PID Controller Design for A Nonlinear Motion Control Based on Modelling the Dynamics of Adept 550 Robot

The multiple link robot is a nonlinear system, in general whose characteristic polynomial is not unique, and whose control schemes depend on particular operating conditions. Among all kinds of controllers used for robot control, nearly 90% is PID controller in the industrial world, not only due to its simplicity and effectiveness, but also to its ability of coping with uncertainties and nonlinearities in the system. Since most of PID controllers are designed based on trial and error method for a black-box plant, for the precision motion control of a robot, it is more desirable to design the PID controller based on the mathematical model of the robot. This paper is dedicated to a PID controller design for a nonlinear motion control based on the mathematical modelling of the dynamics of Adept 550 Robot. The general relationships of the PID controller design on the robotic dynamics and the planned trajectory are derived. The analysis and simulations of its closed loop dynamics indicates its effectiveness in fast and accurate trajectory tracking. The results can be practically generalized to other cases of PID controller design for other robots in the industrial applications.


INTRODUCTION
The multiple link robot is a nonlinear system, in general whose characteristic polynomial is not unique, and whose control schemes depend on particular operating conditions.Among all kinds of controllers used for robot control, nearly 90% is PID controller in the industrial world, not only due to its simplicity and effectiveness, but also to its universality and feasibility.PID controller can be designed according to many PID tuning methods, even if the mathematical model of the system is not available, or uncertainties and nonlinearities were found in the system.PID controller has fairly easily understood physical meanings: the proportional control provides a restoring force that is proportional to the instantaneous error between the set point and the process variable at the moment, but it may end up with steady-state error due to friction, gravity or so; the integral control contributes a corrective force that is proportional to the sum of all past errors multiplied by their own time interval, so that the steady-state error can be eliminated, but it may cause overshoot of the system response; derivative control is introduced to address the overshoot issue by producing a counteractive force that is proportional to the rate of change of error.
With these understandings, often engineers who do not have a systematically training of control theory can design a PID controller for the industrial process whose transfer functions are unknown to them.That is why most of PID controllers are designed based on trial and error method for a black-box plant.
However, for the precision motion control of a robot, the trial and error approach is not an efficient way.In recent years, more and more researchers aim to design different types of PID controllers based on theoretical analysis to deal with nonlinearity or non-explicates for different applications, such as indoor blimp robot, gyro mirror line-of-sight system, two wheeled autonomous balancing robot, parallel manipulator, etc. [1,2,3,4,5,6,7,8].It is demonstrated that the theoretically designed PID controllers can be optimally tuned to fit the needs of a specific systems or a generic task, for better performance than that the trial-and-error learning designed PID controller has.This paper is dedicated to explore a theoretically designed PID controller for a nonlinear motion control on the basis of the mathematical model of Adept 550 robot.As one of the most commonly used robots in the industrial production lines, the Adept 550 robot is a fouraxis SCARA robot with three rotational joints and one translational joint (Fig. 1-a).Its shoulder is embodied with the rotational joint 1, its elbow with the rotational joint 2, and its wrist with the rotational joint 4. Its vertical motion is accomplished with the translational joint 3. Since it features a small motion envelope while its speeds and payloads are relatively high, Adept 550 robot can be found in mechanical assembly, material Zhang IJIEM handling, packaging, machine tending, screwdriving, and many other operations requiring fast and precise automation.Needless to say, a complete study of PID controller design based on the mathematical model of Adept 550 is meaningful for its industrial applications.The geometric dimensions of Adept 550 are shown in Fig. 2 for modelling and simulations.

The Link Positions and Inverse Kinematic Relationships
According to the Denavit -Hartenberg (D-H) coordinates, the z -axis is in the direction of the joint axis, Adept 550 has the special case of parallel z axes, connected with the rigid inner and outer links.The trajectory of the robot is determined by the motion of these two links, and at the wrist the rotational joint rotates about the z axis to adjust the gripper angle, but not to change the trajectory.In order to focus on the trajectory study, without loss its generality, we assume the wrist's rotary angle is zero.Therefore, the coordinate systems to describe trajectory and link positions can be simply presented as follows (Fig. 3): For the inner link (i=1) and outer link (i=2), the coordinate transformation with rotation and translation components is described with the matrix: The simplified transformation matrix of Adept 550 from the base to the gripper is as follows:

Forward and Backward Velocity and Acceleration Relationships
The forward velocity v can be found from Equation (3): where a J is Jacobian matrix And the backward velocity ( ) Also, the relationships of the forward acceleration ( ) The relationships of the backward acceleration ( )

Modelling and Simulation of the Dynamics of Adept 550 Robot
For the simplified structure of the robot as shown in Fig. 4, the inner link and outer link are represented as rigid bars with the central mass For the robotic structure shown in Fig. 4, the total kinematic energy T is expressed as: And the total potential energy V is given by: Thus the dynamics of Adept 550 robot described by Lagrange method is as follows: Substituting Equations ( 13) -(18) into Equation ( 19) results in: , and considering damping torques damping τ of the inner and outer links, Equation (20) can be rewritten as: where It can be observed that the dynamic model of Adept 550 robot is a highly nonlinear system.The simulation of the robot dynamics will be performed with Adams-Bashforth recursive scheme, which has been applied to general open-chain type manipulators and proven to be much more efficient in computational load than the nonrecursive methods.This is especially the case when the system is highly nonlinear.Fig. 5 and 6 display the dynamic responses of angular displacements and velocities with the initial conditions at a given gripper position.The simulation illustrates the dynamics modeling of Adept 550 developed by Lagrange method, and demonstrates the stability of the robot motion.It has been observed that the heavier the load 3 m is, the longer it takes the system to be settled; and when the load is heavier than a certain weight, the gripper system becomes unstable.

Sensitivities to Static Torques in Open Loop Control
In the open loop control of Adept 550 robot, the sensitivities of angular displacements to static torques indicate how well the gripper can hold its position.These sensitivities of angular displacements to static torques are as follows:

Robot Dynamics with PID Controller
It is assumed that the motors driving the inner and outer links are the same type of motors.Considering the motor parameters, dynamics of the two links is described as: , where r is the gear ratio, the two dynamic equations of robot link and its driving motor expressed in Equation ( 24) can be combined into a single one: The PID control law for the controller and motor drive at the inner or outer link is as follows: ) The block diagram based on the transfer functions of Equations ( 25) and ( 26) for the robot control is shown in Fig. 8.The transfer function of the closed loop system is: where the nonlinear terms are treated as disturbances to the system: Its root locus analysis for the inner link and outer link of the tuned PID controllers is demonstrated in Fig. 9. , the dynamics of the inner or outer link can be described as 30) can be rewritten in matrix form: where

Trajectory Planning
The PID controller can be designed to perform a trajectory tracking from the initial position ( ) The trajectory component x P can be expressed as follows: Per Equations (35), the trajectory related matrices and vector given by Equation (32) can be determined.Then the closed loop controlled dynamics of Adept 550 robot expressed in Equation ( 30 In order to verify the effectiveness of the PID controller based on dynamics of Adept 550 robot, the following case is given for simulations: assume the PID controller is with a critical damp for both inner link and outer link, the gripper moves from the initial position It can be observed that the trajectory tracking process is fast and stable.Its evaluation is given in the appendix.

CONCLUSION
The PID controller design for a nonlinear motion control based on the mathematical modelling of the dynamics of Adept 550 Robot is explored.
The general relationships of the PID controller design on the robotic dynamics and the planned trajectory are derived.The analysis and simulations of its closed loop dynamics indicates its effectiveness in fast and accurate trajectory tracking.
The procedure and analysis of this research can be practically generalized to other cases of PID controller design for other robots in the industrial applications.It is meaningful for optimizing the commonly used PID controller without trial-and-error testing, and it is especially important for precision operations.

Figure 3 .
Figure 3. Top view of Adept 550 with joints rotated by angles

Figure 5 .Figure 6 .
Figure 5. Dynamic responses: angular displacements of static torques τ in the open loop will produce errors in angular displacements β .A closed loop position control is desirable to address this issue (Fig.7).

Figure 8 .
Figure 8. PID controlled robotic motions manipulate the root locus yield the following relationships on the PID control parameters:

Figure 9 .
Figure 9. Root locus analysis for inner link (a) and outer link (b)

1
It should satisfy the following requirement over the entire trajectory:

Table 1 . lists the D-H parameters of the inner and outer links. Table 1. D-H Parameters
) or (31) during the planned trajectory can numerically simulated.
) Figure 10.Desired trajectory planning in x axis