Researchers led by Lorenzo Paiola have built RD-RRT*, a path planner that aims for the lowest chance of a collision rather than the shortest path, according to a paper on arXiv that the International Journal of Robotics Research has accepted. It extends the RRT* planner to spaces with both known and uncertain obstacles.
It relies on Risk Density, a measure that judges a whole path at once and approximates its probability of collision. It can also weigh how severe a collision would be, and a hybrid version adds a penalty for length, so paths stay short and the search does not wander through low-risk areas.
In a wide range of scenarios, the planner kept finding paths with a lower estimated collision probability, and Monte Carlo simulations confirmed the approximation. The abstract gives no figures.
Background
More stories about Lorenzo Paiola
RD-RRT*
Path planning method
Path planner that extends RRT* to find the paths with the lowest risk of collision.



