Legged Robot Estimation and Motion Generation with Global Environment Dynamics
Abstract
Approved for Public Release: Problem and Objectives: Legged robots have great potential to advance Naval application areas, including surveillance, emergency response, and inspection and maintenance. However, many Naval settings are globally dynamic environments (GDEs), environments that exhibit absolute dynamic motions in the inertial frame, such as moving ships and aircraft. For example, vessels in sea waves exhibit dynamics that are unknown, unpredictable, and persistently and rapidly changing, preventing reliable and safe robot function due to the high complexity of both the environment and robot dynamics. Safe legged robot locomotion and manipulation in GDEs are impeded by (a) reliance on onboard sensors to obtain movement estimates for the environment, a robot, and objects it interacts with and (b) an overly high planning burden resulting from uncertain and rapidly changing environment dynamics. The research objective is development of a robot estimation and motion generation framework that enables safe locomotion and manipulation inGDEs with unknown and rapidly changing motions. Proposed Approach: A five-part approach draws upon the integration of nonlinear control, theory of hybrid systems, optimization, and geometric control with first-of-their-kind physical experiments simulating shipboard movement: (1) derivation of an estimation approach by extending invariant filters to unknown GDE motions and combining the expanded filter with the existing certifiable object pose estimation method; (2) construction of a motion generation methodology based on the generalization of optimization fabrics to hybrid dynamical systems that include legged robots; (3) design of a footstep planner by applying Lyapunov and barrier function theories to the hybrid, time-varying, unactuated subsystem of a legged robot; (4) integration of the estimator and planner to develop a provably safe framework that explicitly considers the unknown GDE motion and associated complex robot dynamics; and (5) validation of theoretical and algorithmic results through field tests in collaboration with Navy. Outcomes: Three major outcomes are expected: (1) a real-time, provably accurate estimator that relies only on onboard sensors to simultaneously estimate robot, environment, and object movement; (2) a motion generator that translates desired high-level robot behaviors into local motions with fast reactivity, inherent convergence, and provable robustness; and (3) an estimation and motion generation framework that ensures provable safety under a broad spectrum of real-world uncertainties including unknown, rapidly and persistently changing GDE motions. The resulting estimator and motion generator will be valid for general legged robots with arbitrary forms and shapes and extendable to wheeled or tracked robots operating in GDEs with unknown motions, while the motion generator will also be valid for static environments and scale well to complex combinations of motion behaviors. Impact on DoD: Resulting foundationaltheories and algorithms will enable robotic systems with reliable movement capabilities in dynamic environments. Such robots could be empowered with high-level autonomous and intelligent behaviors to assist humans in critical Naval applications, such as facilitation of on-time maintenance and inspection of Navy vessels and assisting Marines in urban operations. Bio-inspired, multi-limb robotic locomotors could be developed to robustly and autonomously operate under diverse, unpredictive, and changing environments, including surf zones, beaches, sea ice, and underway ships. The framework also supports formation of heterogenous robot teams for long-termautonomy; by deploying robots onboard autonomous underwater, surface, and aerial vehicles, unmanned routine maintenance and inspection could be performed, enabling autonomous heterogenous teaming for prolonged Naval operations.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Nov 21, 2023
- Source ID
- N000142412028
Entities
People
- Yan Gu
Organizations
- Office of Naval Research
- Purdue University
- United States Navy