Automated particle swarm optimization based PID tuning for control of robotic arm

In this paper, we revisit the Proportional-Integral-Derivative (PID) controller design for torque control of robotic manipulators, for which, appropriate tuning of the said controller could prove very burdensome, especially with increasing degrees-of-freedom (DOF) and/or when designing a Multi-Input Multi-Output (MIMO) PID controller. That is, when generating and tuning matrix P-I-D gains as opposed to single values, in order to take in account possible coupling effects between involved joints. We tackle both joint space and workspace PID control tuning problems for reference tracking from an optimization standpoint. Using a previously developed stable Adaptive Particle Swarm Optimizer, we are able to automatically and systematically tune P-I-D gains, be it as single gain values or gain matrices, while optimizing a cost or fitness function. The aforesaid cost function can be arranged to feature various aggregated performance measures, `normalized' so as to overcome differences in scale if any. Taking in account some practical limitations, a 2-DOF arm is used here as a case study. Numerical simulations are provided to substantiate the adequacy of our method.