Real-time Collision Risk Estimation based on Stochastic Reachability Spaces

Estimating the risk of collision with other road users is one of the most important modules to ensure safety in autonomous driving scenarios. In this paper, we propose new probabilistic models to obtain Stochastic Reachability Spaces for vehicles and pedestrians detected in the scene. We then exploit these probabilistic predictions of the road-users' future positions, along with the expected ego-vehicle trajectory, to estimate the probability of collision risk in realtime. The proposed stochastic models only depend on the velocity, acceleration, tracked bounding box, and the class of the detected object. This information can easily be obtained through off-the-shelf 3D object detection frameworks. As a result, the proposed approach for collision risk estimation is widely applicable to a variety of autonomous vehicle platforms. To validate our approach, initially we test the stochastic motion prediction on the KITTI dataset. Further experiments in the CARLA simulator, by reproducing realistic collision scenarios, have the goal of demonstrating the effectiveness of the collision risk assessment and are compared with an alternative approach.