Symmetry-Informed Reinforcement Learning and its Application to Low-Level Attitude Control of Quadrotors
- 27 February 2023
- journal article
- research article
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Artificial Intelligence
- Vol. 5 (3) , 1147-1161
Abstract
Symmetry is ubiquitous in nature, physics, and mathematics. However, a classical symmetry-agnostic Reinforcement Learning (RL) approach cannot guarantee to respect symmetry. Researchers have shown that if the symmetry of a system cannot be respected, the performance of a symmetry-agnostic RL approach can be inhibited. To this end, this paper develops a generally applicable Neural Network (NN) module with symmetry that can enforce the symmetry of a system to be respected. Based on the NN module with symmetry, this paper proposes a symmetry-informed Model-Based RL (MBRL) approach that respects symmetry and improves data efficiency. The symmetry-informed MBRL approach is applied to the attitude control of a quadrotor in simulation to evaluate the effectiveness of the approach. The simulation results show that the data efficiency of the symmetry-informed MBRL approach is much superior to that of a symmetry-agnostic MBRL approach. An NN module with symmetry can respect the symmetry of a quadrotor while a naive NN cannot enforce the symmetry of a quadrotor to be respected.Keywords
Funding Information
- Guangdong Basic and Applied Basic Research Foundation for Young Scientists (2021A1515110021)
- Shenzhen Science and Technology (RCBS20210609103819024)
- Research Foundation for Advanced Talents
- Harbin Institute of Technology Shenzhen (CA11409019)
- National Natural Science Foundation of China (12172109, 12172111)
- Natural Science and Engineering (2022A1515011492)
This publication has 54 references indexed in Scilit:
- Design and implementation of a low-complexity flight controller for a quadrotor UAVInternational Journal of Dynamics and Control, 2022
- Tracking Control for a Quadrotor via Dynamic Surface Control and Adaptive Dynamic ProgrammingInternational Journal of Control, Automation, and Systems, 2022
- Learning Humanoid Robot Running Motions with Symmetry Incentive through Proximal Policy OptimizationJournal of Intelligent & Robotic Systems, 2021
- A Survey of Bayesian Calibration and Physics-informed Neural Networks in Scientific ModelingArchives of Computational Methods in Engineering, 2021
- A survey on deep learning in medicine: Why, how and when?Information Fusion, 2021
- A velocity-free adaptive RISE-based trajectory tracking approach for quadrotors with desired model compensationAerospace Science and Technology, 2019
- Physics-informed neural networks: A deep learning framework for solving forward and inverse problems involving nonlinear partial differential equationsJournal of Computational Physics, 2019
- RISE and disturbance compensation based trajectory tracking control for a quadrotor UAV without velocity measurementsAerospace Science and Technology, 2018
- Introduction to Multicopter Design and Control2017
- Machine learning strategies for systems with invariance propertiesJournal of Computational Physics, 2016