Adaptive Bipedal Robot Walking on Industrial Pipes Under Neural Multimodal Locomotion Control: Toward Robotic Out-Pipe Inspection

Abstract

Out-pipe inspection robots encounter challenges in balancing on curved surfaces while moving fast, climbing horizontal and vertical pipes with limited energy, overcoming obstacles, like flanges, and performing stable transition between pipe segments. However, existing robots have yet fully achieved such multimodal locomotion effectively. Therefore, this work presents a novel adaptive bipedal robot and neural multimodal locomotion control toward semiautonomous robotic out-pipe inspection. The locomotion control utilizes eight analyzable neural modules forming a modular structure. Their integration can generate multimodal locomotion with online adaptation of the robot. Taking high-level commands from an operator, such as direction and behavior selection (crawling, obstacle crossing, and transition), the robot autonomously adapts its body posture to balance on pipes, resulting in 100% successful locomotion on horizontal and vertical smooth pipes conducted in our studies. These behaviors are achieved with a speed of 10.24 cm/s, and a cost of transport of 26.3 J/kgm, showing over 200% improvement in terms of speed and energy efficiency compared to existing state-of-the-art out-pipe legged inspection robots. Besides, the robot can overcome obstacles up to 14 cm in height (48% of its height) and perform stable transitions between horizontal and vertical segments. With this level of robustness and efficiency, it has the potential for out-pipe inspection in the oil and gas industry.