What is a Probabilistic Roadmap (PRM) in robotics path planning?

What is a Probabilistic Roadmap (PRM) in robotics path planning?

• A. A deterministic grid search
• B. A sampling-based method to build a roadmap of collision-free configurations and connections, then query a path
• C. A sensor fusion algorithm
• D. A data compression method

Answer: (B)

Probabilistic Roadmap is a sampling-based path planning approach that builds a roadmap of the robot’s configuration space. It works by randomly sampling collision-free configurations (robot poses or joint values) and connecting nearby samples with edges that correspond to collision-free local motions. Once the roadmap is built, you can query a path by attaching the start and goal configurations to the roadmap and using a graph search to find a sequence of configurations from start to goal. This method handles high-dimensional robots much better than grid-based methods because it explores the free space statistically rather than exhaustively. The roadmap can be reused for many queries, making online planning fast after the offline learning phase. The randomness means that with more samples, the chance of finding a valid path when one exists increases. It’s distinct from sensor fusion or data compression, which solve unrelated problems.

Probabilistic Roadmap is a sampling-based path planning approach that builds a roadmap of the robot’s configuration space. It works by randomly sampling collision-free configurations (robot poses or joint values) and connecting nearby samples with edges that correspond to collision-free local motions. Once the roadmap is built, you can query a path by attaching the start and goal configurations to the roadmap and using a graph search to find a sequence of configurations from start to goal. This method handles high-dimensional robots much better than grid-based methods because it explores the free space statistically rather than exhaustively. The roadmap can be reused for many queries, making online planning fast after the offline learning phase. The randomness means that with more samples, the chance of finding a valid path when one exists increases. It’s distinct from sensor fusion or data compression, which solve unrelated problems.