UROP Proceedings 2020-21

School of Engineering Department of Electronic and Computer Engineering 164 Multi-Agent System Control Supervisor: SHI Ling / ECE Student: XU Jinyun / ELEC FU Zhengyu / ISDN Course: UROP1100, Fall UROP1100, Fall A multi-agent system is a robotic system composed of multiple identical robots. It is commonly used to solve problems that are difficult or impossible for an individual agent to solve. In this research, we mainly focus on the formation control of multiple wheeled robots. We have conducted several experiments in system modeling, feedback control, and formation control of differential-drive robots. We are using the graph model to perform both distance-based formation and relative-state formation. Experiments are conducted simultaneously in the simulator and reality. Multi-Agent System Control Supervisor: SHI Ling / ECE Student: WU Bohuai / COSC Course: UROP1100, Spring Robotics is one of the most popular areas that many people would like to do research on this area. Therefore, to teach students about the essential and advanced technologies related to the robotics, there are many courses online that are relevant to robotics. However, the robotic courses online may not be useful enough for a student to know more about the robotics because using physical robot cars can give student more user experience. Thus, preparing more practical and interesting robotic courses and workshops is more important. This project sets up 7 lab materials for students to learn the useful techniques step by step. For all the 7 labs, we will use the robot car shown in Fig.1, which is a robot car with some LEDs and sensors. Multi-Agent System Control Supervisor: SHI Ling / ECE Student: ZHAO Xiaoqi / ELEC WANG Peiqi / ELEC Course: UROP1000, Summer UROP1000, Summer Multiple robot cars can perform different collective tasks, such as formation control and cargo delivery. The project aims at utilizing small two wheeled differential drive robots to accomplish tasks which would otherwise be expensive or even impossible for single machine. Robots used in the experiment run Raspberry Pi OS and communicate with the computer using WebSocket via Wi-Fi. Numerous position feedback information from the robot cars can be collected using methods like fiducial markers and motion capture systems. Various formation control algorithms have been tested such as virtual leader and potential field. A platform is also designed to automate the test procedures. Scenarios such as limited communication between each agent, limited position feedback and presence of obstacles may be simulated if such a platform could be developed.