Learning Modular Robot Control Policies

05/20/2021
by   Julian Whitman, et al.
6

To make a modular robotic system both capable and scalable, the controller must be equally as modular as the mechanism. Given the large number of designs that can be generated from even a small set of modules, it becomes impractical to create a new system-wide controller for each design. Instead, we construct a modular control policy that handles a broad class of designs. We take the view that a module is both form and function, i.e. both mechanism and controller. As the modules are physically re-configured, the policy automatically re-configures to match the kinematic structure. This novel policy is trained with a new model-based reinforcement learning algorithm, which interleaves model learning and trajectory optimization to guide policy learning for multiple designs simultaneously. Training the policy on a varied set of designs teaches it how to adapt its behavior to the design. We show that the policy can then generalize to a larger set of designs not seen during training. We demonstrate one policy controlling many designs with different combinations of legs and wheels to locomote both in simulation and on real robots.

READ FULL TEXT

page 1

page 12

research
10/31/2022

Learning Modular Robot Visual-motor Locomotion Policies

Control policy learning for modular robot locomotion has previously been...
research
10/31/2022

Learning Modular Robot Locomotion from Demonstrations

Modular robots can be reconfigured to create a variety of designs from a...
research
09/08/2023

Sample-Efficient Co-Design of Robotic Agents Using Multi-fidelity Training on Universal Policy Network

Co-design involves simultaneously optimizing the controller and agents p...
research
07/24/2022

N-LIMB: Neural Limb Optimization for Efficient Morphological Design

A robot's ability to complete a task is heavily dependent on its physica...
research
07/09/2020

One Policy to Control Them All: Shared Modular Policies for Agent-Agnostic Control

Reinforcement learning is typically concerned with learning control poli...
research
10/03/2020

Beyond Tabula-Rasa: a Modular Reinforcement Learning Approach for Physically Embedded 3D Sokoban

Intelligent robots need to achieve abstract objectives using concrete, s...
research
03/26/2019

A programmable actuator for combined motion and connection and its application to modular robot

This paper proposes a new type of actuator at millimeter scale, which is...

Please sign up or login with your details

Forgot password? Click here to reset