Motion planning is a cornerstone of the robotics pipeline. Given a mobile object (the robot), an environment, and a start and goal position, the objective of motion planning is to find a valid path from start to goal. A path is valid if it meets a problem-specific set of constraints; such constraints can include the robot does not collide with any obstacles, the robot does not collide with itself, or the robot does not exceed any physical limitations to its velocity.

Our lab focuses its research on sampling-based motion planning, wherein points are sampled according to some distribution in the robot’s configuration space, and then connected togother in a graph to find its goal. We are researching a variety techniques to make this process more performant, including adding smarter sampling to sample more relevant points, parallelization, and improvements to the validity checking process to verify samples more quickly.

Publications

Updated: