Adaptive multi-mode locomotion for bipedal wheel-legged robots via sparse mixture-of-experts deep reinforcement learning

TY - JOUR

T1 - Adaptive multi-mode locomotion for bipedal wheel-legged robots via sparse mixture-of-experts deep reinforcement learning

AU - He, Pan

AU - Zhao, Zeang

AU - Duan, Shengyu

AU - Wang, Panding

AU - Lei, Hongshuai

N1 - Publisher Copyright: Copyright © 2026 He, Zhao, Duan, Wang and Lei.

PY - 2026

Y1 - 2026

N2 - The bipedal wheel-legged robot combines the high energy efficiency of wheeled movement with the terrain adaptability of legged locomotion. However, achieving a smooth transition between these two heterogeneous motion modes within a unified control framework remains challenging. This study proposes a reinforcement learning control framework that integrates the Mixture of Experts (MoE) architecture. This approach employs a “divide and conquer” strategy by introducing a dynamic gating network and a Top-K sparse activation mechanism, which automatically allocates different motion modes to specific expert subnetworks, effectively decoupling conflicting gradients. Simulation results demonstrate that, compared to the single-network PPO method, the MoE-enhanced algorithm exhibits significant improvements in training stability and rewards. The learned policy successfully achieved smooth rolling on flat surfaces and transitioned to dynamic leg-lifting gaits when confronted with obstacles. In various test terrains, it showed a markedly higher success rate compared to the single-network PPO method.

AB - The bipedal wheel-legged robot combines the high energy efficiency of wheeled movement with the terrain adaptability of legged locomotion. However, achieving a smooth transition between these two heterogeneous motion modes within a unified control framework remains challenging. This study proposes a reinforcement learning control framework that integrates the Mixture of Experts (MoE) architecture. This approach employs a “divide and conquer” strategy by introducing a dynamic gating network and a Top-K sparse activation mechanism, which automatically allocates different motion modes to specific expert subnetworks, effectively decoupling conflicting gradients. Simulation results demonstrate that, compared to the single-network PPO method, the MoE-enhanced algorithm exhibits significant improvements in training stability and rewards. The learned policy successfully achieved smooth rolling on flat surfaces and transitioned to dynamic leg-lifting gaits when confronted with obstacles. In various test terrains, it showed a markedly higher success rate compared to the single-network PPO method.

KW - bipedal wheel-legged robot

KW - curriculum learning

KW - gradient conflict

KW - mixture of experts

KW - reinforcement learning

UR - https://www.scopus.com/pages/publications/105032242687

U2 - 10.3389/frobt.2026.1788395

DO - 10.3389/frobt.2026.1788395

M3 - Article

AN - SCOPUS:105032242687

SN - 2296-9144

VL - 13

JO - Frontiers in Robotics and AI

JF - Frontiers in Robotics and AI

M1 - 1788395

ER -