Next-Generation Autonomous Naval Humanoid Robots: Contact-aware Planning and Decision-making for Versatile Locomotion and Navigation
Abstract
Problem objective: My long-term career goal is to design autonomous, safe, and collaborative robot locomotion and manipulation systems in unstructured environments through advanced sensing, planning, and decision-making methodologies. Towards this goal, this ONR YIP proposal focuses on a contact-aware planning and decision-making (CAP-DM) framework for versatile humanoid locomotion and navigation in constrained, human-populated environments, particularly complex naval environments including submarine, ship deck, and offshore rescue boating. To date, the interplay between the complex environment and intricate locomotion dynamics is still under-exploredand poses a long-standing challenge on autonomous locomotion planning. To address this challenge, this proposal employs hierarchically integrated symbolic planning and trajectory optimization (TO) for enabling long-horizon, contact-rich locomotion and manipulation behaviors, sensing and adapting to complex terrain geometries and properties, and interacting with human-surrounded environments proactively. The proposed CAP-DM lies in the intersection of TO, formal methods, AI planning, physical contact sensing and modeling, and learning-based human motion prediction. A large focus of this proposal is placed on experimental evaluations to enable new legged multi-contact and navigation functionalities. This interdisciplinary proposal initiates a first yet significant step towards the larger goal of advancing a revolutionary understanding of interactive locomotion and navigation in complex environments.Technical approach and anticipated outcome: This ONR YIP proposal explores a novel task and motion planning (TAMP) framework for versatile and autonomous interaction between bipedal humanoid robots and constrained environments. The low-level motion planner of this TAMP framework employs distributed trajectory optimization to achieve consensus between centroidal and whole-body dynamics and augments it with multimodal contact sensing and modeling to address complex terrain maneuvering. Built upon this low-level planner, the high-level task planner uses signal temporal logic and AI planning methods for formal task specification design, long-horizon task decomposition,and interactive symbolic decision-making, particularly for handling diverse environmental events composed of irregular terrain and pedestrians. From the contact sensing perspective, a dynamically reconfigurable robotic foot is designed to modulate the contact surface shape, mechanical stiffness, and surface properties in real-time to passively compensate for locomotion uncertainties and hazards. To evaluate the efficiency of this framework, experiments are conducted for dynamically stable balancing on nonstatic surfaces, multi-contact locomotion and manipulation, and proactive interaction with pedestrians in a constrained environment. The proposed framework is pioneering in bridging the fundamental gap between TAMP and humanoid locomotion. If successful, this proposal will advanceversatile and autonomous legged navigation and pave the road for heterogeneous robot teaming in the PI s long-term career goal.Impacts on DoD: The impacts of the research program stem from a significant contribution to fundamental contact sensing, planning, and decision-making methodologies of versatile locomotion and navigation that are fundamentally important to address human interaction with autonomous robot team challenges that Navy will encounter over the next 20 years. The proposed research has great potential to advance various defense applications such as military operational maintenance, disaster first responders, surveillance in warfighting,and the wounded search and rescue. This proposal will not only foster algorithmic and experimental humanoid robotics technologies in Navy, but also revolutionize our fundamental understanding of robot autonomy and intelligence at the national security level. [Approved for Public Release]
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Feb 06, 2023
- Source ID
- N000142312223
Entities
People
- Ye Zhao
Organizations
- Georgia Tech Research Corporation
- Office of Naval Research
- United States Navy