Design, modeling, control, and evaluation of a wearable Centaur robot for load-carriage walking assistance

Abstract

While both autonomous and wearable robots can assist humans in load-carriage tasks, existing autonomous systems face challenges in real-world autonomy and endurance, whereas current wearable systems have shown limited effectiveness in improving human walking efficiency. This paper introduces the Centaur robot, an innovative wearable human augmentation robot that integrates human intelligence with robotic strength for collaborative load-carriage walking. The Centaur robot comprises two independent three-DoF robotic legs and a robotic torso, coupled with the human via a passive softening elastic mechanism, forming a human-Centaur quadruped system. This configuration optimizes vertical load distribution and provides horizontal forward force acting through the center of mass of the human during walking. The compliance-based interaction model established through the elastic mechanism enables dynamic decoupling of the human-Centaur system, allowing the Centaur to be modeled independently. To achieve coordinated locomotion and interaction force control, a novel loco-interaction control strategy is proposed. To further enhance the traversability to varying terrains, a terrain-adaptive swing leg controller is developed to generate a terrain-specific swing trajectory. Experimental evaluation results demonstrate that the Centaur robot effectively adapts to varying human walking directions and speeds while seamlessly collaborating with the human to traverse diverse terrains. In the load-carriage experiment (n = 5), the Centaur robot achieved a load-sharing ratio of 52.22% ± 15.52%, reduced the metabolic cost by 35.16% ± 4.95%, and improved lateral gait stability compared to a regular backpack when carrying a 20 kg load, equivalent to 28.8% ± 4.03% of the participants’ body weight.