Maneuvering over Deformable Terrain: Long-horizon Task and Motion Planning of Bipedal Locomotion via Contact Sensing and Terrain Adaptation

The main aim of this grant is to enhance the planning abilities of robots with legs. Specifically, the team will equip robots with the necessary tools to walk on challenging and ever-changing surfaces, like construction sites filled with loose rocks and debris. This task is quite difficult because robots with legs have many moving parts, and they struggle to plan their movements for such complex terrain.

It becomes even more challenging when the ground is soft or unstable, as the robot must maintain its balance and control how it interacts with the surface. To achieve this goal, the PIs aim to enable robots to plan their actions in advance and adjust in real-time while walking on demanding surfaces. They are taking inspiration from biology to design better robots while bringing together different fields of study, such as understanding the dynamics of robots with legs and how they make contact with surfaces, as well as planning tasks and movements.

This research will train a diverse group of students, including undergraduate students through the Georgia Tech Vertically Integrated Program, and underrepresented high-school students via the Engaging New Generation at Georgia Tech through Engineering & Science (ENGAGES) program. In addition, this award will support the inclusion of a graduate-level curriculum initiative integrated into a cross-disciplinary class to train PhD students at the interface of physics and robotics and broaden participation in computing, engineering, and science disciplines.

If successful, the outcomes of this research could lead to the deployment of legged robots in various fields, including construction, agriculture, and urban environments. Ultimately, the project's overarching goal is to enhance the capabilities of bipedal robots, enabling them to navigate challenging terrains and contribute to various real-world applications, thereby making our lives easier and safer.

Long-horizon, real-time, versatile bipedal robot locomotion over highly complex and deformable terrain is largely underexplored. This research aims to integrate deformable terrain physics, adaptive contact mechanism design, and multi-modal contact sensing into a unified task and motion planning framework and enable bipedal legged machines to maneuver over various harsh terrain.

Particularly, this award will employ a convergent approach including: (i) multi-modal terrain contact mechanism for austere terrain classification and adaptation, (ii) physics-based granular terrain dynamics modeling and deformable patch contact modeling, and (iii) long-horizon planning algorithms solving terrain-adaptive distributed trajectory optimization while symbolically reasoning about versatile gaits and contact decisions. The PIs will uncover fundamental locomotion principles through extensive and representative robot-terrain interaction experiments at various spatiotemporal scales.

This research will enable breakthroughs in the understanding of integrating complex terrain dynamics and rich contact sensing into bipedal locomotion planning and decision-making in a principled manner.

This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.