Project

Autonomous Car Drifting

Optimize car trajectory in constrained time frame using dynamic/kinematic limits and collision as constraints for non-linear optimization problem. Localize with LiDAR scan matching and feed estimated pose to MPC.

Inverse Rendering

Create realistic sensor model that enables optimization of sensor model parameters and reasoning about scene geometry.

Cost-Based APF-RRTs

Speed up RRT with artifical potential field while being able to escape local minima.

Hierarchical Multi-Agent Path Finding

Spatially divide large maps into sub-regions and find high-level paths to provide guidance to low-level ECBS solvers in each sub-region.

Task and Contact Optimization for Robot Manipulation

Formulate high-level symbolic planning as a smooth optimization problem and combine with low-level CIO framework to optimize over complex contact planning problems.

Physics-based LiDAR Simulation

Accurately model LiDAR sensor in a differentiable framework supporting gradient-based optimizations

Motion Planning Benchmark

Benchmark state-of-the-art motion planners and techniques for wheeled mobile robots to highlight their strengths and weaknesses.