In SurenaV, one of the main focuses of our project was to increase previous generation walking speed up to 1.4 km/hr (2 times faster than Surena IV). To reach this speed, DCM-based trajectory planning has been used to generate a CoM pattern. DCM has unstable second-order dynamics, so it needs to be controlled. In this project, we have implemented CoM trajectory planning, geometric inverse kinematic, and 5th order polynomial trajectory for ankles.
To the left, you can see a simulation of a Surena IV humanoid robot walking at a speed of 0.56 km/hr in Choreonoid. (Github repo: https://github.com/Kassra-sinaei/SurenaV)
The mentioned algorithm for planning CoM trajectory has several design parameters: step time, double support time, pelvis height, step length, etc. To achieve a robust high-speed gait, we are trying optimization with a Genetic Algorithm. PyBullet has been used for this part of simulation alongside ROS (Robotic Operating System). Before attempting to test the Surena V, we repeated the procedure with a model of Surena IV too. To the right, you can see part of the PyBullet simulation.
In the Surena V project, we are working on stabilizing the robot's gait using online sensors feedback. The robot has an IMU and Gyro sensor attached to its pelvis, FT sensors mounted at each sole and an RGBD camera. By implementing state estimators robot's COM velocity and ZMP tracking are possible. We use these reference points to control the robot's gait.
We used GA (single objective and multiobjective) to obtain the best walking parameters for the offline gait pattern generation; we used GA (single objective and multiobjective). Results and details of this work have been presented in the bellow paper.
