RcLab 대학원생 및 학부연구생을 모집합니다!
RUBI is a waterproof bipedal robot developed to extend its operating range into water-exposed environments. It achieves IP67-level waterproofing by sealing its actuators and electronic modules. It is controlled by a reinforcement-learning-based locomotion policy to operate robustly across changes in terrain and environmental conditions.
RUBI is controlled using reinforcement learning and is capable of stable walking on a variety of ground environments, including flat ground, slopes, gravel, and slippery snow.
Development and Reinforcement Learning based Control of RUBI
RUBI at the Humanoid 2025
To expand the robot's operation to water-exposed environments, experiments were conducted under both spray and submersion conditions. These tests verified the robot's waterproof design performance and the stability and robustness of the control algorithm.
Walking of RUBI in Spray and Underwater Environments
RUBI's point-foot structure makes it structurally disadvantageous for stable self-sufficiency while stationary. Nevertheless, we implemented an algorithm that learns stand-up motion through reinforcement learning and enables the transition to walking after independence.
RUBI, Stand Up!