WAM-Plan: Contact-Rich Global Planning for Whole-Arm Manipulation and Control

WAM-Plan is a first-author paper in preparation combining offline trajectory optimization with online Model Predictive Control to enable safe whole-arm manipulation in contact-rich human environments. Rather than avoiding all contact, the planner globally reasons about the robot's ability to regulate forces at potential contact locations, producing trajectories that keep the arm in configurations where contact-aware MPC can effectively manage interactions. This is enabled by distributed tactile sensing across the robot arm. By planning for contact rather than against it, the approach expands the robot's usable workspace without sacrificing safety.

This experience has been absolutely crucial with regards to understanding and experiencing the full cycle of research from a first author perspective. Working on my own research project with complete independence over the course of a year with weekly advisor meetings made it very clear that I greatly enjoy research and would like to pursue a PhD. I performed literature review, narrowed down key exploration topics and strategies, built code infrastructure, derived and implemented various original algorithms, tested my methods, hit many dead ends, adjusted my focus, iterated, and wrote reports on my findings.

(ours) RRT Baseline