• Corpus ID: 233209774

Fast and Efficient Locomotion via Learned Gait Transitions

@article{Yang2021FastAE,
  title={Fast and Efficient Locomotion via Learned Gait Transitions},
  author={Yuxiang Yang and Tingnan Zhang and Erwin Coumans and Jie Tan and Byron Boots},
  journal={ArXiv},
  year={2021},
  volume={abs/2104.04644}
}
: 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… 

Coupling Vision and Proprioception for Navigation of Legged Robots

This work exploits the complementary strengths of vision and pro-prioception to develop a point-goal navigation system for legged robots, called VP-Nav, which shows superior performance compared to wheeled robot baselines, and ablation studies which have disjoint high-level planning and low-level control.

Coupling Vision and Proprioception for Navigation of Legged Robots

This work exploits the complementary strengths of vision and proprioception to develop a point-goal navigation system for legged robots, called VP-Nav, which shows superior performance compared to wheeled robot baselines, and ablation studies which have disjoint high-level planning and low-level control.

Coupling Vision and Proprioception for Navigation of Legged Robots

This work exploits the complementary strengths of vision and proprioception to develop a point-goal navigation system for legged robots, called VP-Nav, which shows superior performance compared to wheeled robot baselines, and ablation studies which have disjoint high-level planning and low-level control.

Learning Low-Frequency Motion Control for Robust and Dynamic Robot Locomotion

It is shown that low-frequency policies are less sensitive to actuation latencies and variations in system dynamics, and to the extent that a successful sim-to-real transfer can be performed even without any dynamics randomization or actuation modeling.

Advanced Skills by Learning Locomotion and Local Navigation End-to-End

This work proposes to solve the complete problem of local navigation on challenging environments with legged robots by training an end-to-end policy with deep reinforcement learning and shows that the time dependence of the task reward is critical to successfully learn these new behaviors.

Learning and Deploying Robust Locomotion Policies with Minimal Dynamics Randomization

This work investigates a simple strategy of random force injection (RFI) to perturb system dynamics during training and shows that the application of random forces enables it to emulate dynamics randomization, which allows it to obtain locomotion policies that are robust to variations in system dynamics.

Learning to Walk by Steering: Perceptive Quadrupedal Locomotion in Dynamic Environments

A hierarchical learning framework, named PRELUDE, is presented, which decomposes the problem of perceptive locomotion into high-level decision-making to predict navigation commands and low-level gait generation to realize the target commands.

Model-free End-to-end Learning of Agile Quadrupedal Locomotion over Challenging Terrain

The method of combining reinforcement learning with inverse kinematics to generate foot trajectories in the model-free mode is adopted and the standard and smooth trot gait only depends on the robot’s proprioceptive observations, and robot can also achieve steady walking by curriculum learning in the case of with no camera obtaining terrain's information.

A Walk in the Park: Learning to Walk in 20 Minutes With Model-Free Reinforcement Learning

This work demonstrates that the recent advancements in machine learning algorithms and libraries combined with a carefully tuned robot controller lead to learning quadruped locomotion in only 20 minutes in the real world.

Hierarchical Reinforcement Learning for Precise Soccer Shooting Skills using a Quadrupedal Robot

A hierarchical framework is proposed that leverages deep reinforcement learning to train a robust motion control policy that can track arbitrary motions and a planning policy to decide the desired kicking motion to shoot a soccer ball to a target.

References

SHOWING 1-10 OF 58 REFERENCES

Learning agile and dynamic motor skills for legged robots

This work introduces a method for training a neural network policy in simulation and transferring it to a state-of-the-art legged system, thereby leveraging fast, automated, and cost-effective data generation schemes.

Dynamic Locomotion in the MIT Cheetah 3 Through Convex Model-Predictive Control

This paper presents an implementation of model predictive control (MPC) to determine ground reaction forces for a torque-controlled quadruped robot, capable of robust locomotion at a variety of speeds.

Variational Contact-Implicit Trajectory Optimization

A new family of trajectory optimization algorithms are proposed that leverage ideas from discrete variational mechanics to derive higher-order generalizations of the classic time-stepping method of Stewart and Trinkle, and it is possible to perform contact-implicit trajectory optimization with significantly higher accuracy.

A direct method for trajectory optimization of rigid bodies through contact

A novel method for trajectory planning of rigid-body systems that contact their environment through inelastic impacts and Coulomb friction is presented, which eliminates the requirement for a priori mode ordering.

Sim-to-Real: Learning Agile Locomotion For Quadruped Robots

This system can learn quadruped locomotion from scratch using simple reward signals and users can provide an open loop reference to guide the learning process when more control over the learned gait is needed.

Blind Bipedal Stair Traversal via Sim-to-Real Reinforcement Learning

This paper shows that sim-to-real reinforcement learning (RL) can achieve robust locomotion over stair-like terrain on the bipedal robot Cassie using only proprioceptive feedback, and only requires modifying an existing flat-terrain training RL framework to include stair- like terrain randomization, without any changes in reward function.

From Pixels to Legs: Hierarchical Learning of Quadruped Locomotion

It is shown that hierarchical policies can concurrently learn to locomote and navigate in these environments, and show they are more efficient than non-hierarchical neural network policies.

Learning Task Space Actions for Bipedal Locomotion

This work integrates learning a task space policy with a model-based inverse dynamics controller, which translates task space actions into joint-level controls and demonstrates the approach in simulation and shows that the learned policies are able to transfer to the real bipedal robot Cassie.

Learning quadrupedal locomotion over challenging terrain

The presented work indicates that robust locomotion in natural environments can be achieved by training in simple domains.

Learning a Contact-Adaptive Controller for Robust, Efficient Legged Locomotion

A hierarchical framework that combines model-based control and reinforcement learning (RL) to synthesize robust controllers for a quadruped (the Unitree Laikago) that can adapt to challenging environmental changes on the fly.
...