Path Planning for Non-Circular, Non-Holonomic Robots in Highly Cluttered Environments.

Area of Science:

• Robotics
• Artificial Intelligence
• Path Planning Algorithms

Background:

• Autonomous mobile robots require efficient path planning in large, cluttered environments.
• High-resolution state spaces are necessary for navigating narrow passages and near obstacles.
• Computational cost of collision detection increases exponentially with resolution.

Purpose of the Study:

• To develop a novel algorithm for feasible path planning for slender autonomous mobile robots.
• To address the computational challenges of high-resolution state spaces in path planning.
• To enhance real-time performance and support dynamic navigation in changing environments.

Main Methods:

• Graph search on a high-resolution, kinodynamic-feasible lattice state space.
• Utilizing offline precomputed maps updated via partial convolutions to reduce collision detection costs.
• Proposing alternative fast convolution methods for varying environmental clutter.

Main Results:

• Significant reduction in computational expense for path search algorithms.
• Improved run-time performance compared to traditional methods.
• Demonstrated support for dynamic path planning in evolving environments.

Conclusions:

• The presented algorithm offers an efficient solution for autonomous robot path planning in complex environments.
• The use of precomputed maps and partial convolutions effectively mitigates computational bottlenecks.
• The approach enhances the robot's ability to navigate dynamic and cluttered spaces effectively.