Smooth Scenario-Based Model Predictive Control for Autonomous Collision Avoidance in Inland Waterways

Scenario-Based Model Predictive Control (SB-MPC) is an autonomous collision avoidance algorithm primarily designed for open and coastal waters. Over the years, the algorithm has shown significant promise in both simulations and real-world experiments. However, one of the challenges in adapting SB-MPC for autonomous inland waterway collision avoidance is the inherent resolution insufficiency due to deriving solutions from a finite set of discretized solutions using exhaustive search. To increase the resolution of the solution space, a derivative-based optimization strategy would be required. However, the non-smooth components of the SB-MPC cost function prohibit this approach. Therefore, we propose a newer variant of the algorithm, Smooth Scenario-Based Model Predictive Control (Smooth-SBMPC), specifically designed for the highly constrained and complex navigational environments inherent to inland waterways. It utilizes different techniques, such as using Fuzzy Logic, to convert the non-smooth components of the original SB-MPC cost function into smooth, derivable ones. The effectiveness of Smooth-SBMPC is validated through a comprehensive simulation study.