Daniel Seita

This work presents HOTICE, a whole-body humanoid learning framework for transporting objects through cluttered environments, and introduces Humanoid-Object Decoupled Potential Fields, which jointly encode collision-avoidance guidance for the robot and the carried object, enabling coordinated, obstacle-aware motion for both.

Generating tactile signals from vision could revolutionize how robots manipulate objects in complex environments where touch is essential but hard to measure.

Prioritizing resource-aware grasps, where finger usage is explicitly modeled, dramatically improves the success rate of sequential dexterous manipulation tasks.

Grounding reward learning in natural language rationales makes policies 2x more robust to spurious correlations and distribution shifts.