Gait generation for real-time quadruped locomotion
Published in RAPDASA-RobMech-PRASA-AMI, 2023
Recommended citation: PENDING https://www.matec-conferences.org/articles/matecconf/abs/2023/15/matecconf_rapdasa2023_04022/matecconf_rapdasa2023_04022.html
Abstract: Gait generation for a legged robot in real-time is no simple task. Typically, powerful processors are required to perform real-time trajectory optimisation techniques to generate the trajectories, or the platform requires full state feedback and controllers to track a pre-generated trajectory. Here we present a novel method for generating gaits in real-time by varying three parameters (RideHeight, StepLength, and StepHeight). Various successful walking experiments were performed with the developed quadrupedal platform, demonstrating that this method worked with minimal state feedback (only leg angles) and limited processing power (Teensy 4.0 microcontroller was used). The resulting gait was generated in real-time and maintained stability in open loop. The platform was capable of traversing obstacles (50mm x 50mm) and ascending or descending slopes (15 degrees).