Design Of RGB-D-based Mobile Robot Following Systems

With the rapid development of robots and AI tech,following robot has become a research hotspot in recent decades.Compared with traditional robots,following robots can conformably interact and cooperate with human beings and can perform dynamic,flexible and fine tasks in complex environments.Nowadays,following robots have been widely accepted in people's daily life.For instance,in airports,following robots can be used to help elderly people to fetch baggage;in warehouses,staff can guide following robots to take goods to designated locations;in supermarkets,people can put goods on robots for pleasant shopping.Therefore,it can be concluded that the following robots have broad application prospects and significant research importance.The shortcomings of the off-the-shelf methods for researching mobile following robots and the traditional Meanshift algorithm and PID-controlling algorithm have been deeply analyzed in this thesis,which came up with the idea to improve the traditional Meanshift algorithm and the PID-controlling algorithm,applying the algorithm to the tracking system of mobile following robots so as to improve the performance of the system.Finally,PI and fuzzy PI-controlling algorithms are validated in the tracking system.The experimental results demonstrated that the fuzzy PI-controlling algorithm has better controlling effect in the tracking system of mobile following robots.The major work and results are as follows:(1)A mobile robot tracking system is constructed through ASUS xtion Pro live camera,NVIDIA TX2 embedded development board and Open mobile robot,and an upper monitor software is designed that provides a reliable hardware and software basis for the realization of the algorithm in this thesis;(2)Aiming at the problem that the size of the core window of the traditional Meanshift algorithm isn't changeable,the depth information and Kalman filter algorithm are adopted to improve the Meanshift algorithm to make the core window adaptive,heighten the robustness of the visual tracking algorithm and provide accurate position information for the subsequent controlling algorithm;(3)The tracking error model of mobile robot is established and analyzed.When the state changes,the parameters of PI controller need to be adjusted to achieve the original control effect in that the target user's state is uncertain(namely,walking speed).A fuzzy controller is used to simulate manual adjustment of PI controller parameters so as to achieve better controlling effect in this thesis.Finally,the validity of the design scheme of the tracking system for mobile robots is verified throughout experiments;(4)Draw a conclusion for the whole page,looking ahead for the potential problems worth deeper researching.