the workspace 186 (see FIG. 4). The workspace 186 is defined to be the volume of space containing the various possible positions in which the robotic manipulator 34 and the end effector 50 are capable of servicing the fuel stem 90. Thus, when the cameras 182, 184 are disposed proximate to the fuel stem 90, such that the fuel stem 90 lies within the workspace 186, the vision controller 174 is able to determine a three-dimensional position P_e of the fuel stem 90.

[0077] Before explaining the various modes of operation of the fueling system 24, the methodology and mathematics for determining a three-dimensional position P_e of the fuel stem 90 will now be explained. In order to determine the position P_e, a mathematical camera model of each of the cameras 182, 184 is utilized. These mathematical models are identical except for the numerical values of parameters measured after the cameras 182,184 have been mounted in the camera arm 56 and the aperture and focus of their lenses have been adjusted. For the purposes of simplicity and clarity, only the mathematical model for camera 182 will be further described hereinbelow.

[0078] The camera model comprises two parts: an intrinsic camera model and an extrinsic camera model. Referring to FIGS. 12 and 14, the intrinsic camera model specifies the relationship between a point (i.e., point (x_i,y_i)) with respect to a digital image coordinate system and a direction in space (i.e., unit vector D_i) with respect to a camera coordinate system CS_ci. The intrinsic camera model for camera 182 depends on the internal geometric, electronic and optical characteristics of camera 182 and on the electronic characteristics of the frame grabber 180 which will be discussed in greater detail below.

[0079] During run time, the extrinsic camera model of each the cameras 182, 184 specifies the position and orientation of the camera coordinate system CS_c (i.e., coordinate system CS_ci for the camera 182 and coordinate system CS_co for the camera 184) with respect to a predetermined coordinate system that is fixed with respect to the cameras, such as CS2. In a constructed embodiment, the extrinsic camera model parameters are measured in the factory and are stored in non-volatile memory (e.g., hard disk) of the vision controller with respect to the camera arm 56 mounting frame. During run time the vision controller 174 always supplies position and orientation vectors with respect to CS2 but the robot controller 46 always converts and uses these vectors with respect to CS1. Because each extrinsic camera modelwhich is a coordinate system transformation matrixmay be readily determined by one skilled in the art, only the intrinsic camera model will be discussed.

[0080] The intrinsic camera model for camera 182 will now be discussed. The intrinsic camera model comprises several equations (discussed hereinafter) that are used to determine a direction vector D_i pointing in space to the center point P_i of the fuel stem 90. Referring to FIGS. 12, 13, and 14, the intrinsic camera model utilizes three coordinate systems: a digital image coordinate system, a camera sensor coordinate system, and a camera coordinate system. Referring to FIG. 12, the two-dimensional digital image coordinate system defined by the axes X_i and Y_i is illustrated. The digital image coordinate system is defined by a plurality of rows and columns of pixels 185. Further, the Principal Point (C_x, C_y) is the point projected onto the digital image coordinate system that corresponds to the optical axis of a camera lens (not shown) of the camera 182. Further, the image point (x_i, y_i) corresponds to the center of the fuel stem 90 and is generated by the commercially available vision software previously discussed.

[0081] Referring to FIG. 13, the two-dimensional camera sensor coordinate system defined by the axes X_s and Y_s is illustrated. The CCD camera 182 has a sensor plane 188 comprising a plurality of rows and columns of CCD sensor cells 187. The center of the fuel stem 90 is imaged on the sensor point (x_s, y_s) (coordinates with respect to the camera sensor coordinate system) and after digitization by the frame grabber 180, this point corresponds to the image point (x_i, y_i) (coordinates with respect to the digital image coordinate system).

[0082] Referring to FIG. 14, the three-dimensional camera sensor coordinate system CS_ci is partially illustrated in two dimensions. The camera coordinate system CS_ci is defined by the axes X_ci, Y_ci, Z_ci has an origin O_ci that corresponds to the center of a lens (not shown) of the camera 182. The origin O_ci is also located on the optical axis of the camera 182. As shown, the unit direction vector D_i is determined with respect to the camera coordinate system CS_ci.

[0083] Before discussing the intrinsic camera model equations, the parameters used in the equations will be set forth. The parameters include:

	f	effective focal length of the camera lens
		(meters),
	Sx	camera aspect ratio,
	Cx	x image coordinate (pixel) of the Principal
		Point,
	Cy	y image coordinate (pixel) of the Principal Point,
	K1	second degree radial distortion coefficient of the
		camera lens,
	dx	width of a sensor cell (meters),
	dy	height of a sensor cell (meters),
	Ncx	number of sensor cells on a row of the camera
		Sensor,
	Ncy	number of sensor cells on a column of the camera
		Sensor,
	N1x	number of pixels on an image row.

[0084] The parameters d_x, d_y, N_cx, N_cy, N_ix may be supplied by the manufacturer of the camera 182, and the parameter N_ix is a known parameter of the frame grabber 180. The parameters f, S_x, C_x, C_y, K₁, may be readily determined by one skilled in the art utilizing an algorithm set forth in the following publication: A Versatile Camera Calibration Technique for High-Accuracy 3D Machine Vision Metrology Using Off-the-Shelf TV Cameras and Lenses, IEEE Journal of Robotics and Automation Vol RA-3, No 4, August 1987which is incorporated by reference in its entirety.

[0085] Referring to FIG. 15, camera model equations 1-8 are used to map the image point (x_i, y_i) with respect to a digital image coordinate system to a sensor point (x_s, y_s) with respect to a camera sensor coordinate system and to further calculate an undistorted sensor point (x_u, y_u) with respect to the camera sensor coordinate system, from which is calculated a direction vector D_i with respect to the camera coordinate system CS_ci. In particular, equations (1), (2), (7), and (8) are used to map the image point (x_i, y_i) to the sensor point (x_s, y_s)where light from a point P_i of the fuel stem 90 is projected. Referring to FIG. 14, sensor point (x_s, y_s) is shown on the sensor plane 188. It should be understood, however, that camera 182 (and all conventional cameras) radially distorts projected light beams. Accordingly, the equation (3) is utilized to calculate a radial distortion factor k for the camera 182 to compensate for the radial distortion of camera 182. Next, equations (4) and (5) are utilized to calculate the undistorted sensor point (x_u, y_u) (coordinates with respect to the camera sensor coordinate system) utilizing the radial distortion factor k. Referring to FIG. 14, the point (x_u, y_u) is shown on the sensor plane 188. The preferred embodiment uses an improved version of Tsai's algorithm that was developed by Reg Willson and that is freely available on Internet at the following URL: http://www.cs.cmu.edu/afs/cs.cmu.edu/user/rgw/www/TsaiCode.html, as also shown in the computer program listing Appendix on a compact disc submitted herewith.

[0086] Referring to FIG. 15, the equation (6) is used to calculate the unit vector D_i which is centered at the origin O_ci and points toward the center point P_i of the fuel stem 90. The numerator of the equation (6) represents a vector pointing toward the center point P_i. The denominator of equation (6) represents the magnitude of the vector D_i. Thus, the vector D_i represents a unit vector.

[0087] The foregoing camera model equations are valid if the following criteria regarding the camera 182, the lens (not shown) and the frame grabber 180 are true. First, the optical axis of the camera 182 is presumed to be perpendicular to the sensor plane 188. Second, the lens is focused so that an object in the camera workspace 186 is in focus. Third, the focus of the lens is fixed. In other words, a zoom lens is not used. Fourth, the digitization of a digital image by the frame grabber 180 is accurate.

[0088] It should be understood that the intrinsic camera model for camera 184 may also be defined by the equations (1)-(8) utilizing the ten parameters f, S_x, C_x, C_y, K₁, d_x, d_y, N_cx, N_cy, N_ix determined for the camera 184. Referring to FIG. 11, the camera 184 may have an origin O_co with a coordinate system CS_co. Generally, to determine a position of an object, an observer first establishes a stereo match of the object and then uses triangulation to determine a distance to the object with respect to the observer. The vision system 164 works in a similar manner. When the cameras 182, 184 and the frame grabber 180 generate first and second digital images, respectively, of the workspace 186 the conventional vision software (discussed above), within the vision controller 174, performs image template matching and generates first and second image match locations. We will hereafter refer to the first image match location as the inner image location and to the second image match location as the outer image location.

[0089] The conventional vision software determines an inner image point (i.e., x_i, y_i) and an outer image point (not shown) that are representative of the center point of the fuel stem 90 in the first and second digital images, respectively. The inner image point (i.e., x_i, y_i) is utilized to calculate the direction vector D_i as explained above. Similarly, the outer image point is utilized to calculate the direction vector D_o. It should be understood that the direction vector Do may be calculated utilizing the equations (1)-(8) discussed above. The two direction vectors D_i and D_o are utilized in conjunction with the extrinsic camera model of the cameras 182, 184 to triangulate as estimated position P_e of the fuel stem 90 with respect to the coordinate system CS2. This position with respect to CS2 is transmitted by the vision controller 174 to the robot controller 46 which transforms it with respect to CS1.

[0090] Referring to FIG. 11, to triangulate the position of the estimated triangulation point P_e with respect to the coordinate system CS2, the position and orientation of the inner and outer camera coordinate systems CS_co and CS_ci with respect to coordinate system CS2, i.e., the extrinsic camera models of cameras 182, 184 must be known. The transformation of a first coordinate system (i.e., CS_co or CS_ci) to a second coordinate system (i.e., CS2) is well known in the art and will not be discussed in any further detail. The vision controller 174 calculates an inner triangulation point P_i along the line O_ci, D_i that is the closest point to the line O_co, D_o. The controller 174 then calculates an outer triangulation point P_o that is the closest point on the line O_co, D_o to the inner triangulation point P_i. Finally, the controller 174 calculates the midpoint between the inner triangulation point P_i and the outer triangulation point P_o, which is the triangulated point P_e. As previously discussed, the triangulated point P_e is the estimated position of a center point of the fuel stem 90 with respect to the coordinate system CS2. Further, the vision controller 174 calculates a triangulation error which is the distance between points P_e and P_i. If the triangulation error is less than a threshold error value, the vision controller 174 transmits the position P_e to the robot controller 46. In particular, the position P_e may be transmitted to the controller 174 utilizing the following variables: CARRIAGE_TO_STEMX, CARRIAGE_TO_STEMY, CARRIAGE_TO_STEMZ.

[0091] During fueling of the fuel stem 90, force feedback control may be utilized to correct the position of the robotic arm 34. In particular, if a force vector applied to the end effector 54 exceeds a threshold magnitude, the end effector 54 may be moved to reduce the force vector to a desired magnitude. The desired magnitude may comprise a zero value or a non-zero value depending upon the amount of force needed to seal the end effector 54 against the fuel stem 90.

[0092] Referring to FIG. 16, a boot 102 of the end effector 54 is illustrated. Further, the coordinate system CS5 is shown on the boot 102 which corresponds to a contact point of a fuel stem 90 with the boot 102. When the boot 102 mates with the fuel stem 90, a force vector FV is applied to the boot 102 with force components f_x5, f_y5 and f_z5 along the X5, Y5 and Z5 axes, respectively. It should be noted that the Y5 axis extends outwardly from the page. The undesirable force f_y5 results from an error in position for the robotic arm 34 along the J1 axis. The undesirable force f_x5 results from an error in position of the robotic arm 34 about the J3 axis. As shown, the load cells 106, 108, 110, 112 may be utilized to determine a force vector applied to the front housing 104. Thereafter, the force vector FV and the component forces f_x5 and f_y5 may be readily determined utilizing simple force vector equations well know to those skilled in the art.

[0093] The forces f_x5, f_y5 and f_z5 applied to the boot 102 create (i) an undesirable force f₁ on the carriage 40 along the J1 axis and (ii) an undesirable torque ₃ on the pitch arm 52 about the J3 axis. Referring to FIG. 17, the force f₁ and the torque ₃ may be calculated utilizing a Jacobian Transpose equation. In particular, the Jacobian Transpose equation utilizes (i) the geometric DH constants g, h, k of the robotic arm 34 (see FIG. 7), (ii) the current joint values ₂, ₃, ₄, J5, and (iii) the forces f_x5, f_y5, f_z5to calculate the force f₁ and the torque ₃. The Jacobian Transpose equation may be readily determined by one skilled in the art.

[0094] The force f₁ and the torque ₃ are utilized to calculate displacement error values for the carriage 40 along the J1 axis and the pitch arm 52 about the J3 axis. Because the boot 102 is compliant, it can be modeled as a spring utilizing the spring equation:

[0095] ddisplacement of boot 102,

[0096] kthe spring constant

[0097] Thus, an error in force or torque (i.e., f₁ and ₃) is proportional to an error in displacement (i.e., displacement d) of the robotic arm 34 along the J1 and J3 axes. In a constructed embodiment, the force f₁ is input into a first PID controller (implemented in software) that calculates a displacement error value _J1 for the carriage 40 on the J1 axis. Similarly, the torque ₃ is input into a second PID controller that calculates an angular displacement error value ₃ for the pitch arm 52 about the J3 axis. As previously discussed, the values _J1, and ₃ may be utilized by the robot controller 46 to correct the position of the carriage 40 and the pitch arm 52, respectively, to reduce the undesirable forces applied to the end effector 54 during fueling of the fuel stem 90.

[0098] The robot controller 46 utilizes three types of sensor feedback to position the robotic arm 34 along the J1 axis. The types of sensor feedback include a line encoder feedback, a vision system feedback, and a force feedback. Referring to FIG. 2, the line encoder feedback includes encoder counts from the conveyer line encoder 162 that are utilized to determine a gross position of the fuel stem 90. The encoder counts are also indicative of the speed of the conveyer line 28. Referring to FIG. 11, the vision system feedback includes a three-dimensional position P_e of the fuel stem 90 with respect to the coordinate system CS1. Finally, the force feedback includes a measured force exerted on the end effector 54 by the fuel stem 90 during fueling of the fuel stem 90.

[0099] Referring to FIGS. 10A and 10D, the robot controller 46 moves the robotic arm 34 along the J1 axis using the tracking equation (9). The equation (9) utilizes a STEM_DISPLACEMENT variable that corresponds to the distance that the fuel stem 90 (and the vehicle 26) have traveled along conveyer axis 170 since the vehicle 26 passed the light sensor trigger position. The STEM_DISPLACEMENT variable is calculated using the following equation:

[0100] first encoder countencoder count from conveyer line encoder 162 when the V_TRIG signal is generated;

[0101] CFconversion factor for converting an encoder count to a distance along the conveyer line axis 170.

[0102] Accordingly, the STEM_DISPLACEMENT variable is updated to a new value whenever the current encoder count is updated by the conveyer line encoder 162.

[0103] The equation (9) also utilizes the DESIRED_CARRIAGE_TO_STEMY constant obtained from the VID record. The DESIRED_CARRIAGE_TO_STEMY constant represents the desired distance along the J1 axis from the origin of coordinate system CS1 (on the carriage 40) to a point P1 directly across from the fuel stem 90to allow the end effector 54 to mate with the fuel stem 90.

[0104] The equation (9) also utilizes the HOME_TO_STEM constant obtained from the VID record. The HOME_TO_STEM constant represents the distance along the J1 axis from the origin of coordinate system CS0 to a point P0 directly across from the fuel stem 90when the vehicle 26 passes the light sensor trigger position.

[0105] The equation (9) also utilizes a VISION_TRACKING_ERROR variable. As previously discussed, the vision system 164 calculates a three-dimensional position P_e of the fuel stem 90 with respect to a coordinate system CS1. Further, the vision system 164 transfers the position P_e to the robot controller 46 using the following variables: CARRIAGE_TO_STEMX, CARRIAGE_TO_STEMY, CARRIAGE_TO_STEMZ. Because the CARRIAGE_TO_STEMY variable represents a current distance from the carriage 40 (and coordinate system CS1) to the fuel stem 90 along the J1 axis, the variable can be used to calculate the VISION_TRACKING_ERROR. In particular, the VISION_TRACKING_ERROR variable may be calculated using the following equation:

[0106] Referring to FIG. 10A, the carriage 40 (and coordinate system CS1) is perfectly positioned along the J1 axis to allow engagement of the end effector 54 and the fuel stem 90. As shown, the CARRIAGE_TO_STEMY distance returned by the vision system 164 is equal to the DESIRED_CARRIAGE_TO_STEMY. Accordingly, the VISION_TRACKING_ERROR value equals a zero value.

[0107] Referring to FIG. 10B, the carriage 40 (and coordinate system CS1) are positioned too far in front of the fuel stem 90 along the J1 axis to allow engagement of the end effector 54 and the fuel stem 90. Accordingly, the VISION_TRACKING_ERROR is equal to a negative number (i.e., a negative value along the Y1 axis) which decreases the COMMANDED_J1_POSITION value. In response, the carriage 40 and the robotic arm 34 are moved to a position along the J1 axis corresponding to the new COMMANDED_J1_POSITION.

[0108] Referring to FIG. 10C, the carriage 40 (and coordinate system CS1) are positioned too far behind a desired position on the J1 axis to allow engagement of the end effector 54 and the fuel stem 90. Accordingly, the VISION_TRACKING_ERROR is equal to a positive number (i.e., a positive value along the Y1 axis) which increases the COMMANDED_J1_POSITION variable. In response, the carriage 40 and the robotic arm 34 are moved to a position along the J1 axis corresponding to the new COMMANDED_J1_POSITION.

[0109] The equation (9) also utilizes a FORCE_FEEDBACK_ERROR variable. As previously discussed, the robot controller 46 monitors a force exerted on the robotic arm 34 by the fuel stem 90 during the fueling of the stem 90. Further, the controller 46 calculates a displacement error value _J1 to correct the position of the carriage 40 along the J1 axis responsive to the force. Note that the force control also compensates for misalignments in the vertical directional using the J3 axis. When the boot 102 is mated with the fuel stem 90, the FORCE FEEDBACK_ERROR variable is set equal to the calculated displacement error value _J1. When the boot 102 is not mated with the fuel stem 90, the FORCE_FEEDBACK_ERROR is set equal to a zero value.

[0110] The modes of operation of the fueling system 24 in accordance with the present invention will now be discussed. Referring to FIG. 21A, the modes of operation include a power up mode 190, an auto mode 192, and a manual mode 194. The various modes of operation are implemented in software that is stored in the ROM of the robot controller 46.

[0111] During the power up mode 190, the robot controller 46 and the vision controller 174 establish communication with each other via the communication bus 172. Further, power is applied to the motor drivers 44, 72.

[0112] Next, the robot controller 46 advances to a step 191 that prompts the user of the fueling system 24 to select between the different modes of system operation. As discussed, the modes of operation include auto mode 192 or manual mode 194. The robot controller 46 also provides for the option to shut down the fueling system 24. The modes of operation may be chosen using a GUI and a pointing device (not shown), a push-button based operator panel (not shown), or similar controls on the supervisory PLC 178.

[0113] When the user selects the auto mode 192 of operation, the software modules comprising the auto mode 192 are executed to implement the method for fueling a vehicle 26 in accordance with the present invention. Referring to FIG. 21B, the auto mode 192 includes a stow module 196, an idle module 198, a gross position module 200, a track fuel stem module 202, an insert module 204, a fuel module 206, a purge fuel module 208, and an extract module 210.

[0114] Referring to FIGS. 2 and 4, the stow module 196 performs steps to move the carriage 40 to a home position (i.e., origin of the coordinate system CS0) along the J1 axis. Further, the joints 66, 82 are moved to predetermined stowed positions.

[0115] Referring to FIG. 21C, the idle module 198 includes a step 212 of determining if a vehicle trigger signal V_TRIG was received. Referring to FIG. 10A, when the vehicle 26 passes the light sensor trigger position, a continuously emitted light beam is not detected by the light sensor 160. In response, the light sensor 160 generates the signal V_TRIG which is received by the robot controller 46. Referring to FIG. 21C, if the signal V_TRIG is received, the module 198 advances to a step 214 which stores a first encoder count from the conveyer line encoder 162. Thereafter, the module 198 is exited and the auto mode 192 advances to the gross position module 200. Alternately, if the signal V_TRIG is not received, the module 198 iteratively performs the step 212 until a vehicle 26 is detected by the light sensor 160.

[0116] Referring to FIG. 21D, the gross position module 200 includes a step 216 of requesting a VID record from the vision controller 174. As previously discussed, the VID record contains vehicle dependent information for tracking the fuel stem 90. In a preferred embodiment, unit vector a_x, a_y, a_z is obtained from the VID record, and is then used to calculate the optimum J2 and J4 angles. The J2 angle and J4 hard stop are adjusting during the gross position module 200. In an alternate embodiment the J2 and J4 angles are automatically calculated but are manually adjusted off line prior to automatic operation.

[0117] The module 200 further includes a step 218 which iteratively calculates a gross position of the fuel stem 90 with respect to the J1 axis. Referring to FIG. 10, the step 218 utilizes the tracking equation (9) to calculate the COMMANDED_J1_POSITION which also represents the gross position of the fuel stem 90. Further, both the VISION_TRACKING_ERROR and the FORCE_FEEDBACK_ERROR are equal to a zero value during the step 218.

[0118] Referring to FIG. 21D, the module 200 further includes a step 220 of moving the robotic arm 34 along the J1 axis to position the end effector 54 proximate to the gross position of the fuel stem 90. Further, the module 200 moves the robotic arm 34 (and the end effector 54) at a speed substantially equal to the speed of the conveyer line 28. During the step 200, the cameras 182, 184 on the robotic arm 34 should be positioned along J1 such that the field of view of the cameras 182, 184 covers the fuel stem 90.

[0119] The module 200 further includes a step 222 of triggering the frame grabber 180. In particular, the robot controller 46 generates a signal V_TR that causes the frame grabber 180 to transfer first and second digital images of the fuel stem 90 from the cameras 182, 184 to the vision controller 174.

[0120] The module 200 further includes a step 224 of requesting a three-dimensional position P_e of the fuel stem 90 from the vision controller 174. As previously discussed, the vision controller 174 performs template matching on the first and second digital images and calculates first and second correlation scores, respectively. Further, the controller 174 returns the position P_e if the first and second correlation scores are above a threshold correlation score and a triangulation error of the position P_e is below a predetermined triangulation error.

[0121] The module 200 further includes a step 226 which determines whether a three-dimensional position P_e was received from the vision controller 174. If the position P_e was received, the module 200 is exited and the auto mode 192 advances to the track fuel stem module 202. Alternately, if the position P_e was not received, the module 200 returns to step 222.

[0122] Referring to FIG. 21B, the auto mode 192 advances to the track fuel stem module 202 after module 200. Referring to FIG. 21E, the module 202 includes a step 228 of triggering the frame grabber 180 to obtain first and second digital images of the fuel stem 90 generated by the cameras 182, 184, respectively.

[0123] The module 202 further includes a step 230 that requests a three-dimensional position P_e of the fuel stem 90 from the vision controller 174.

[0124] The module 202 further includes a step 232 which determines whether a three-dimensional position P_e of the fuel stem 90 was received by the robot controller 46. If the position P_e was received, the module 202 advances to the step 234. Otherwise, the module 202 returns to the step 228.

[0125] The module 202 further includes a step 234 of calculating the VISION_TRACKING_ERROR. As previously discussed, the VISION_TRACKING_ERROR is utilized in the equation (9) to more accurately position the robotic arm 34 along the J1 axis relative to the fuel stem 90. It should be understood that the VISION_TRACKING_ERROR is iteratively calculated in the track fuel stem module 202 and the insert module 204 so long as the fuel stem 90 is viewed by both cameras 182, 184.

[0126] The module 202 further includes a step 236 which moves the robotic arm 34 along the J1 axis proximate the fuel stem 90 responsive to the COMMANDED_J1_POSITION.

[0127] Referring to FIG. 21B, the auto mode 192 advances to the insert module 204 after the module 202. Referring to FIG. 21F, the module 204 includes the steps 238, 240, 242, 244, 246. The step 238 calculates joint variables ₃ and J5 using the inverse kinematic model previously discussed. The step 240 moves the pitch arm 52 about the J3 axis to the calculated angle ₃. The step 242 moves the end effector 54 about the J4 axis to a predetermined angle ₄.

[0128] The step 244 moves the boot 102 along the J5 axis a distance J5 to allow the boot 102 to mate with the fuel stem 90. Referring to FIG. 20, during the steps 238, 240, 242, the end effector 54 is positioned to allow the step 244 to move the end effector 54 along a docking line toward the fuel stem 90. As shown, the docking line extends between a point P_b on the nozzle axis 245 to the fuel stem point P_e. Further, the docking line forms an approach angle _D with respect to the stem axis 243. Finally, the step 246 moves the fuel hose 100 into the fuel stem 90.

[0129] Referring to FIG. 21B, the auto mode 192 advances to the fuel module 206 after the module 204. Referring to FIG. 21G, the fuel module 206 includes the steps 248, 250, 252. The step 248 calculates the FORCE_FEEDBACK_ERROR which is utilized by the controller 46 to calculate the COMMANDED_J1_POSITION of the carriage 40 (and the robotic arm 34) along the J1 axis. The step 250 opens a fuel valve 116 in the end effector 54 to supply fuel to the fuel stem 90. The step 252 closes the fuel valve 116 after a predetermined amount of fuel is pumped into the fuel stem 90.

[0130] Referring to FIG. 21B, the auto mode 192 advances to the purge fuel module 208 after the module 206. Referring to FIG. 8, the purge fuel module 208 performs steps to apply air pressure to the check valves 140, 142 to force any residual fuel in the manifold 98 and the fuel hose 100 into the fuel stem 90.

[0131] The auto mode 192 advances to the extract module 210 after the module 208 which extracts the fuel hose 100 from the fuel stem 90 and moves the robotic arm 34 to a predetermined retract position.

[0132] Referring to FIG. 21B, the auto mode 192 after exiting module 210 advances to a step 254 which determines if another vehicle 26 has passed the light sensor trigger position. If another vehicle 26 is detected, the auto mode 192 advances to the gross position module 200. Otherwise, the auto mode 192 advances to the stow module 196.

[0133] There is a self-test operating mode during which the robot controller 46 places the robotic arm 34 such that the two robot tip markers TM0 and TM1 as shown in FIG. 20 are within the field of view of the cameras 182, 184. The vision controller 174 then acquires a first digital image and a second digital image from the cameras 182, 184, and the frame grabber 180, and locates the first and second image locations of the tip marker 0 TM0 and of the tip marker 1 TM1. Using a triangulation method already described for the localization of the fuel stem point, it calculates the 3D positions of both tip markers with respect to CS2. This information may be used by the robot controller 46 to periodically update the kinematic parameters which vary due to dimensional instability of the elastomeric boot hose 103.

[0134] Pneumatic cylinder 114 is provided to extend the filler hose 100 through the boot 102 in order to pump fuel into the filler neck from a point below the no-lead insert. When the filler hose 100 is retracted, the cylinder 114 is prevented from reaching its full stroke due to the fact that the tip of the filler hose is too large to fit through the boot 102. Should the tip of the hose 100 be lost, the cylinder 114 can be fully retracted. This configuration allows the robot to sense if the hose 100 has been lost or damaged by simply using two limit switches on the cylinder 114 retract stroke, one of which is triggered when the hose 100 is retracted far enough to reach the boot, and one of which is triggered when the cylinder 114 is fully retracted.

[0135] The inventive fueling system 24 and the method related thereto represent a significant improvement over conventional fueling systems and methods. In particular, the inventive fueling system 24 allows the automatic fueling of an automotive vehicle 26 without the need for an operator. Thus, the inventive fueling system 24 and method result in labor savings. Further, the fueling system 24 can move an end effector 54 to mate with the fuel stem 90 while the vehicle 26 is moving. Thus, the vehicle 26 can be fueled without stopping the conveyer line 28 resulting in manufacturing cost savings and increased assembly line efficiency.

[0136] While the invention has been particularly shown and described with reference to the preferred embodiments thereof, it is well understood by those skilled in the art that various changes and modifications can be made in the invention without departing from the spirit and the scope of the invention.