Real-time Adaptive GPU multi-agent path planning
The project implements a GPU path planning algorithm that is derived from the sequential A* algorithm to allow massively parallel, real time execution. The new algorithm employs a limited look ahead strategy similar to the wave fronts of a breath-first-search algorithm. Using a heuristic to estimate the most profitable direction for moving along a particular direction at leach step, the algorithm strikes a balance between work set size and optimality. The implementation of the algorithm further employs a windowed strategy to reduce the amount of information that need to be maintained for fast access.
|GPU RTAA.zip||858.79 KB|