Choreographing Safety: Planning via Ice-cone-Inspired Motion Sets of Feedback Controllers for Car-Like Robots

Safe navigation of car-like robots operating via onboard sensing on crowded roads is a challenging task in general. Our primary objective in this work is the development of a novel sensor-based control structure that directly computes safe control inputs for navigation based on the motion sets of pre-designed feedback controllers. We demonstrate an input-constrained feedback controller to induce motion sets whose geometric structure is an ice-cone. The key implication of this is that the naturally induced system trajectories evolve within this motion set during stabilization. We then propose an instantaneous planning strategy for safe navigation that choreographs a sequence of ice-cones within the bounds of safety. This enables the robot to navigate safely without the explicit requirement of adhering to pre-computed paths/trajectories. We demonstrate the proposed methodology via ROS simulations on test scenarios using the CAT vehicle testbed. Further, we demonstrate its application in a traffic intersection scenario using the trajec-tory information collected from the interaction dataset. The proposed approach especially finds applications in scenarios where pre-computing safe trajectories based on the predicted trajectories of neighboring agents is difficult/unreliable during run-time.