This project addresses the interaction of highly automated vehicles with vulnerable road users (VRU) such as pedestrians and cyclists, in the context of future urban mobility. It covers VRU sensing, cooperative localization, behaviour modeling, intent recognition, and vehicle control. Within the AMR group we focus on vehicle control for safe and efficient autonomous driving. See the project website.
Obstacle avoidance for quadrotors with limited sensing and compute in dynamic environments. We developed a Dual-Structure Particle-based (DSP) map for continuous, real-time occupancy prediction on onboard CPUs, together with risk-aware sampling and optimization-based trajectory planners.
MAVs with active sensing to overcome FOV limits in obstacle avoidance. Sensing direction for a camera with an independent DOF is planned from the environment and flight condition, paired with a computationally efficient trajectory planner. The system reached over 6 m/s in woods with robust performance in dynamic settings.
A systematic study of environment representations for visual navigation through sim-to-real transfer learning, with criteria for suitable representations and an end-to-end navigation model.