Ask1: Development and Reinforcement Learning-Based Control of a Custom Quadruped Robot

TL;DR

This paper details the design and control of a custom quadruped robot, Ask1, using reinforcement learning to achieve robust terrain navigation without relying on traditional motion priors or reference trajectories.

Contribution

It introduces a novel RL-based control method for a custom quadruped robot, extending previous approaches and demonstrating real-world applicability and generalization across different robots.

Findings

Ask1 successfully navigates rugged terrains in real-world tests.

The RL method generalizes well between different robot morphologies.

Elimination of Adversarial Motion Priors simplifies the control architecture.

Abstract

In this work, we present the design, development, and experimental validation of a custom-built quadruped robot, Ask1. The Ask1 robot shares similar morphology with the Unitree Go1, but features custom hardware components and a different control architecture. We transfer and extend previous reinforcement learning (RL)-based control methods to the Ask1 robot, demonstrating the applicability of our approach in real-world scenarios. By eliminating the need for Adversarial Motion Priors (AMP) and reference trajectories, we introduce a novel reward function to guide the robot's motion style. We demonstrate the generalization capability of the proposed RL algorithm by training it on both the Go1 and Ask1 robots. Simulation and real-world experiments validate the effectiveness of this method, showing that Ask1, like the Go1, is capable of navigating various rugged terrains.