Adjustment of vehicle-implement trajectories to compensate for lateral implement offset

[0002] This application claims the benefit of the following application:

[0003] U.S. Provisional Patent Application No. 60/249,313, entitled, “Method for the Adjustment of Vehicle-implement Trajectories to Compensate for Lateral Implement Offset,” filed Nov. 15, 2000, naming Michael O'Connor and Thomas Bell as inventors, and under an obligation of assignment to IntegriNautics, Inc. of Menlo Park, Calif.

[0004] U.S. Provisional Patent Applications No. 60/249,108 is incorporated by reference herein.

[0005] The invention generally relates to operator control of autonomous heavy machinery and specifically to methods for human operators to use autonomous vehicles in agriculture, construction, and open-pit mining.

[0006] A significant amount of work has been done to improve the efficiency of heavy machinery in areas such as agriculture, construction, and open-pit mining. In particular, some efforts strive to improve effectiveness of the equipment by providing very accurate position information to the driver or even automatically steering the vehicle.

[0007] For processes that require an implement attached to the vehicle, the accuracy of the combined vehicle-implement pair may be insufficient for the overall process, even though the accuracy of the automatically controlled (or sensor-assisted) vehicle would otherwise be adequate. Prior art does not go beyond control of the vehicle to address discrepancies between the accuracy of the vehicle alone versus the accuracy of the vehicle-implement pair.

[0008] When the goal of the overall system is to accurately control the path of the implement (rather than the vehicle), there are at least two approaches that can be taken: non-co-located sensing and control, on the one hand, and co-located sensing and control, on the other.

[0009] Regarding non-co-located sensing and control, the sensor is not located on the body to be controlled. In such a system, the sensor is on the vehicle, but the object to be controlled is the implement—typically at its point of contact with the ground. The system has a very accurate model of the implement and how it is connected to the vehicle. The position of the implement is inferred from the precisely measured position of the vehicle by a transformation matrix and, possibly, a dynamic model and additional intermediary sensors at the hitch.

[0010] The advantages of this method include avoiding the additional costly sensors on the implement and avoiding the relative complexity of vehicle control in the co-located method below. A disadvantage is that this technique may not provide sufficient accuracy at the point of interest due to magnification of sensor error by inaccuracies in the model.

[0011] In a co-located sensing and control system, the sensor is located on the body to be controlled. In this case, the sensor could be a position sensor located on the implement. An advantage of this method is that the error signal used in closed-loop feedback control is directly measured at the point of interest and can be more accurate than non-co-located control. Disadvantages are that it requires additional sensors on the implement, a more elaborate controller on the vehicle, and possibly a different controller for each type of implement.

[0012] The following patents address aspects such as GPS position sensing for machine control and path planning, but do not discuss control of implements nor ways to calibrate the implement:

[0013] Parkinson et al., U.S. Pat. No. 6,052,647, “Method and System for Automatic Control of Vehicles Based on Carrier Phase Differential GPS,” describes an auto control system for a land vehicle based on carrier phase differential GPS for position and heading sensing.

[0014] Gudat et al., U.S. Pat. No. 5,646,843, “Apparatus and Method for Surface Based Vehicle Control System,” and Gudat et al., U.S. Pat. No. 5,615,116, “Apparatus and Method for Autonomous Vehicle Navigation Using Path Data,” discuss auto control of a land vehicle with distinct operating modes and path following. Simple methods for switching between autonomous and manual modes are described.

[0015] Brandt et al., U.S. Pat. No. 6,085,130, “Method and Apparatus for Selecting a Transition Scheme for Use in Transitioning a Mobile Machine from a First Path to a Second Path” relates to the transition from one path to another path, presumably under complete autonomy.

[0016] Zimmerman et al., U.S. Pat. No. 6,041,870, “Method and Apparatus for Steering a Dozing Machine,” describes actuation of a dozer blade to laterally adjust the centerline of the vehicle.

[0017] Keller et al., U.S. Pat. No. 5,987,383, “Form Line Following Guidance System,” covers a method for automatically generating paths.

[0018] Rao, U.S. Pat. No. 5,684,696, “System and Method for Enabling an Autonomous Vehicle to Track a Desired Path,” describes generation of a desired trajectory based on current location and desired trajectory.

[0019] Sarangapani et al., U.S. Pat. No. 6,064,926, “Method and Apparatus for Determining an Alternate Path in Response to Detection of an Obstacle,” Sarangapani, U.S. Pat. No. 6,173,215, “Method for Determining a Desired Response to Detection of an Obstacle,” and Sarangapani, U.S. Pat. No. 5,752,207, “Method and Apparatus for Determining a Path for a Machine between a Predetermined Route and a Final Position,” all describe automatic path planning and obstacle avoidance.

[0020] Motz et al., U.S. Pat. No. 6,236,924, “System and Method for Planning the Operations of an Agricultural Machine in a Field,” describes a path planning method.

[0021] The following patents pertain to vehicle control:

[0022] Motz et al., U.S. Pat. No. 6,205,381, “System and Method for Planning the Operations of an Agricultural Machine in a Field,” relates to coordination of multiple vehicles under automatic control.

[0023] Staub et al., U.S. Pat. No. 6,236,916, “Autoguidance System and Method for an Agricultural Machine,” relates to general auto control of farm tractors.

[0024] Gharsalli et al., U.S. Pat. No. 6,062,317, “Method and Apparatus for Controlling the Direction of Travel of an Earthworking Machine,” relates to control systems for construction scraper.

[0025] Sin et al., U.S. Pat. No. 5,657,226, “System and Method for Causing an Autonomous Vehicle to Track a Path,” relates to general auto control of farm tractor.

[0026] The following patents address aspects of implement control, but do not discuss ways to adjust the implement based on observed performance. These patents also deal with actuated implements only:

[0027] Gharsalli et al., U.S. Pat. No. 6,062,317, “Method and Apparatus for Controlling the Direction of Travel of an Earthworking Machine,” discusses use of the implement to affect the steering on an earth-grader type of vehicle.

[0028] Gharsalli et al., U.S. Pat. No. 6,062,317, “Method and Apparatus for Controlling the Direction of Travel of an Earthworking Machine,” discusses methods for planning operations of an agricultural machine in a field. Information about a work implement (a model) is provided to the system, but no provision to calibrate based on performance is discussed.

[0029] Lombardi, U.S. Pat. No. 6,129,155, “Method and Apparatus for Controlling a Work Implement Having Multiple Degrees of Freedom,” Bailey et al., U.S. Pat. No. 5,987,371, “Apparatus and Method for Determining the Position of a Point on a Work Implement Attached to and Movable Relative to a Mobile Machine,” Skinner et al., U.S. Pat. No. 6,185,493, “Method and Apparatus for Controlling an Implement of a Work Machine,” and Wiechman, U.S. Pat. No. 6,099,236, “Apparatus for Controlling Movement of an Implement Relative to a Frame of a Work Machine,” all discuss control of implements using hydraulic actuators and linear and rotary sensors to infer position. No provision to calibrate implement based on performance is discussed.

[0030] Krieg et al., U.S. Pat. No. 6,286,606, “Method and Apparatus for Controlling a Work Implement,” and Meduna, et al., U.S. Pat. No. 6,295,746, “Method and Apparatus for Controlling Movement of a Work Implement,” discuss changing control modes on a grader-type implement. No provision is made to calibrate the implement based on observed performance.

[0031] Herein is taught a system including a mobile vehicle, an implement, a position sensor, a controller, a user interface (a touch-screen monitor, video monitor or keypad, as examples), a software program to compute a calibration trajectory, and a steering system for steering the vehicle to the desired or adjusted trajectory based on the error between the vehicle's desired and actual positions.

[0032] In one embodiment, non-co-located sensing and control combine with a calibration procedure to help eliminate model error. Another embodiment uses co-located sensing (but not control) to calibrate the model.

[0033] FIG. #_1 illustrates a system incorporating an embodiment of the invention.

[0034] FIG. 2 illustrates a calibration trajectory.

[0035] FIG. 3 illustrates a post-calibration trajectory.

[0036] FIG. 4 illustrates intermediary rows in cross-section.

[0037] FIG. 5 illustrates another embodiment of the invention.

[0038] FIG. 6 illustrates yet another embodiment of the invention.

[0039] (Some of the drawings are not to scale.)

[0040] FIG. #_1 illustrates a system (1) incorporating an embodiment of the invention. The system (1) may include a farm tractor (101) and a local global positioning system (GPS) reference station (510).

[0041] The system (1) uses non-co-located sensing and control combined with a calibration procedure to eliminate model error. The accuracy of this method is limited by the accuracy of the manual measurement of row widths, described below. The system compensates for model error by adjusting the nominal trajectory of the vehicle, causing the implement to go precisely where it is intended to go.

[0042] The farm tractor (101) may have position and heading sensors (102), antennae (103-i), antenna (104), a controller (105), a steering system (106), a touch-screen monitor (107) and a communications link (108).

[0043] The GPS reference station (510) may include an antenna (111 ) and a transmitter (112).

[0044] In this example, the position and heading sensors (102) are based on the Global Positioning System (GPS). Using the local GPS reference station (510) and differential carrier phase methods, the sensors (102) sense the tractor position to inch-level accuracy. (The station (510) and the methods are well known in the art.)

[0045] The reference station (510) receives GPS signals at its location via its antenna (111) and sends them via radio modem (112) to the tractor (101) via the tractor's antenna (104). The tractor heading is measured to sub-degree accuracy using three GPS antennas (103-1:103-3), using yet another GPS technique well known in the art.

[0046] The controller (105) may be a computer that uses the GPS position and heading information to autonomously steer the tractor (101). The controller (105) generates a trajectory for the implement-offset calibration procedure. The trajectory may be a double U-turn pattern, with equal spacing between the passes, as shown by (120). The controller (105) runs software that generates a steering signal (108), which is a function of the error between the measured position and heading, on the one hand, and the desired trajectory, on the other. The steering signal (108) is applied to the vehicle's steering system (106), typically either a hydraulic actuator or another controller.

[0047] The touch-screen monitor (107) displays a graphical user interface (GUI) (109) controlled by software on the controller (105). The GUI (109) may feature icons (120, 121) representing the tractor and the desired trajectory to provide visual feedback of the system performance to the user. The graphical user interface may also feature a set of virtual buttons (122) that allow the user to enter row spacing measurements into the system.

[0048] The vehicle (101) tows or pushes an implement (113) that is attached by a hitch (114). The centerline of the implement (115) is not necessarily aligned with the centerline of the vehicle (117) due to the hitch or other model errors, creating an undesired implement offset (116). A nominal measure of the implement offset (116) may be provided by the manufacturer of the implement and entered into the system via database or hand, The implement (113) may feature equally distributed blades (118-i) for working the ground.

[0049] FIG. 5 illustrates another embodiment of the invention. This embodiment employs co-located sensing to calibrate the model, as well as non-co-located control of the vehicle-implement pair. The accuracy of the alignment of the sensor on the implement to the point of interest (one of the plowing blades, for example) limits the accuracy of this method.

[0050] This embodiment differs from the preceding embodiment in at least two ways: One, there is an extra GPS antenna (503-4) that is placed on the implement (513). Its signal is fed into the GPS sensor (502). Second, the user interface does not require a way for the user to enter measured row widths. The location of the antenna (503-4) on the implement (513) may be provided through a database or may be provided by the manufacturer of the implement (513).

[0051] With reference to FIG. 2, a tractor (201) tows or pushes an implement (202). The implement centerline is laterally offset from the tractor centerline by an unknown amount (203).

[0052] The tractor (201) may be commanded to follow a double U-turn trajectory (204) that is equally spaced between passes (209-i). The implement (202) forms a set of equally spaced rows (206). The pass spacing (209-i) is wider than the implement (202), so as to form intermediary rows (207, 208). The intermediary rows (207, 208) are narrower or wider than the implement-formed rows (206) due to the implement offset.

[0053] FIG. 4 illustrates the intermediary rows (401, 402) in cross-section. The width (403, 404) of an intermediary row (401, 402) is measured from top edge to top edge of an adjacent row. This may be done manually—with a tape measure, for example.

[0054] The measured distances (403, 404) are entered into the tractor's controller through the user interface as “ROW MEASUREMENT 1” and “ROW MEASUREMENT 2”. The measurements are used to compute a lateral offset to be applied to future trajectories. The offset is computed as follows: The difference between the two measurements, divided by four, gives the lateral offset. The sign of the result indicates whether the offset should be applied as a left adjustment or a right adjustment.

[0055] The tractor may be again commanded to follow a double U-turn trajectory (304) with lateral offset applied. The spacing (309-1, 309-2) between passes may no longer be equal.

[0056] The intermediary rows (307, 308) are measured, as before. If row widths (307, 308) are not spaced to within acceptable tolerance, the procedure is repeated from the entering of measured distances into the controller.

[0057] In an alternate embodiment, a tractor (601) tows or pushes an implement (602). The implement centerline is laterally offset from the tractor centerline by an unknown amount (603). The location of the implement sensor with respect to the implement centerline is accurately known (604).

[0058] The tractor may be commanded to follow a straight-line trajectory (606). The controller on the tractor collects several (e.g., more than 10) simultaneous measurements of the tractor position and the implement position as the tractor moves.

[0059] The average of the differences between the tractor positions and implement positions is calculated. The lateral component of this average (605) minus the known sensor location on the implement (604) gives the implement lateral offset (603).

[0060] The embodiments are by way of example and not limitation.

[0061] Modifications to the invention as described will be readily apparent to one of ordinary skill in the art.