2020/07/01 by Digby Chappell, Ke Wang, Chappell, Digby +3
Biochemistry, Genetics and Molecular Biology · Engineering · #FOS: Computer and information sciences #FOS: Electrical engineering #Prosthetics and Rehabilitation Robotics #Robotic Locomotion and Control #Robotics (cs.RO) #Systems and Control (eess.SY) #Zebrafish Biomedical Research Applications #electronic engineering #information engineering
paper · pdf · doi:10.48550/arxiv.2007.00385
openalex publication_date 2020/07/01 · openalex created_date 2022/07/26 · openalex updated_date 2026/07/28
Online footstep planning is essential for bipedal walking robots to be able\nto walk in the presence of disturbances. Until recently this has been achieved\nby only optimizing the placement of the footstep, keeping the duration of the\nstep constant. In this paper we introduce a footstep planner capable of\noptimizing footstep placement and timing in real-time by asynchronously\ncombining two optimizers, which we refer to as asynchronous real-time\noptimization (ARTO). The first optimizer which runs at approximately 25 Hz,\nutilizes a fourth-order Runge-Kutta (RK4) method to accurately approximate the\ndynamics of the linear inverted pendulum (LIP) model for bipedal walking, then\nuses non-linear optimization to find optimal footsteps and duration at a lower\nfrequency. The second optimizer that runs at approximately 250 Hz, uses\nanalytical gradients derived from the full dynamics of the LIP model and\nconstraint penalty terms to perform gradient descent, which finds approximately\noptimal footstep placement and timing at a higher frequency. By combining the\ntwo optimizers asynchronously, ARTO has the benefits of fast reactions to\ndisturbances from the gradient descent optimizer, accurate solutions that avoid\nlocal optima from the RK4 optimizer, and increases the probability that a\nfeasible solution will be found from the two optimizers. Experimentally, we\nshow that ARTO is able to recover from considerably larger pushes and produces\nfeasible solutions to larger reference velocity changes than a standard\nfootstep location optimizer, and outperforms using just the RK4 optimizer\nalone.\n