RGB-D Sensor Based Human Comfortable Following Behavior for Service Robots in Indoor Environments

Combining the scientific research related to human behavior and social interaction, a human following method is proposed in consideration of human comfort feelings for service robots to achieve a more friendly and effective following behavior with a high degree of acceptance. With high-reliability and low-cost RGB-D (RGB-depth) sensor, a people detection algorithm based on HOG (histogram of oriented gradient) and a people tracking algorithm based on UKF (unscented Kalman filter) are studied, and real-time and accurate detection and tracking of human are realized in dynamic unstructured environment. On the basis of discussing the factors that affect the following comfort, a comfortable following model and a behavior utility function are established, and the motion control and planning based on human comfort are realized. Systematic analysis and evaluation on the proposed human comfort following method are carried out, which verifies the validity and usability of the following behavior, and further provides the possibility of friendly interaction between the human and the robot.