Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

ATRos: Learning Energy-Efficient Agile Locomotion for Wheeled-legged Robots

Jingyuan Sun, Hongyu Ji, Zihan Qu, Chaoran Wang, Mingyu Zhang

TL;DR

The paper tackles the challenge of energy-efficient, agile locomotion for wheeled-legged robots by introducing ATRos, an RL-based framework that combines a Predictive Response Policy Network (PRPN) with a PPO-based actor–critic to coordinate wheels and legs using proprioceptive inputs. It formulates the problem as a partially observable Markov decision process and employs an energy-aware reward to encourage efficient, stable, terrain-adaptive motion, achieving zero-shot sim-to-real transfer. The authors validate ATRos in simulation and on a real Go2W platform across grass and stair terrains, demonstrating improved energy efficiency and robust, agile performance with diverse terrain generalization. The work advances practical hybrid locomotion by balancing rolling efficiency and legged versatility, with implications for long-duration missions in real-world environments.

Abstract

Hybrid locomotion of wheeled-legged robots has recently attracted increasing attention due to their advantages of combining the agility of legged locomotion and the efficiency of wheeled motion. But along with expanded performance, the whole-body control of wheeled-legged robots remains challenging for hybrid locomotion. In this paper, we present ATRos, a reinforcement learning (RL)-based hybrid locomotion framework to achieve hybrid walking-driving motions on the wheeled-legged robot. Without giving predefined gait patterns, our planner aims to intelligently coordinate simultaneous wheel and leg movements, thereby achieving improved terrain adaptability and improved energy efficiency. Based on RL techniques, our approach constructs a prediction policy network that could estimate external environmental states from proprioceptive sensory information, and the outputs are then fed into an actor critic network to produce optimal joint commands. The feasibility of the proposed framework is validated through both simulations and real-world experiments across diverse terrains, including flat ground, stairs, and grassy surfaces. The hybrid locomotion framework shows robust performance over various unseen terrains, highlighting its generalization capability.

ATRos: Learning Energy-Efficient Agile Locomotion for Wheeled-legged Robots

TL;DR

The paper tackles the challenge of energy-efficient, agile locomotion for wheeled-legged robots by introducing ATRos, an RL-based framework that combines a Predictive Response Policy Network (PRPN) with a PPO-based actor–critic to coordinate wheels and legs using proprioceptive inputs. It formulates the problem as a partially observable Markov decision process and employs an energy-aware reward to encourage efficient, stable, terrain-adaptive motion, achieving zero-shot sim-to-real transfer. The authors validate ATRos in simulation and on a real Go2W platform across grass and stair terrains, demonstrating improved energy efficiency and robust, agile performance with diverse terrain generalization. The work advances practical hybrid locomotion by balancing rolling efficiency and legged versatility, with implications for long-duration missions in real-world environments.

Abstract

Hybrid locomotion of wheeled-legged robots has recently attracted increasing attention due to their advantages of combining the agility of legged locomotion and the efficiency of wheeled motion. But along with expanded performance, the whole-body control of wheeled-legged robots remains challenging for hybrid locomotion. In this paper, we present ATRos, a reinforcement learning (RL)-based hybrid locomotion framework to achieve hybrid walking-driving motions on the wheeled-legged robot. Without giving predefined gait patterns, our planner aims to intelligently coordinate simultaneous wheel and leg movements, thereby achieving improved terrain adaptability and improved energy efficiency. Based on RL techniques, our approach constructs a prediction policy network that could estimate external environmental states from proprioceptive sensory information, and the outputs are then fed into an actor critic network to produce optimal joint commands. The feasibility of the proposed framework is validated through both simulations and real-world experiments across diverse terrains, including flat ground, stairs, and grassy surfaces. The hybrid locomotion framework shows robust performance over various unseen terrains, highlighting its generalization capability.

Paper Structure

This paper contains 12 sections, 3 equations, 2 figures.

Table of Contents

  1. INTRODUCTION
  2. Related Work
  3. Hybrid Locomotion of Wheeled-Legged Robots
  4. Learning Approaches for Locomotion
  5. Wheeled-Legged Locomotion Policy
  6. Framework Overview
  7. Training Details
  8. Experiments
  9. Experimental setup
  10. Cross grass terrains
  11. Cross stairs
  12. CONCLUSION

Figures (2)

  • Figure 1: The overview framework of ATRos. PRPN encodes partial observations into predictive signals for a PPO-based actor–critic framework, enabling adaptive and energy-efficient locomotion with zero-shot sim-to-real transfer.
  • Figure 2: Snapshots of Go2W robot crossing different types of terrains: grass and stairs with the height of 13 cm. (a) Wheel-based traversal of grassy terrain and small slopes. (b) Wheel-based traversal of stair descent. (c) Forward stair ascent achieved through wheeled-legged locomotion. (d) Backward stair ascent through wheeled-legged locomotion.