A Nonlinear Model Predictive Controller for Trajectory Planning of Skid-Steer Mobile Robots in Agricultural Environments

This research presents an integrated trajectory planning strategy with a motion control approach using a Nonlinear Model Predictive Control (NMPC) framework for Skid-Steer Mobile Robots (SSMRs) in agricultural scenarios. In a single architecture, the proposed NMPC strategy is aimed at trajectory tracking and involves real-time re-planning of pre-scheduled points in a given crop mapped against static and dynamic obstacles. A Real-Time Iteration (RTI) scheme was adopted to ensure feasibility in the optimization process, even when meeting tightened constraints. A set of potential field functions is formulated to minimize tracking errors and control effort while maximizing obstacle avoidance. The benefits of the proposed strategy regarding performance, constraint satisfaction, and computational were tested via simulations and field trials on an SSMR Husky A200. The results evidenced that prioritizing the robot position and obstacle speeds reduced the tracking error and input effort by 45.3% and 40.8% respectively, compared to the scenario prioritizing only obstacle positions. Thus, prioritizing the obstacle model further mitigates the collision risks in the agricultural field.