PathCover's RISP speeds up robot navigation 10x
New algorithm cuts point cloud processing time by 10x for autonomous robots...
A University of Delaware team (Kunal Narkhede, Abhijeet Kulkarni, Guoquan Huang, Ioannis Poulakakis) has developed PathCover, a groundbreaking framework for autonomous robot navigation that leverages RISP (Randomized Iterative Space Partitioning) to process raw point cloud data in expected linear time.
The core innovation solves a critical bottleneck in robotics: traditional corridor generators can't meet real-time computation requirements for trajectory planning. PathCover addresses this by creating sequences of overlapping, obstacle-free convex polytopes directly from LiDAR scans. Benchmarks show a 10x speedup over state-of-the-art methods while maintaining comparable corridor volumes. The team validated the pipeline through high-fidelity quadrotor simulations and physical deployment on a quadrupedal robot navigating constrained environments using live LiDAR perception.
- PathCover uses RISP algorithm to process point clouds in expected linear time, 10x faster than existing methods
- Validated on synthetic and real-world LiDAR datasets with identical corridor quality
- Deployed on both quadrotor and quadrupedal robots for real-world navigation
Why It Matters
Enables real-time obstacle avoidance for autonomous robots in dynamic environments