Vision-based walking control of bipedal and quadrupedal transformable robot for nuclear power plant working

Abstract

The purpose of this study is to develop all the technologies necessary to develop and operate a remote control robot for emergency measures and normal monitoring in case of accidents in nuclear power plants. In a high-risk environment or a disaster environment, stable mobility without overturning the robot should be a top priority. Track and wheel type robots have limitations in their ability to overcome rough terrain such as stairs, so they are suitable for overcoming complex environments, but bipedal walking robots are difficult to control, and stability is poor due to high center of gravity and narrow stable area. . In the case of four-legged walking, stable walking is possible due to a wide stable area, but there is a limit to workability. Therefore, by utilizing the advantages of bipedal walking and quadruped walking, we developed a robot that can perform a task stably, especially by imitating a chimpanzee that is capable of agile movement and does not fall over. In order to solve the problems of the existing robot hardware, the weight and the moment of inertia are reduced, and light and fast movement is possible through optimal design. This was verified through simulation and analyzed by comparing it with the existing robot performance. In addition, the robot platform was used to perform tasks such as walking, overcoming stairs, valve work, and environment and location recognition in the nuclear power plant. In particular, an algorithm for stably walking against external shocks was developed, and an autonomous walking method that minimizes human intervention was proposed by recognizing the surrounding environment through depth information and object recognition and generating a walking path and footstep. To verify this, we conducted simulations and experiments and confirmed the feasibility of robots for work in nuclear power plants.