Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Reinforcement Learning-Based Model Matching in COBRA, a Slithering Snake Robot

Harin Kumar Nallaguntla

TL;DR

This work employs a reinforcement learning-based model identification method aimed at enhancing the accuracy of the dynamics for COBRA's dynamical model using experimental and simulated data, highlighting its potential to address sim-to-real gap in robot implementation.

Abstract

This work employs a reinforcement learning-based model identification method aimed at enhancing the accuracy of the dynamics for our snake robot, called COBRA. Leveraging gradient information and iterative optimization, the proposed approach refines the parameters of COBRA's dynamical model such as coefficient of friction and actuator parameters using experimental and simulated data. Experimental validation on the hardware platform demonstrates the efficacy of the proposed approach, highlighting its potential to address sim-to-real gap in robot implementation.

Reinforcement Learning-Based Model Matching in COBRA, a Slithering Snake Robot

TL;DR

This work employs a reinforcement learning-based model identification method aimed at enhancing the accuracy of the dynamics for COBRA's dynamical model using experimental and simulated data, highlighting its potential to address sim-to-real gap in robot implementation.

Abstract

This work employs a reinforcement learning-based model identification method aimed at enhancing the accuracy of the dynamics for our snake robot, called COBRA. Leveraging gradient information and iterative optimization, the proposed approach refines the parameters of COBRA's dynamical model such as coefficient of friction and actuator parameters using experimental and simulated data. Experimental validation on the hardware platform demonstrates the efficacy of the proposed approach, highlighting its potential to address sim-to-real gap in robot implementation.
Paper Structure (40 sections, 37 equations, 30 figures, 2 tables, 1 algorithm)

This paper contains 40 sections, 37 equations, 30 figures, 2 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. State-Of-The-Art Modeling Tools
  3. Sim-to-Real Transfer for RL-based Strategies
  4. Modeling Frameworks Background
  5. Objectives and Outline of Thesis
  6. Contributions
  7. Literature Review
  8. Central Pattern Generators
  9. Learning-based Strategies for Locomotion Tasks
  10. Simulation-to-Reality Transfer of Trained Policies
  11. Research Analysis
  12. Simulators and Physics Engines
  13. Physics Engines
  14. Open Dynamics Engine
  15. Constraint Force Mixing
  16. ...and 25 more sections

Figures (30)

  • Figure 1.1: Popular Solutions for Solving Sim2real Problem
  • Figure 1.2: Shows successful transfer of an RL-based policy on 1) ANYmal robot lee_learning_2020 2) & 6) Unitree A1 yang_learning_2022feng_genloco_2022 3) Hexapod li_learning_2020 4) Cassie li_reinforcement_2021 5) Sirius feng_genloco_2022 7) Mini Cheetah feng_genloco_2022 8) 7-DoF Yumi Robot chebotar_closing_2019 9) & 13) ANYmal C miki_learning_2022 10) ANYmal B rudin_learning_2022 11) Digit humanoid radosavovic_learning_2023 12) Soft Snake Robot liu_reinforcement_2023
  • Figure 1.3: Highlights the sim2real gap oberserved between COBRA and Webots simulator. The robot performs sidewinding motion in webots with an untuned simulator model. The red ball represents the position attained by the actual robot when executing analogous joint trajectories.
  • Figure 1.4: Illustrates sim-to-real gap observed in COBRA's head link trajectory while performing sidewinding gaits at frequencies of 0.35 Hz, 0.5 Hz, and 0.65 Hz respectively
  • Figure 1.5: Illustrates sim-to-real gap observed in COBRA's joint angles while performing sidewinding gaits at frequencies of 0.35 Hz, 0.5 Hz, and 0.65 Hz respectively
  • ...and 25 more figures