Risk-aware motion planning with CVaR maps for collision avoidance in dynamic environments

Abstract

Nonlinear model predictive control (NMPC) is an efficient and proven method for optimization-based autonomous vehicle motion planning. In safety-critical control systems, controllers should address inequality-constrained optimization problems. In this work, we design a single unified constraint using an grid map. A single discrete barrier state is added to the system model to transform a constrained optimization problem into an unconstrained optimization problem. This approach can simplify complex motion planning problems and reduce computational cost. In addition, we defined a new safety metric for collision avoidance with dynamic objects and measured the risk using Conditional-Value-at-Risk (CVaR) to account for uncertainty. For safe motion planning that minimizes collision risk, we design a CVaR map, and design a single constraint that allows the vehicle to drive only in cells with risk below a predefined threshold. Finally, we propose a risk-aware motion planning method by applying it to the grid map-based integrated motion planning method. We verify the performance of the proposed method and the conventional MPC based motion planning method through real-time simulations in CarMaker and ROS environments. In the pop-up obstacle avoidance scenario and overtaking scenario in a dynamic environment, the proposed risk-aware motion planning method shows advantages in computation time and driving stability. Finally, the performance of the proposed method is experimented and verified on a real autonomous vehicle.