Overall Concerns

Not Complete

Sample-based planning is not complete, it is probabilistically complete. In applications where decisions need to be made quickly, PRM & RRT may fail to find a path in difficult environments, such as the one shown below.

To path plan in an environment such as the one presented above, alternate means of sampling can be introduced (such as Gaussian or Bridge sampling). Alternate methods bias their placement of samples to obstacle edges or vertices of the open space.

Not Optimal

Sample-based path planning isn’t optimal either - while an algorithm such as A* will find the most optimal path within the graph, the graph is not a thorough representation of the space, and so the true optimal path is unlikely to be represented in the graph.

Conclusion

Overall, there is no silver bullet algorithm for sample-based path planning. The PRM & RRT algorithms perform acceptably in most environments, while others require customized solutions. An algorithm that sees a performance improvement in one application, is not guaranteed to perform better in others.

Ultimately, sample-based path planning makes multi-dimensional path planning feasible!