Model Predictive Control of a Two-Wheeled Robot

Ronald P.M. Chan, Karl A. Stol, and C. Roger Halkyard

Keywords

Predictive control, Modelling and simulation, Twowheeled robots, Constraint Tightening

Abstract

Two-wheeled robots, with two wheels located coaxially on either side of the robot and centre of mass above the wheel axles, are statically unstable systems. This paper applies linear model predictive control (MPC) as an effective control strategy which allows constraints to be specified. The MPC controller predicts future states to avoid control input exceeding motor current and voltage limits, which would otherwise result in the system becoming unstable. To account for the non-linearity of the two-wheeled robot, the constraint tightening method is used – a sufficiently large margin is reserved to regulate any difference between the internal linear model and the true nonlinear model. The model predictive controller is compared to a linear-quadratic regulator (LQR) in simulation. While the LQR response becomes unstable for aggressive reference speed inputs, the MPC controller can successfully operate the robot within speed and acceleration limits to ensure continued stability.

Important Links:



Go Back