Path Planner for Dynamic Environments

Research is proposed to investigate the feasibility of a path planning module that makes use of time-bounded lattices to avoid potential collisions in dynamic environments. The challenge of planning in a dynamic environment is two-fold. First, tracking and predicting the trajectories of dynamic obstacles (e.g., cars, humans, and aircraft) is very noisy. Second, planning with dynamic obstacles requires adding time as a dimension to the planning problem. Paths need to be re-computed frequently because dynamic obstacles are constantly in motion. To deal with the uncertainty inherent in predictions, our research will investigate a novel representation of dynamic obstacles that enables the planner to generate safe paths efficiently yet is general enough for use with wide variety of perception modules and sensor suites. Second, to provide real-time guarantees on the performance of the planner, we propose to utilize a novel data structure for planning – a time-bounded lattice. The innovative time-bounded lattice will be well-suited for use with previously developed re-planning algorithms that have already been successfully employed on a variety of unmanned ground vehicles.