CVT-based 2D motion planning with maximal clearance

Maximal clearance is an important property that is highly desirable in multi-agent motion planning. However, it is also inherently difficult to attain. We propose a novel approach to achieve maximal clearance by exploiting the ability of evenly distributing a set of points by a centroidal Voronoi tessellation (CVT). We adapt the CVT framework to multi agent motion planning by adding an extra time dimension and optimize the trajectories of the agents in the augmented domain. As an optimization framework, our method can work naturally on complex regions. We demonstrate the effectiveness of our algorithm in achieving maximal clearance in motion planning with some examples.