Safe Occlusion-Aware Autonomous Driving via Game-Theoretic Active Perception

Abstract

Autonomous vehicles interacting with other traffic participants heavily rely on the perception and prediction of other agents' behaviors to plan safe trajectories. However, as occlusions limit the vehicle's perception ability, reasoning about potential hazards beyond the field of view is one of the most challenging issues in developing autonomous driving systems. This paper introduces a novel analytical approach that poses safe trajectory planning under occlusions as a hybrid zero-sum dynamic game between the autonomous vehicle (evader) and an initially hidden traffic participant (pursuer). Due to occlusions, the pursuer's state is initially unknown to the evader and may later be discovered by the vehicle's sensors. The analysis yields optimal strategies for both players as well as the set of initial conditions from which the autonomous vehicle is guaranteed to avoid collisions. We leverage this theoretical result to develop a novel trajectory planning framework for autonomous driving that provides worst-case safety guarantees while minimizing conservativeness by accounting for the vehicle's ability to actively avoid other road users as soon as they are detected in future observations. Our framework is agnostic to the driving environment and suitable for various motion planners. We demonstrate our algorithm on challenging urban and highway driving scenarios using the open-source CARLA simulator.

Toward Safer Autonomous Vehicles: Occlusion-Aware Trajectory Planning to Minimize Risky Behavior

“…Given the study's methodology, we haven't considered additional uncertainties such as sensor noise, road friction coefficients, or controller inaccuracies. While we haven't factored in uncertainties like perception [27], it's worth mentioning that our method primarily deals with static obstacles that limit the field of view, typically identified reliably by perception systems. The method works for occluded areas behind clearly visible static obstacles and given semantic information.…”

“…New concepts and planning extensions: There is also work developing a new planning concept for the problem of occluded areas, for example, using game theory [27]. Zhang et al formulate the problem as a dynamic zero-sum game between the autonomous vehicle and an initially occluded road user.…”

Occlusion-Aware Motion Planning for Autonomous Driving

“…Inspired by works [34] and [35]we incorporate the Probability distribution of agent emergence from obscured regions predicted from contextual information which obtained in previous work [2] into the incomplete information virtual game to regulate the speed of the autonomous vehicle on motion planning.…”

An Intelligent Self-Driving Truck System for Highway Transportation

“…However, there is still a long way to full L4 autonomous driving truck, especially dealing with more complex traffic and pedestrians. We noticed many recent advanced studies have been released targeting several areas in autonomous driving, e.g., collision avoidance (Zhang and Fisac, 2021), left-turn planning (Shu et al, 2020). For the future study, we plan to cover more logistic truck operation scenarios in addition to highways, e.g., on-ramp/off-ramp scenarios, left-turn/right-turn planning in crowd intersections.…”