Integrating Whole-Body Occupancy and Manipulability Map into Inverse Reachability Maps for Quality-Driven Base Positioning

Mobile manipulators—industrial robots like the UR5 mounted on mobile platforms—promise transformative flexibility for manufacturing and logistics. However, mobility introduces a critical challenge: determining where to position the base such that the manipulator executes multi-target trajectories with collision-free, kinematically optimal configurations. Existing inverse reachability methods treat collision detection as post-processing validation, requiring expensive iterative checking that becomes prohibitive in cluttered environments. This thesis introduces the Inverse Hybrid Dynamic Reachability Map (iHDRM) framework, fundamentally integrating whole-body collision awareness through a precomputed Occupation Map. For every valid joint configuration, the framework maps which workspace voxels the robot occupies, enabling collision detection through set intersection rather than geometric computation. This reduces collision checking from O(Nconfigs) to O(|Vobstacle|). For trajectories, the framework computes the intersection of candidate base poses across targets (Bcommon = B1 ∩ · · · ∩ BN ), filters by collision constraints, and selects optimal locations through manipulability-based quality aggregation, providing completeness, correctness, and optimality guarantees. Experimental validation demonstrates successful base placement for diverse trajectory shapes in cluttered environments. GPU-accelerated parallel computation enables practical offline map construction at high spatial resolution. The framework extends industrial manipulators with mobile flexibility while maintaining collision safety and kinematic quality, providing a practical path toward autonomous mobile manipulation.