Fast and Efficient Locomotion via Learned Gait Transitions
arXiv:2104.04644
Abstract
We focus on the problem of developing energy efficient controllers for quadrupedal robots. Animals can actively switch gaits at different speeds to lower their energy consumption. In this paper, we devise a hierarchical learning framework, in which distinctive locomotion gaits and natural gait transitions emerge automatically with a simple reward of energy minimization. We use evolutionary strategies (ES) to train a high-level gait policy that specifies gait patterns of each foot, while the low-level convex MPC controller optimizes the motor commands so that the robot can walk at a desired velocity using that gait pattern. We test our learning framework on a quadruped robot and demonstrate automatic gait transitions, from walking to trotting and to fly-trotting, as the robot increases its speed. We show that the learned hierarchical controller consumes much less energy across a wide range of locomotion speed than baseline controllers.
arXiv commentPublished in CoRL 2021. Website: Website: https://sites.google.com/view/fast-and-efficient Code: https://github.com/yxyang/fast_and_efficient
References in corpus (13)
- Soft Actor-Critic: Off-Policy Maximum Entropy Deep Reinforcement Learning with a Stochastic Actor
- Learning agile and dynamic motor skills for legged robots
- Evolution Strategies as a Scalable Alternative to Reinforcement Learning
- DeepMimic: Example-Guided Deep Reinforcement Learning of Physics-Based Character Skills
- Emergence of Locomotion Behaviours in Rich Environments
- Highly Dynamic Quadruped Locomotion via Whole-Body Impulse Control and Model Predictive Control
- Sim-to-Real: Learning Agile Locomotion For Quadruped Robots
- Policies Modulating Trajectory Generators
- Learning a Contact-Adaptive Controller for Robust, Efficient Legged Locomotion
- Multi-Agent Manipulation via Locomotion using Hierarchical Sim2Real
- From Pixels to Legs: Hierarchical Learning of Quadruped Locomotion
- Contrasting Exploration in Parameter and Action Space: A Zeroth-Order Optimization Perspective
- Learning Task Space Actions for Bipedal Locomotion