Uncertainty-Aware Motion Planning for Autonomous Driving in Mixed Traffic Environment

The paper introduces Uncertainty-Aware Motion Planning (UAMP), a novel approach for autonomous vehicle (AV) motion planning that incorporates uncertainty in predicting human driver intents. UAMP employs a proximity-aware uncertainty estimator to create a joint intent distribution and utilizes Uncertainty-Calibrated Value Learning (UCVL) to mitigate biases in value function learning. Experimental results demonstrate that UAMP enhances safety and driving comfort in mixed-traffic environments, which is critical for practitioners developing AV systems that must operate alongside human drivers.