Conflict-Based Lazy Search for Fast Multi-Manipulator Planning

This paper introduces Conflict-Based Lazy Search (CBLS) for planning with multiple robot manipulators in cluttered workspaces. CBLS builds on Conflict-Based Search by precomputing a controlled-sparsity lazily evaluated graph for one manipulator and using Lazy Edge-based A* (LEA*) for single-agent pathfinding. Because collision and feasibility checks on manipulator-planning edges are expensive, LEA* delays those evaluations and maintains an edge queue. The authors report improved performance over standard CBS and RRT-Connect, while claiming optimal vertex efficiency for LEA* and better edge efficiency than A*.