Dynamic Control for Manipulator

The invention is related to programmed manipulators or in a more limited embodiment referred to as industrial "robots"; and more specifically to dynamic controls for manipulators having a fast nonlinear dynamics control loop.

The development of programmed manipulators or robots is progressing at a rapidly increasing pace in the United States as elsewhere in the world. Currently, the industrial robots, such as taught by Hohn in U.S. Pat. No. 3,909,600, employ dynamic control laws which ignore the complex nonlinear dynamics of the various moving elements which result in less than optimal response. The computational bottleneck of many advanced dynamic control schemes is an algorithm for computation of the actuator torques and/or forces required to produce the desired joint accelerations, for a given set of joint velocities and angles.

The dynamics of an "n" link manipulator can be described by the following set of n nonlinear differential equations.

where

For a typical 6 link manipulator, the functions J, f and g in Equation 1 are very complex, nonlinear expressions. Because of this complexity, most proposed control schemes are based on breaking down or decoupling Equation 1 into n linear differential equations. The decoupling feedback is given by:

where

Equation 2 transforms the original system given by equation 1 into the following linear, decoupled system. ##EQU1##

The system described by Equation 3 is easily controlled by any standard control technique.

Equation 2 constantly changes when the manipulator moves, because J, f and g are dependent on the joint angles. Thus to achieve decoupling at all times, the decoupling torques T must be continuously recomputed in real time. In order to avoid introduction of undesirable dynamic effects the required frequency of computation is about 100 times per second. This is the bottleneck in any control scheme, because it involves an excessive number of floating point additions and multiplications.

Numerous schemes for computation of Equation 2 have been proposed during the last fifteen years. A comprehensive comparison of these techniques is presented by J. M. Hollerbach in his article "A Recursive Lagangian Formulation of Manipulator Dynamics and a Comparative Study of Dynamics Formulation Complexity" IEEE Transactions on Systems, Man and Cybernetics, Vol. 10, No. 11, November 1980, pp 730-736. The most elegant formulation of the problem, based on Lagrange's equation, is due to J. J. Uicker "On the Dynamic Analysis of Spatial Linkages Using 4.times.4 Matrices" Ph. D. Dissertation, Northwestern University, August 1965, and M. E. Kahn "The Near-Minimum-Time Control of Open Loop Articulated Kinematic Chains", Stanford University A. I. Memo 177 September 1969. Unfortunately, the (Uicker/Kahn) formulation is unsuitable for real-time computation, because for a typical six link manipulator it involves about 118,000 additions and multiplications per trajectory point. J. Y. S. Luh, M. M. Walker and R. P. C. Paul, in their article "On Line Computational Scheme for Mechanical Manipulators" ASME Journal of Dynamic Systems, Measurement and Control, Vol. 102, No. 2, June 1980, pp 69-76 disclosed that the execution time of Equation 2 for a six link manipulator using Lagrange's equation is 7.9 seconds. This is approximately 790 times longer than the 0.01 second goal for current systems. The fastest exact computational scheme known today is the recursive Newton-Euler technique disclosed by Luh et al in the above article. This scheme first computes the velocities and the accelerations of the manipulator links, going from the manipulator base to the last link. It then computes the torques and forces acting on the links, going from the last link back to the base. All the computations are performed in link coordinates, which significantly reduces the number of coordinate transformations required and reduces the number of arithmetic operations to 1590 for a six link manipulator. This techniques requires 0.0335 seconds to perform the necessary arithmetic operations. Hollerbach concluded that only very minor improvements can be made to the recursive Newton-Euler formulation disclosed by Luh et al. However, Hollerbach's conclusion only applies to formulations which are very general and can handle manipulators with any number of links and any configuration.

The dynamic manipulator control described herein is based on the fact that most industrial manipulators are relatively simple geometrically and have from 4 to 7 links. Its computational time, in high level languages, is about five or ten times less than that using the technique taught by Luh et al depending on the implementation. The proposed control divides the manipulator into two parts. The first three links (counting from the base) are modeled using classical (non-matrix) Lagrange's equations, with velocities computed in link coordinates. The remaining links (i.e., the wrist) are modeled using recursive Newton-Euler equations.

The invention is a control system for a multiple link mechanical manipulator or robot having a nonlinear dynamic control loop computing correction factors added to signals generated by the primary linear dynamic control loop correcting the actuator signals for the non-linear dynamics of the manipulator. The primary control comprises a central processing unit having adequate storage and computation capabilities for generating signals indicative of the desired motion of the individual links, a linear dynamic control loop for generating actuator signals controlling the actuators moving the individual manipulator links, and sensors responsive to the position and motion of the individual manipulator links for generating feedback signals to the central processing unit and the linear dynamic control loop. The improvement comprising a nonlinear dynamic control loop for correcting the linear control signals for the nonlinear dynamics of the manipulators, said nonlinear dynamics control loop characterized by a nonlinear dynamics control computer for generating correction factors for each of the actuators moving the manipulator links in response to the motion signals generated by the central processing unit and the feedback signals using Lagrangian dynamics for the first three links and recursive Newton-Euler dynamics for the remaining links, and means for summing the correction factors with the associated actuators signals to correct the actuator signals for the nonlinear dynamics of the manipulator.

The advantages of the manipulator control system are that it is more effective and efficient than the linear control systems of the prior art and the computations for the nonlinear dynamics are capable of being performed at a rate from 5 to 10 times faster than prior art nonlinear control schemes.

These and other factors will become more apparent from a reading of the detailed description of the invention in conjunction with the appended figures.

FIG. 1 is a schematic presentation of the first three links of a typical manipulator.

FIG. 2 is a perspective of the first three links of a manipulator identifying the respective joints.

FIG. 3 is a schematic presentation of a typical manipulator wrist.

FIG. 4 is a block diagram of the manipulator control system.

FIG. 5 is a block diagram showing nonlinear dynamics control in greater detail.

FIG. 6 is a flow diagram for the nonlinear dynamic control computer.

FIG. 7 is a block diagram of the nonlinear dynamics control system using two microprocessors for the nonlinear dynamics control computer.

FIG. 8 is a flow diagram showing the division of the computations between the two microprocessors in the nonlinear dynamic control loop of FIG. 7.

The dynamic manipulator control is based on division of the manipulator into two parts. The first part consists of the first three links (counting from the base), and the second part consists of all the remaining links. Referring to FIG. 1 there is shown schematically the first three links of a typical robot arm 10 comparable to the robot arm disclosed by Hohn in U.S. Pat. No. 3,909,600. The coordinate systems of the links are assigned using the standard convention as taught by Hollerbach cited above, except for link 3, which always has the origin of its coordinates at its end and the (z) axis along its longitudinal axis. In particular, the first three links, Links 1, 2 and 3 are supported from a base 18 as shown. The first link, Link 1 is mounted for rotation from the Base 18 about a vertical axis (z.sub.0), the second link, Link 2, is mounted at the end of Link 1 and is rotatable about a first horizontal axis (z.sub.1) and the third link, Link 3 is mounted at the end of Link 2 and is rotatable about a second horizontal axis (z.sub.2). The z axis (z.sub.3) of Link 3 is along its longitudinal axis as shown. The x and y axes complete the right hand system, with x in the vertical plane. This proposed model provides a standard interface between the two parts of robots' arm 10, by means of the forces and moments acting on the end of Link 3 to which the wrist is attached. Thus, any first-three-link arm and any type of wrist will always match. The standard convention previously described does not allow such modularity, because coordinates of Link 3 are determined by the type and orientation of the joint at the end of Link 3. For convenience the joints between the links are identified on FIG. 2. Conventionally there is the Waist Joint 11, the Shoulder Joint 12, the Elbow Joint 13, the Wrist Joint 14 etc.

A typical three degree of freedom Wrist 20 is shown in FIG. 3. Its reference frame is the system of coordinates of Link 3. As shown, Wrist 20 has a fourth link, Link 4 mounted to the end of Link 3 and is rotatable about the axis (z.sub.3) of the Link 3, a fifth link, Link 5 attached to the end of Link 4 and is rotatable about axis (Z.sub.4) of Link 4 and Link 6 is rotatable about axis (Z.sub.5) of Link 5.

The actuator torques (forces) of the Wrist 20 required by the actuators moving the various links are computed using recursive Newton-Euler equations as taught by Luh et al cited above. The computations are referred to the base coordinates through Link 3 coordinates, the velocities and accelerations of which are simple functions of the parameters and variables of the first three links. The computation starts with the Link 4, and propagates the velocities and the accelerations out to the last link, and then evaluates the actuator torques (forces) starting from the last link back to actuator moving Link 4. Because of the simplicity of most manipulator wrists, the equations can be written out explicitly for each link. Such a formulation uses less than 40% of the arithmetic operations required by the general recursive Newton-Euler algorithm. Further reductions in the computational time for the nonlinear dynamics are due to program execution optimization. Because the first-three-link models and the wrist models are very simple, these models appear in symbolic form in the program. This allows for the grouping of similar terms and other operation saving manipulation well known to those skilled in the art.

In addition to the actuator torques, the nonlinear dynamics control also computes the forces and moments required at the end of Link 3, in order to support the Wrist 20. These latter forces and moments are always computed in the standard coordinates of Link 3.

The second part of the nonlinear dynamics control computes the torques (forces) of actuators for Links 1, 2, and 3, using Lagrangian dynamics. These torques (forces) are primarily due to four effects:

a. Translational inertia (T.sub.t) of Links 1, 2, and 3.

b. Rotational inertia (T.sub.r) of Links 1, 2, and 3.

c. Gravity forces (T.sub.g) acting on Links 1, 2, and 3

d. Forces and torques (T.sub.w) required to support the wrist 20.

The total torque (T.sub.i) required from actuator i is given by

The first two components are evaluated as follows: ##EQU2##

The contribution due to gravity, T.sub.g, and the contribution due to the wrist, T.sub.w, can be written down by inspection.

The above version of Lagrange's equation requires less than 40% of the arithmetic operations required by method taught by Luh et al cited above. Again further savings of computational time can be realized by program execution optimization. The equations for the first three links, Links 1, 2 and 3, appear in symbolic form in the program and are conveniently separated into terms corresponding to inertia forces, centripetal forces, Coriolis forces and gravity forces. The Wrist 20 is represented in the computations for the first three links by the forces and moments it applies to the Link 3, in numerical form. The use of Lagrangian dynamics for the first three links is advantageous for controlling of manipulators which are not open kinematic chains. The developed control easily accommodates for various types of manipulator configurations, because the first-three-link model is derived directly from basic principles of dynamics and are typical to most robot arms.

The complete algorithm can be summarized by the following seven steps using the robot arm as disclosed by Hohn.

1. Compute the velocities and accelerations V.sub.3, w.sub.3, V.sub.3 and w.sub.3 acting on Wrist Joint 14, in Link 3 coordinates using Lagrangian dynamics as described by Luh et al cited above. The equations for the velocities and accelerations are as follows:

2. Compute the torques (forces) for the actuators for Links 4, 5 . . . n and the forces and torques required to support the Wrist 20 at the end of Link 3 (Link 3 and Link 4 interface) using the recursive Newton-Euler algorithm described by Hollerbach cited above. A summary of this algorithm, presented in the Section entitled "Recursive Newton-Euler Dynamics", is as follows: ##EQU4## where: w.sub.i =the angular velocity of Link i

w.sub.i =the angular acceleration of Link i

F.sub.i =total external force or Link i

N.sub.i =total external torque on Link i

T.sub.i =Link i actuator torque (force)

I.sub.i =inertial tensor of Link i about its center of mass

f.sub.i =force exerted on Link i by Link (i-1)

n.sub.i =moment exerted on Link i by Link (i-1)

p.sub.i =vector from the base coordinate origin to the joint i coordinate orgin.

p.sub.i =vector from coordinate origin (i-1) to origin i

r.sub.i =vector from the base coordinate origin to Link i center of mass

r.sub.i =vector from coordinate origin i to Link i center of mass

M.sub.i =mass of Link i

z=vector along the axis of joint i+1

3. Using Equation 5 compute T.sub.ti, the torques (forces) required from the actuators for Links 1, 2 and 3 due to translational inertia of Links 1, 2, and 3.

4. Using Equation 6 compute T.sub.ri, the torques (forces) required from the actuators for Links 1, 2 and 3 due to rotational inertia of Links 1, 2, and 3.

5. Compute T.sub.gi, the torques (forces) required from the actuators for Links 1, 2 and 3 due to gravity action on Links 1, 2 and 3. ##EQU5## where s.sub.i is the location of the center of mass of link i in Link i coordinates.

6. Compute T.sub.wi, the torques (forces) required from the actuators for Joints 11, 12 and 13 to balance the force and moments the wrist applies on the end of link 3.

where F.sub.x, F.sub.y, F.sub.z, T.sub.x, T.sub.y, T.sub.z are the interface forces and moments between Link 3 and Link 4.

7. Using Equation 4 compute T.sub.i, the total torques (forces) required from the actuators for Links 1, 2 and 3.

FIG. 4 is a detailed block diagram for a manipulator control system. The kinematic portion of the system comprises a Central Processing Unit (CPU) 102, a Control Console 104, a Program Store 106 and a Data Store 108. The Control Console 104 provides the operator with the necessary controls and displays to generally operate a Manipulator 100 as is known in the art. The system will also include other conventional peripheral devices (not shown) such as a machine input/output device concerned with machine oriented functions such as switches, solenoids and other similar devices associated with the Manipulator 100 and a process input/output device for integrating or coordinating the operating of the Manipulator with other machines.

The operational program of the Central Processing Unit 102 is stored in the Program Store 106 while the data used by the Central Processing Unit 102 in the execution of its program is stored in Data Store 108. The data stored in Data Store 108 will include constants used in the calculations, feedback data received from Sensors 110 detecting the actual position, velocity, and or, acceleration of the various links of the Manipulator 100 as well as any partial solutions generated in the execution of the program as may be required.

The Central Processing Unit 102, in response to the inputs received from the Control Console 104 will execute the program stored in the Program Store 106 using the data stored in Data Store 108, and generate electrical signals indicative of the desired motion of the individual links in the Manipulator 100.

These desired motion signals are converted into actual motion of the manipulator by a conventional dynamic control loop comprising Sum Amplifier Means 112, Servo Amplifiers 114 and Power Amplifier 116. Sum Amplifier 112 subtracts the feedback signals indicative of the actual position, velocity and acceleration of each link from the associated motion signal to produce a difference signal for each manipulator link. The Servo Amplifiers 114 convert the individual difference signals to dynamic actuator signals indicative of the torques or forces required to move the associated links of the manipulator along the desired path. The individual servo amplifiers execute the linear control laws for manipulator motion and include dynamic compensators for achieving fast and stable response.

Power Amplifiers 116 amplify the actuator signals to power levels required by the individual actuators to produce the desired torque or force. The outputs from Power Amplifiers 116 are received by the individual actuators which move the individual links of the Manipulator 100 with the desired motion. When the actuators in the Manipulator are electric motors, the output of Power Amplifiers 116 will be electric signals actuating the electric motors. Alternatively if the manipulator actuators are hydraulic or pneumatic devices, the output of the Power Amplifiers 116 will be hydraulic or pneumatic signals as is known in the art.

The control system as described above represents the current state of the art of manipulator control systems. As previously indicated, the control law used in these systems is based on ignoring the complex (nonlinear) dynamics of the various moving elements. Currently this approach is satisfactory since the complex dynamics of the various elements may be adequately compensated for by programing the control law for a worst case situation and using brute force to impart to each link the desired motion. This approach is not the most effective or efficient way to control a manipulator and results in less than optimal response.

The dynamic response of the Manipulator 100 can be improved by providing a second dynamic control loop which corrects the torque signals output from the Servo Amplifiers 114 for the complex (nonlinear) dynamics of the various moving elements. Refering back to FIG. 4 this second dynamic control loop comprises a Nonlinear Dynamics Control Computer 118 for computing a correction factor for each link in the manipulator and a Sum Amplifier Means 120 for summing the computed correction factors with the corresponding torque signals output from the Servo Amplifiers 114. The correction factors correct each of the actuator signals for the complex dynamics of the individual links previously ignored. The Nonlinear Dynamics Control Computer 118 receives the desired motion signals generated by the Central Processing Unit 102 and the feedback signals generated by Sensors 110. The Nonlinear Dynamics Control Computer 118 then computes the correction factors for the first three links using Lagrange's equations and the correction factors for the remaining links using the recursive Newton-Euler algorithm as previously described. The computed correction factor for each link is then summed with the associated torque signals in Sum Amplifier Means 120. Sum Amplifier Means 120 comprises a plurality of sum amplifiers, one associated with each of the servo amplifiers of Servo Amplifiers 114 as shall be discussed relative to FIG. 5.

Referring now to FIG. 5, there is shown in greater detail, the second dynamic control loop for a six (6) link manipulator. The Nonlinear Dynamics Control Computer 118 may be any commercially available microprocessor, such as the 16 bit Intel 8086 or Motorola 68000 microprocessor or any of the newer 32 bit microprocessors recently introduced or currently under development. The Nonlinear Dynamics Control Computer 118 receives directly the desired motion signals from the Central Processing Unit 102 and the feedback signals from Sensors 110 individually identified as Sensors 110-1 through 110-6 associated with the individual Links 1 through 6. Each link may have one or more sensors, generating signals indicative of the links current position, velocity, and/or acceleration as is known in the art.

The Nonlinear Dynamic Control Computer 118 computes the correction factors for the actuator signals output from the individual Servo Amplifiers 114-1 through 114-6 in response to the desired motion signals and the feedback signals. The correction factor signals output from the Dynamic Control Computer 118 are individually summed with the associated actuator signals generated by Servo Amplifiers 114-1 through 114-6 individual Sum Amplifiers 120-1 through 120-6 as shown. The corrected actuator signals output from Sum Amplifiers 120-1 through 120-6 are individually amplified in Power Amplifiers 116-1 through 116-6 to drive the individual Actuators 117-1 through 117-6 associated with Links 1 through 6 respectively. The individual Sensors 110-1 through 110-6 detect the current state of their respective links and generate feedback signals communicated back to Central Processing Unit 102, Sum Amplifiers 112 and the Nonlinear Dynamics Control Computer 118.

The program flow diagram for Nonlinear Dynamics Control Computer 118 is illustrated in FIG. 6. The program begins by recording the desired motion signals received from the Central Processing Unit 102 and the feedback signals received from the Sensors 110-1 through 110-6 as indicated by block 122. The program then proceeds to compute the linear velocity V, linear acceleration V, angular velocity w and angular acceleration w at the end of Link 3 in Link 3 coordinates as indicated by block 124 using the Lagragian dynamics equations previously described.

The next step is to compute the torques and/or forces required from actuators 4, 5 . . . n associated with Links 4 through n of the Wrist 20 and the forces and torques required to support the Wrist 20 at the interface between Links 3 and 4 using the recursive Newton-Euler equation as indicated by block 126. The details of the procedure are the same as algorithm (8) previously described.

The program then proceeds to compute the torques and forces due to inertia on Links 1, 2 and 3 as indicated by block 128. These computations are conducted through the use of Equation 5.

Next the torques or forces due to rotational inertia of Links 1, 2 and 3 are computed as indicated by block 130. These forces are computed using Equation 6.

The program next proceeds to compute the forces due to gravity as indicated by block 132. Equations for the forces due to gravity are well known and are easily derived for the specific manipulator. These forces are computed using Equation 9.

This computation is followed by the computation of the torques or forces T.sub.wi required from the actuators for Links 1, 2 and 3 to balance the forces and moments the wrist applies to the end of Link 3 as indicated by block 134. These computations are made using Equation 10.

Finally, the total torques or forces for the actuators of Links 1, 2 and 3 are computed as indicated by block 136. These torques are computed using Equation 4 as previously indicated. Namely:

The last operation of the Nonlinear Dynamics Control Computer 118 is to output the computed torques and forces as correction factor signals to Sum Amplifiers 120-1 through 120-6, as indicated by Block 137, where they are summed with the actuator signals.

Inspection of the program represented by the flow diagram illustrated on FIG. 6 shows that the computations performed by blocks 124 and 126 are independent of the computations performed by blocks 128, 130 and 132 and need not be performed in the indicated sequence. Additionally, the computer time required for the computations of blocks 124 and 126 is approximately the same as the computer time required for the computations of blocks 128, 130 and 132. Therefore, the computer time for making these computations can be cut in half by independently performing the computations of blocks 124 and 126 in parallel with the computations of blocks 128, 130, and 132 using two microprocessors such as Microprocessor 138 and Microprocessor 140 as shown in FIG. 7.

Referring to FIG. 7, Microprocessor 138 computes the correction factors for the nonlinear dynamics for the wrist actuators using the recursive Newton-Euler Equation computations of blocks 124 and 126 while Microprocessor 140 computes the correction factors for the nonlinear dynamics of the arm actuators (actuators for Links 1, 2 and 3) using the Lagrange Equation computations of blocks 128 through 132. The computations of blocks 134 and 136 are then completed in Microprocessor 140 using the results of the computations of block 126 in Microprocessor 138 in combination with the results of the computations performed in blocks 128, 130, and 132.

The correction factors computed by Microprocessor 138 for the wrist are added to the outputs from the Servo Amplifiers 114-4, 114-5, . . . 114-n collectively designated Servo Amplifiers 114-B by a plurality of individual sum amplifiers, such as Sum Amplifiers 120-4, 120-5 . . . 120-n, collectively designated Sum Amplifier 120-B. The outputs of the individual Sum Amplifiers 120-B are individually amplified by a like plurality of power amplifiers, Power Amplifiers 116-B which activate the actuators, Actuators 142 associated with Links 4, 5, . . . n. The sensors associated with Links 4, 5, . . . n, Sensors 110-B, generate position and velocity feedback signals for the Central Processing Unit 102 and the Dynamic Control Computer 118.

In a like manner, the correction factors generated by Microprocessor 140, for the first three links, Link 1, 2 and 3, are added to the outputs from Servo Amplifiers 114-1, 114-2 and 114-3, collectively designated Servo Amplifier 114-A by Sum Amplifiers 120-1, 120-2, and 120-3, collectively designated Sum Amplifier 120-A. The summed signals are amplified in individual Power Amplifiers 116-1, 116-2, and 116-3, designated Power Amplifiers 116-A, which activate the actuators for Links 1, 2, and 3, designated Actuators 144. The Sensors 110-1, 110-2, 110-3, designated Sensors 110-A generate position and velocity feedback signals to the Central Processing Unit and Dynamic Control Computer as previously described.

The parts of the program shown on the flow diagram of FIG. 6 executed by individual Microprocessors 138 and 140 are shown on FIG. 8. Referring to FIG. 8, Microprocessor 138 computes the torques and forces for Links 4, 5 . . . n using the recursive Newton-Euler Equation as previously described with reference to blocks 124 and 126. The torques computed in block 126 for Links 4, 5 . . . n are transmitted to Microprocessor 140 to complete the computations of the torques for Links 1, 2 and 3 as indicated by block 146. The torques computed by block 126 are also transmitted to Sum Amplifiers 120-4, 120-5, . . .120-n as the correction factors for the actuator signals output from Servo Amplifiers 114-4, 114-5, . . . 114-n. Simultaneously and in parallel, Microprocessor 140 computes the torques for Links 1, 2 and 3 to balance translation inertia, rotational inertia and gravity using Lagrange's Equations as indicated by blocks 128, 130 and 132. The computed torques and forces for Links 4, 5 . . . n received from Microprocessor 138 are then utilized in Microprocessor 140 to compute the torques and/or forces required to balance the forces of Links 4, 5 . . . n acting on the end of Link 3 as indicated by block 134. Finally, all of the forces acting on Links 1, 2, and 3 are summed together to generate the nonlinear torques acting on Links 1, 2, and 3 which are output as the correction factors for the respective links.

Using this second dynamic control loop, the manipulator control system is capable of compensating for the nonlinear dynamics of the manipulator system in a more precise and efficient manner. By dividing the links of the manipulator into two sets and placing the origin of the third link coordinates at its end, the nonlinear forces acting on links 4, 5, . . . n which normally comprise Wrist 20 can efficiently be computed using the recursive Newton-Euler Equations and the nonlinear torques and forces acting on Links 1, 2 and 3 can be computed using Lagrangian Dynamics. This nonlinear dynamic control system, reduces the required computer time to compute the desired correction factors for each of the links to within a range from 0.003 to 0.004 seconds which is about 5 times faster than using Newton-Euler dynamics as taught by Luh et al. Further by using two microprocessors in the Nonlinear Dynamic Control Computer, the computer time can be further reduced by a factor of approximately 2 making the dynamic manipulator control system almost 10 times faster than the prior art and well within the desired 0.01 seconds.

Having disclosed the invention it is recognized that those skilled in the art may take changes to the embodiment and algorithms disclosed herein without departing from the spirit of the invention as described herein and set forth in the appended claims.