Pronking
CoM displacement: 80cm per step.
We present CAJun, a novel hierarchical learning and control framework that enables legged robots to jump continuously with adaptive jumping distances.
CAJun consists of a high-level centroidal policy and a low-level leg controller. In particular, we use reinforcement learning (RL) to train the centroidal policy, which specifies the gait timing, base velocity, and swing foot position for the leg controller. The leg controller optimizes motor commands for the swing and stance legs according to the gait timing to track the swing foot target and base velocity commands using optimal control. Additionally, we reformulate the stance leg optimizer in the leg controller to speed up policy training by an order of magnitude. Our system combines the versatility of learning with the robustness of optimal control.
By combining RL with optimal control methods, our system achieves the versatility of learning while enjoys the robustness from control methods, making it easily transferable to real robots. We show that after 20 minutes of training on a single GPU, CAJun can achieve continuous, long jumps with adaptive distances on a Go1 robot with small sim-to-real gaps. Moreover, the robot can jump across gaps with a maximum width of 70cm, which is over 40% wider than existing methods.
We implement two dynamic jumping gaits using CAJun. In Pronking, all legs jump simultaneously. In Bounding, the front and rear legs jump in alternating order.
CoM displacement: 80cm per step.
CoM displacement: 70cm per step.
In CAJun, the robot can adapt its jumping distance based on user command. For example, when the distance command shifts between 0.3m and 1m, the robot can alternate between long and short jumps.
Trajectories of the base, front leg and rear leg.
The hierarchical design of CAJun makes it robust under external perturbations. For example, the robot is able to recover from an agressive backward leash-pull and resume jumping.
The robot suffers from foot slipping when entering grass from asphalt, but is able to resume normal jumping after a few steps.
We compare CAJun with end-to-end RL methods trained under a similar setup, and find that end-to-end method cannot jump as far and suffers from a larger sim-to-real gap.
CAJun: 4.76m, E2E: 1.67m
CAJun: 4.34m, E2E: 3.47m
@article{yang2023cajun,
title={CAJun: Continuous Adaptive Jumping using a Learned Centroidal Controller},
author={Yang, Yuxiang and Shi, Guanya and Meng, Xiangyun and Yu, Wenhao and Zhang, Tingnan and Tan, Jie and Boots, Byron},
journal={arXiv preprint arXiv:2306.09557},
year={2023}
}