3-D Contouring Control of Fully Actuated Bipedal Walking
- Our research group has been selected to receive an award of $30,000 to further our participation in MassTLC's 5G Robotics Challenge, sponsored by Verizon and Ericsson!! (01/2019)
In this research project, we develop nonlinear control strategies to realize exponential tracking of the desired position trajectory on a walking surface for fully actuated bipedal robots. Satisfactory global position tracking is important for accomplishing complex tasks such as multi-agent coordination and obstacle avoidance. However, such capabilities have not been fully explored with nonlinear control based walking strategies. To enable exponential position tracking on a walking surface, our research attempts to adapt contouring control from machining tasks such as cutting and milling to bipedal robotic walking. A contour is defined as a 1-D geometric path on the walking surface, and the contouring control problem will be decomposed into two subproblems. One is a stabilization problem with the objective of realizing exponential convergence to the shape of the desired contour. The other is a position tracking problem with the objective of realizing exponential convergence to the desired position trajectory along the desired contour. Inspired by the Hybrid-Zero-Dynamics Framework, our contouring control strategy is synthesized based on full-order hybrid walking dynamics, input-output linearization, and formal stability analysis. Through achieving exponential global position tracking, the results from this project will greatly enhance the walking versatility for fully actuated bipedal robots.
Publications:
- Y. Gu, B. Yao, C. S. G. Lee, “Bipedal Gait Recharacterization and Walking Encoding Generalization for Stable Bipedal walking,” in Proc. of the IEEE International Conference on Robotics and Automation (ICRA), 2016, pp. 1788-1793. pdf
- Y. Gu, B. Yao, C. S. G. Lee, “Exponential Stabilization of Fully Actuated Bipedal Robotic Walking,” ASME Journal of Dynamic Systems, Measurements, and Control, in press. pdf
- Y. Gu, B. Yao, C. S. G. Lee, “Straight-Line Contouring Control of Fully Actuated 3-D Bipedal Robotic Walking,” submitted to American Control Conference (ACC) for publication. pdf
Integrated Non-periodic Planning and Exponential Stabilization
In this study, we propose a unified planning and control framework for bipedal robotic walking over rough terrains. The overall framework is composed of three layers:
- A high-level hybrid phase-space planner for designing highly agile locomotion behaviors;
- A middle-layer trajectory generation interface;
- A low-level controller capable of robustly stabilizing the planned locomotion behaviors.
Publication:
- Zhao, Y. Gu, “An Integrated Non-periodic Planning and Exponential Stabilization Framework of Dynamic Legged Locomotion,” submitted to ACM International Conference on Hybrid Systems: Computation and Control (HSCC) for publication. pdf
Time-dependent Orbital Stabilization of Underactuated Bipedal Walking
In this research, we study orbitally exponential stabilization of underactuated bipedal walking with time-based control. Orbital stabilization of underactuated walking has been extensively studied through phase-based stabilization, but implementation imperfection such as sensor noise may cause practical issues for phase-based stabilization. These issues can be effectively solved by time-based stabilization, which will result in a hybrid, aperiodically varying control system whose stability is much more difficult to evaluate. To tackle this research problem, we aim to establish sufficient conditions to evaluate the orbitally exponential stability for underactuated walking under time-based control.
Publication:
- Y. Gu, B. Yao, C. S. G. Lee, “Time-dependent Orbital Stabilization of Underactuated Bipedal Walking,” in Proc. of American Control Conference (ACC), 2017, pp. 4858-4863. pdf
Biologically Inspired Bipedal Walking
This research investigates energy optimization of bipedal robotic walking based on observations and analysis of human walking. Due to high energy cost and limited battery life, today's bipedal robots can only operate for a short period of time on batteries, which are far from satisfactory for real-world applications. In contrast, human walking is highly efficient. As our previous research has shown that a bipedal robot’s posture affects the required joint torques to reach certain CoM acceleration, we will study biologically inspired energy optimization by:
- Collecting and analyzing normal human walking data;
- Extracting the posture selection strategy of human walking.
Publication:
- Y. Gu, C. S. G. Lee, B. Yao, “Feasible Center of Mass Dynamic Manipulability of Humanoid Robots,” in Proc. of the IEEE International Conference on Robotics and Automation (ICRA), 2015, pp. 5082-5087. pdf