Vortragsdetails

Reinforcement Learning for Ground Force Prediction in Unstructured Environments

In robotic navigation, vision is essential for understanding unstructured environments and enabling effective locomotion. This presentation highlights the significance of vision-driven decision-making for footstep planning in both quadruped and biped robots. Existing methodologies are reviewed, focusing on approaches utilizing mixed integer programming (MIP) and reinforcement learning (RL) to address challenges in predicting ground forces.

The proposed strategy employs direct RL in conjunction with centroidal dynamics to enhance ground force predictions. These predictions will be incorporated into a Model Predictive Control (MPC) framework, where the actions generated by the policy will guide real-time decision-making. This integration aims to facilitate the transfer of the trained policy to a physical robotic platform. By advancing the state of robotic navigation, this approach seeks to improve the stability and adaptability of robots operating in complex, real-world terrains.

In der Regel sind die Vorträge Teil von Lehrveranstaltungsreihen der Universität Bremen und nicht frei zugänglich. Bei Interesse wird um Rücksprache mit dem Sekretariat unter sek-ric(at)dfki.de gebeten.

© DFKI GmbH
zuletzt geändert am 31.03.2023