In the control community, the control of robots is extensively studied. In most cases, it assumes that the robot is governed by the nonlinear Euler-Lagrange dynamics possibly with uncertainties (e.g., the inertia and mass properties of the links may not be exactly known). Robust control and adaptive control are two standard approaches for the control problem of robots, in most cases, relying on the full nonlinear robot model.
On the other hand, in the applications of robots, especially industrial robots or robots with large gear ratio, the above mentioned adaptive/robust nonlinear controllers are not applicable or preferable. The major obstacle is that the hardware cannot implement the n joint torques simultaneously under a small sampling period (e.g., less than 10 ms) due to the communication constraint between the central computer and each joint controller unit.
For these reasons, in most applications, the independent joint control (e.g., PD controller) is preferable rather than the multi-variable robust/adaptive nonlinear control, although the latter is actively studied in the academic field.
How to think about this gap? Is it reasonable?
the computer speed in calculation is a big problem, i have the same problem but i am not working on it now but have some suggestions,. what about using high speed technologies as DSP or FPGA or other embedded systems that have a high speed of computing or parallel computing ?
With the advent of the high DSP's and FPGA, I doubt that the processing speed is the main reason behind the potential to use the PD controllers rather than more powerful robust adaptive controllers. I think that the simplicity of implementation of the PD controllers plays a vital role in this issue versus the complexity of the realization of the robust adaptive controllers. However, a key point in reducing the gap between the advent in control of robots and that is being implemented in industrial robots, is to look for the disadvantages of the PD controllers (like the limited mode of operation and inability of accommodating possible dynamics drifts) and escalating them to the robots societies will help in reducing such a gap.
I think that the industry prefers the well-known PID control because it works well around the operating point at low velocities, it is well understood, does not require a mathematical model, and it is actually easy to implement in some industrial platforms such that PLCs. However, it is a linear SISO controller. In the independent joint control, the robot is considered as a perturbation that depends on the control configuration, so, this must be implemented with DC motors and grain trains. In the computed torque control an accurate and precise mathematical model is required, that is used to cancel the non-linearities, so, it is not robust. Hence, if the control applications require high speeds provided for instance by induction motors without grain trains, maybe the industry will look for an advanced controller. From my point of view an experience an LQR control or a linear robust control can be implemented with the sampling period required by the hardware, and should be considered first.
I thank Rene Galindo for the excellent answer to this question or problem. I totally agree with you on the reliability of the independent robot joint controllers, which are very common in most industrial robots that intend to manipulate heavy loads. The rationality of using independent control for such robots stems from the fact that the gear ratio is so large that the nonlinear coupling forces can indeed be considered to be small disturbances for each joint, especially at low velocities.
On the other hand, the robots nowadays are expected to accomplish versatile tasks in unstructured environments, which motivates, e.g., the active study of visual servoing control in the robotics community. Fast end-effector motion in large scope irrespective of the uncertainties is often necessary for the successful completion of the task. In this case, it seems not so reasonable to neglect the coupling forces among the robot joints, and it is now necessary to develop the coupling robot torque control to fully accommodate the uncertainties and the nonlinear coupling effect. Yet, we encounter trouble in applications due to the concealing nature of the torque control module of most industrial robots. In fact, most robot-production companies are not willing to open this door to us.
Small aerial vehicles (UAVs) is an specific application that due to their small masses and inertias has fast dynamics, and the couplings can not be neglected. Also, external disturbance of wind must be considered.
- Badges
- Science topic
- Similar topics
- Engineering
- Control Theory