Moving Obstacle Avoidance for Large, High-Speed Autonomous Ground Vehicles

This work introduces the use of hard constraints to avoid moving obstacles for navigating a large, high-speed autonomous ground vehicle in an unstructured environment using nonlinear model predictive control in a single-level architecture, where path planning and tracking are combined into a single optimization problem. Additionally, the hard constraints approach is compared to the traditional approach in this context which implements obstacle avoidance by augmenting the obstacle avoidance requirements into the cost function as soft constraints. In both approaches, the control signals, which are steering angle command and reference longitudinal speed, are optimized using a nonlinear vehicle dynamics model, where the objective is to minimize the time-to-goal. Results indicate that the hard constraints approach outperforms the soft constraints approach both in terms of obstacle avoidance performance and optimization time.