SKATER: Synthesized Kinematics for Advanced Traversing Efficiency on a Humanoid Robot via Roller Skate Swizzles

TL;DR

This work presents SKATER, a 25-DoF humanoid endowed with four passive inline wheels per foot to enable roller-skate swizzle locomotion, combined with a PPO-based deep reinforcement learning controller trained in simulation and transferred to the real robot via domain randomization. The approach yields a continuous, low-impact rolling gait that emerges from an implicit gait reward rather than explicit timing, achieving substantial reductions in joint loads and energy use compared with walking. Across simulation and real-world tests, the swizzle gait achieves a $I$-type impact reduction of $75.86\%$ and a $CoT$ reduction of $63.34\%$, while maintaining robust performance on multiple friction surfaces and showing pronounced joint-torque savings. The results support roller skating as a viable, energy-efficient locomotion modality for humanoids, with opportunities to extend to higher-speed gaits and perception-enabled autonomy.

Abstract

Although recent years have seen significant progress of humanoid robots in walking and running, the frequent foot strikes with ground during these locomotion gaits inevitably generate high instantaneous impact forces, which leads to exacerbated joint wear and poor energy utilization. Roller skating, as a sport with substantial biomechanical value, can achieve fast and continuous sliding through rational utilization of body inertia, featuring minimal kinetic energy loss. Therefore, this study proposes a novel humanoid robot with each foot equipped with a row of four passive wheels for roller skating. A deep reinforcement learning control framework is also developed for the swizzle gait with the reward function design based on the intrinsic characteristics of roller skating. The learned policy is first analyzed in simulation and then deployed on the physical robot to demonstrate the smoothness and efficiency of the swizzle gait over traditional bipedal walking gait in terms of Impact Intensity and Cost of Transport during locomotion. A reduction of $75.86\%$ and $63.34\%$ of these two metrics indicate roller skating as a superior locomotion mode for enhanced energy efficiency and joint longevity.

Figures (9)

• Figure 1: The SKATER system: a humanoid robot equipped with roller skates for learning swizzle locomotion through deep reinforcement learning.
• Figure 2: Deep reinforcement learning control framework for SKATER. The policy network processes proprioceptive and exteroceptive sensor data to generate joint-level commands, enabling adaptive roller skate swizzle locomotion.
• Figure 3: Hardware specifications of the SKATER humanoid robot.
• Figure 4: Comparison of foot contact force profiles: (a) roller skating locomotion with continuous ground contact and stable force distribution, and (b) conventional bipedal walking with periodic impact peaks and zero-force swing phases.
• Figure 5: Velocity tracking performance of SKATER following step reference commands.