Stochastic motion planning via probabilistic occupancy function

Mar 30, 2019 1 min read

The video below shows the example discussed in our CDC 2018 paper. The successive convexification-based planner computes these trajectories within $0.25$ seconds (sampling time). The obstacle motion prediction is done on-the-fly using my work on probabilistic occupancy function.

  1. The colored circles are obstacles that we must avoid. The obstacles have known stochastic dynamics. The figure on the right shows the $\alpha$-probability occupied sets.
  2. The black dot is the current robot location.
  3. The blue asterisks show the planner’s intended future trajectory. We apply the same code to an environment with more obstacles.
stochastic-optimal-control
Research Scientist

Researcher with experience in optimization, control, stochastic modeling, and reinforcement learning