Self-driving Electric Wheelchair in Crowded Environments Using a Fuzzy Potential Model Predictive Control

In the recent development of small autonomous vehicles, such as electric wheelchairs, one of the challenging environments to be operated is a congested area in which vehicles are mixed with pedestrians. In addition to prompt avoidance in response to the surrounding nonstationary environment, optimal maneuvering must be conducted between the multiple feasible trajectories. In the present study, the fuzzy potential method is combined with model predictive control to avoid obstacles based on predictions of future behavior. Furthermore, the global search for a solution and flexible switching of avoidance paths were achieved utilizing the Monte Carlo optimization in model predictive control. This also enables the optimization of a nonlinear and discontinuous evaluation function including membership functions for fuzzy inference. The method's effectiveness was verified through simulations and real-time experiments for a selfdriving electric wheelchair equipped with a LiDAR, and verification confirmed flexible obstacle avoidance. Copyright (c) 2023 The Authors.