Research on improved RRT path planning algorithm based on multi-strategy fusion.

Area of Science:

• Robotics
• Artificial Intelligence
• Computer Science

Background:

• Rapid-expanding random trees (RRT) algorithms face challenges like search blindness, randomness, slow convergence, and non-smooth paths.
• Existing RRT variants, including RRT-Star and RRT-Connect, offer improvements but can be further optimized.

Purpose of the Study:

• To propose and evaluate a Multi-Strategy Fusion RRT (MSF-RRT) algorithm for enhanced path planning.
• To address the limitations of traditional RRT algorithms in terms of efficiency and path quality.

Main Methods:

• Implemented a target bias strategy to increase sampling probability in the target region.
• Introduced a bias expansion strategy for orderly expansion towards the target.
• Utilized an adaptive step size strategy adjusting expansion based on map complexity.
• Optimized the planned path using pruning and cubic B-spline curves.

Main Results:

• MSF-RRT reduced search time by 90.53%, path length by 16.84%, and node count by 88.43% compared to traditional RRT.
• Achieved average reductions of 79.33% in search time, 14.58% in path length, and 77.71% in node count versus RRT-Star.
• Demonstrated average reductions of 49.74% in search time, 14.89% in path length, and 68.74% in node count compared to RRT-Connect.
• The algorithm showed improved efficiency and performance, aligning with kinematic properties.

Conclusions:

• The MSF-RRT algorithm offers a significant improvement over existing RRT methods for path planning.
• It effectively addresses issues of search blindness, convergence speed, and path smoothness.
• MSF-RRT demonstrates superior performance across various map complexities and is suitable for applications requiring efficient and kinematically feasible paths.