CAREER: Trajectory Planning for Highly Dynamic Legged Robots in Complex Environments

University Of Illinois At Urbana-Champaign, Urbana IL

Investigators

Abstract

This Faculty Early Career Development Program (CAREER) project will improve the ability of legged robots to traverse complex environments. The ultimate objective of the project is to rival the ability of animals to navigate daunting obstacles in order to reach a desired goal. In contrast to standard quasi-static approaches to robot path planning, these bio-inspired methods will allow highly dynamic behavior -- that is, during its motion the robot will be able to pass through positions that would be impossibly unbalanced if the robot was standing still. This may occur as the robot makes a continuous series of jumps, without pausing at the intermediate points. Numerical efficiency, provided by a hierarchical decomposition into well-behaved subproblems, will allow the trajectories to be computed as they are needed. First, solutions will be generated based on a greatly simplified dynamic model, in terms of "global" quantities that describe the translation and rotation of a robot-fixed reference frame. Of particular interest are the force pulses generated during the instants when the jumping robot is in contact with other objects. Possible solution trajectories will be grouped by the number of hops required to reach the destination and the foot locations on the obstacles. Finally a detailed solution will incorporate the "shape" variables that describe the configuration of the robot components with respect to the robot-fixed frame. The findings from this study will help lead to the creation of legged robots for use in inhospitable or dangerous real-world environments, including for disaster response, military applications, and exploration of inaccessible or inhospitable locations. The project will also build upon the appeal of agile legged robots for education and outreach to high school, undergraduate, and graduate students, including underrepresented minorities, to encourage the development of the next generation of engineers. The objective of this study is to establish close connections between the mathematical modeling of legged robot systems, trajectory planning strategies, and the hardware design principles needed to create highly dynamic legged robots navigating complex unstructured environments. To realize the goal, this project will develop a hierarchical understanding of the dynamic behavior of position variables (global position and orientation) and shape variables (internal configuration) under the application of contact forces. Based on this understanding, a multilevel process will be sought for designing a collision-free feasible trajectory for the position variables, and then deforming the trajectory while gradually taking into account the effects of the shape variables. This will be achieved by (1) introducing model structures for legged robot systems that provide a hierarchical understanding and abstraction of complex dynamics, (2) establishing a trajectory-planning framework that combines approaches from convex optimization and numerical continuation methods by utilizing the newly obtained model structures, and (3) exploring hardware design principles that are tightly integrated with the trajectory-planning framework for rigorous validation and demonstration of the methods using torque-controlled robotic quadrupeds. Realization of these aims will lay the foundation for a new class of trajectory-planning problems for the locomotion of highly dynamic legged robots navigating complex environments. 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.