Reinforcement Learning With Evolutionary Trajectory Generator: A General Approach for Quadrupedal Locomotion

نویسندگان

چکیده

Recently reinforcement learning (RL) has emerged as a promising approach for quadrupedal locomotion, which can save the manual effort in conventional approaches such designing skill-specific controllers. However, due to complex nonlinear dynamics robots and reward sparsity, it is still difficult RL learn effective gaits from scratch, especially challenging tasks walking over balance beam. To alleviate difficulty, we propose novel RL-based that contains an evolutionary foot trajectory generator. Unlike prior methods use fixed generator, generator continually optimizes shape of output given task, providing diversified motion priors guide policy learning. The trained with residual control signals fit different gaits. We then optimize network alternatively stabilize training share exploratory data improve sample efficiency. As result, our solve range simulation by including on beam crawling through cave. further verify effectiveness approach, deploy controller learned 12-DoF robot, successfully traverse scenarios efficient provide video show YouTube. 1 1 [Online]. Available: youtube.com/watch?v=hgBLR09MEOw, code available Github: github.com/PaddlePaddle/PaddleRobotics

برای دانلود رایگان متن کامل این مقاله و بیش از 32 میلیون مقاله دیگر ابتدا ثبت نام کنید

اگر عضو سایت هستید لطفا وارد حساب کاربری خود شوید

منابع مشابه

Machine Learning for Fast Quadrupedal Locomotion

For a robot, the ability to get from one place to another is one of the most basic skills. However, locomotion on legged robots is a challenging multidimensional control problem. This paper presents a machine learning approach to legged locomotion, with all training done on the physical robots. The main contributions are a specification of our fully automated learning environment and a detailed...

متن کامل

Central Pattern Generator Based Omnidirectional Locomotion for Quadrupedal Robotics

Trajectory generation for quadruped robots is a challenging task since they are underactuated systems which must balance using sensory feedback and satisfy ground contact constraints. There is a substantial body of evidence that many animals use central pattern generators (CPGs) for generating joint trajectories and regulation through sensory feedback. However, CPG models formulated in the join...

متن کامل

The Development of Modular Evolutionary Networks for Quadrupedal Locomotion

Artificial Neural Networks have so far failed to produce a convincing route to Robotic Intelligence. Training and Organizational Algorithms (such as Evolutionary Algorithms) are presently not flexible or sophisticated enough to configure large networks which fuse data from different sensory domains in a complex and changing environment. The approach outlined here is different in that it allows ...

متن کامل

Evolutionary Reinforcement Learning for Simulated Locomotion of a Robot with a Two-link Arm

In this paper we present a neural controller design for robots using an evolutionary reinforcement learning system that is suitable for learning through interaction. The method starts with networks of minimal structures determined by the domain expert and increases their complexity along the evolution path. It uses a nature inspired meta-level evolutionary process where new structures are explo...

متن کامل

High speed locomotion for a quadrupedal microrobot

Research over the past several decades has elucidated some of the mechanisms behind high speed, highly efficient and robust locomotion in insects such as cockroaches. Roboticists have used this information to create biologically-inspired machines capable of running, jumping, and climbing robustly over a variety of terrains. To date, little work has been done to develop an at-scale insect-inspir...

متن کامل

ذخیره در منابع من


  با ذخیره ی این منبع در منابع من، دسترسی به آن را برای استفاده های بعدی آسان تر کنید

ژورنال

عنوان ژورنال: IEEE robotics and automation letters

سال: 2022

ISSN: ['2377-3766']

DOI: https://doi.org/10.1109/lra.2022.3145495