Experiences with online local model predictive control for WMR navigation

Path-tracking by using MPC (model predictive control) is a suitable control science solution for mobile robot navigation applications. Online MPC is reported by using short-term horizons that allow dealing with flexible path-tracking and reactive behaviors. The majority of MPC experimental research developed is based on the fact that the reference trajectory is known beforehand. However, under dynamic environments the global solution becomes unfeasible for the majority of applications where the scenario should be considered as partially unknown due to the lack of global sensors or the existence of dynamic obstacles. Moreover, traditional motion control of wheeled mobile robots (WMRs) is achieved by using discontinuous control laws that are implemented through low level velocity PID controllers. Instead of using such methods, this work proposes to use local MPC as a useful methodology for WMR navigation under dynamic environments or as obstacle avoidance strategy. In this way, the desirable path coordinates are used in the control law as a way for obtaining the robot speed commands. Simulation results are used for addressing online MPC implementations. The system is on-robot tested by using simple on board perception systems. In this context, a local occupancy grid is built by using dead-reckoning and monocular data.