Static and Dynamic Path Planning Using Incremental Heuristic Search
This addresses path planning for automated vehicles in dynamic settings, representing an incremental improvement over existing methods.
The paper tackled path planning for car-like agents in dynamic environments, finding that traditional methods fail to respect dynamic constraints and proposing a novel 4D state space scheme that efficiently produces near-optimal trajectories with bounded optimality.
Path planning is an important component in any highly automated vehicle system. In this report, the general problem of path planning is considered first in partially known static environments where only static obstacles are present but the layout of the environment is changing as the agent acquires new information. Attention is then given to the problem of path planning in dynamic environments where there are moving obstacles in addition to the static ones. Specifically, a 2D car-like agent traversing in a 2D environment was considered. It was found that the traditional configuration-time space approach is unsuitable for producing trajectories consistent with the dynamic constraints of a car. A novel scheme is then suggested where the state space is 4D consisting of position, speed and time but the search is done in the 3D space composed by position and speed. Simulation tests shows that the new scheme is capable of efficiently producing trajectories respecting the dynamic constraint of a car-like agent with a bound on their optimality.