AI-Driven Constrained Optimal Control for Bi-manual Loco-Manipulation

Research Project, 2024 – 2029

The goal of this project is to develop an AI-driven constrained optimal control framework for bi-manual loco-manipulation with mobile robots of humanoid form factor. This framework will be able to efficiently deal with large number of geometric constraints (in order of hundreds) including both external constraints coming from bi-manual manipulation task and internal constraints coming from closed loop transmissions used in the robot design. The optimal control framework will be designed such that it can be used for both kino-dynamic motion planning as well as stabilizing control considering the mobile/floating base. Lastly, we propose to use learning from demonstration to automatically identify the geometric constraints for the bi-manual manipulation task and seamlessly switch between different tasks over time. The framework will be experimentally evaluated on Mobile YuMi dual arm platform at the WARA Robotics Lab at ABB Coorporate Research Center and on the series-parallel hybrid RH5 Manus humanoid at DFKI Robotics Innovation Center in Germany for bi-manual assembly applications.