Nearest-Neighbourless Asymptotically Optimal Motion Planning with Fully Connected Informed Trees (FCIT*)

Tyler S. Wilson,Wil Thomason,Zachary Kingston,Lydia E. Kavraki,Jonathan D. Gammell

2024-11-27

Abstract:Improving the performance of motion planning algorithms for high-degree-of-freedom robots usually requires reducing the cost or frequency of computationally expensive operations. Traditionally, and especially for asymptotically optimal sampling-based motion planners, the most expensive operations are local motion validation and querying the nearest neighbours of a configuration. Recent advances have significantly reduced the cost of motion validation by using single instruction/multiple data (SIMD) parallelism to improve solution times for satisficing motion planning problems. These advances have not yet been applied to asymptotically optimal motion planning. This paper presents Fully Connected Informed Trees (FCIT*), the first fully connected, informed, anytime almost-surely asymptotically optimal (ASAO) algorithm. FCIT* exploits the radically reduced cost of edge evaluation via SIMD parallelism to build and search fully connected graphs. This removes the need for nearest-neighbours structures, which are a dominant cost for many sampling-based motion planners, and allows it to find initial solutions faster than state-of-the-art ASAO (VAMP, OMPL) and satisficing (OMPL) algorithms on the MotionBenchMaker dataset while converging towards optimal plans in an anytime manner.

Robotics

What problem does this paper attempt to address?

The problem that this paper attempts to solve is: how to improve the performance of high - degree - of - freedom robot motion planning algorithms, especially in the asymptotically optimal case. Traditionally, the most expensive operations in asymptotically optimal sampling - based motion planning algorithms are local motion verification and nearest - neighbor query of the configuration. Although recent research has significantly reduced the cost of motion verification by using single - instruction - multiple - data (SIMD) parallelism, these advances have not yet been applied to asymptotically optimal motion planning.

Specifically, the paper proposes a new algorithm - Fully Connected Informed Trees (FCIT*) - which aims to use SIMD parallelism to significantly reduce the cost of edge evaluation, thereby constructing and searching a fully - connected graph. This not only eliminates the need for the nearest - neighbor structure, but also allows FCIT* to outperform existing asymptotically - super - optimal (ASAO) algorithms (such as VAMP, OMPL, etc.) in terms of the speed of finding the initial solution, and can converge to the optimal solution in an any - time manner.

### Main Problem Summary:
1. **Improve motion planning efficiency**: Reduce the cost of computationally - intensive operations, especially local motion verification and nearest - neighbor query.
2. **Achieve asymptotically optimal solution**: Ensure that the algorithm can gradually approach the optimal solution as the computation time increases.
3. **Eliminate the need for the nearest - neighbor structure**: Simplify the search process by constructing a fully - connected graph and improve the overall efficiency of the algorithm.

### Main Contributions of FCIT*:
- **Fully - connected graph**: By considering all possible edges, make full use of the current sample set and avoid the overhead of the nearest - neighbor structure.
- **Efficient edge evaluation**: Significantly reduce the cost of edge evaluation by using SIMD parallelism.
- **Ordered search**: Search in the order of the quality of potential solutions, similar to the A* algorithm, but without the need to pre - discretize the search space.

Through these improvements, FCIT* performs well in multiple benchmark tests. In particular, it can find higher - quality initial solutions more quickly in complex environments (such as cage environments) and gradually converge to the optimal solution as the computation time increases.