Research On Some Key Issues Of Multi-robot Cooperation Based On Vision

Intelligent mobile robots can work in relatively complex environment, with the abilities ofself-planning, self-organizing and self-adaptability. Intelligent mobile robots can automaticallyimplement some routine and dangerous tasks instead of human in many cases. The multi-robotcooperation system, as one of research hot spots in recent years, has many advantages comparedwith single robot system, such as low development cost, comprehensive functions and so on. Asone of the most important supporting technologies, the vision technology research promotes theresearch and development of robot. With the improvement of visual detection, identification,tracking algorithms and other aspects algorithms, and the development of network andcommunication technologies, vision technology has been widely applied in the field ofmulti-robot control.Considering the requirement of security application, the multi-robot system platform isdesigned and fulfilled under the LAN environment. The cooperation technologies of multi-robotsystem based on vision are concerned. Some key issues and relative algorithms, including robotlocation, obstacle avoidance and formation are emphasized. Furthermore, the task of invadercapturing, one of the most typical tasks of the multi-robot system is implemented to verify theeffectiveness of the system, algorithms and software.Firstly, the multi-robot technology research status of domestic and international isintroduced, many major achievements of the domestic and foreign institutions in the field ofrobotics are summarized, and the purpose and significance of this study is described. Secondly,the overall structure of multi-robot system is correspondingly researched and designed. Theappropriate architecture of the multi-robot system is conceived, and a hardware platform of themulti-robot system is established using Pioneer robots P3-DX and Amigo. Robots cancommunicate with each other via the wireless LAN, which lays the foundation for themulti-robot system. Thirdly, vision technologies and algorithms, including SURF algorithm, thecombination model of SURF with Kalman Filter, and global visual monitoring technology basedon CAMShift etc, are studied. It is proved by the corresponding experiments in thesemi-structured environment that multi-robots system has acceptable real-time, robustness and acertain ability to adapt to the environment. These technologies and algorithms offer necessaryprerequisite of achieving more accurate robot position estimation to complete a complex task,meanwhile the accuracy and real-time of multi-robot systems to finish tasks can be improved by application of these vision algorithms. Fourthly, the corresponding vision-based algorithms ofrobot localization, obstacle avoidance, formation, capturing are analysized. Simulation andphysical experiments show the effectiveness of the related algorithms. Subsequently, the overallcommissioning test process is designed to test the properties of real-time, accuracy andadaptability of system and algorithms. It is testified that the design of multi-robot system andrelated algorithms have real time and good accuracy, but adptive ability still can be improved atfurther. Fifthly, the work achieved is summarized. The futher improvemet of this system and thefuture development of multi-robot systems are prospected.