Efficient Quadruped Robot Control Through Human Pose Recognition Using Combined Vision Models

Quadrupeded robot control is in an indoor or outdoor environment is an important research area for human-robot collaboration (HRC). This paper introduces a novel method for real-time human pose recognition tailored for controlling quadruped mobile robots in indoor environments (). We propose an integrated system based on Yolov5, OpenPose, and DeepSORT tracking algorithm. This system is able to accrurately recognize human postures, which is crucial for human pose recognition in. By leveraging vision-based posture detection, our system establishes a robust, non-verbal communication channel that enhances the collaboration between humans and robots. This innovative method highlights the potential of posture-based controls in enhancing the adaptability and intuitiveness of human-robot interactions.