A Mobile Agent-Based Framework for Automatic Coordination of Autonomous Multi-Robot Systems
By: Binsen Qian Advisor: Professor Harry Cheng
10:00 AM – 11:00 AM 2033 Bainer Hall Wednesday, March 7, 2018
Robots can protect humans from risk in many applications, such as search and rescue, outer-space exploration, and toxic cleanup. Multi-robot systems have a huge potential to benefit versatile applications through reconfiguring heterogeneous or homogeneous robots in different ways. This includes combining different capabilities for heterogeneous systems and forming different physical configurations for homogeneous systems. Additionally, multi-robot systems can reduce overall time consumption by pursuing goals by parallel execution. For example, for those applications that contain spatially separate tasks, multiple robot groups can work in different locations simultaneously and therefore improve the performance in terms of time. More importantly, due to the multiple agents and data sharing, the entire system is more fault-tolerance. While having these advantages, it also raises numerous software design challenges, such as communication, task assignment, and resource conflicts. Ideally, a cooperation framework needs to be robust, adaptable, efficient, and flexible to an extent. This research thesis presents the design and implementation of RoboCoop, a mobile agent-based framework for automatic coordination of multi-robot systems. RoboCoop consists of several modules, such as knowledge base, Input/Output, task procession/execution, and sensor reading. The agent-based framework utilizes the innate advantages of the multi-threading of each agent, such that each module can run persistently without blocking others. Modules can exchange information and data through inter-agent communication based on the standards of the Foundation for Intelligent Physical Agents (FIPA). The presented framework allows robots to coordinate, manage, and execute tasks automatically. Also, a backup mechanism is developed to ensure the robustness of robot systems. Moreover, it allows customized algorithms and strategies for task allocation and execution. Three case studies have been completed with three different robots, Linkbot, CPSBot, and Lego Mindstorms NXT/EV3, to demonstrate the potentials of the framework. In the box pushing application, there are over-sized boxes that require two robots pushing simultaneously to make it move straight, which demonstrates the ability of synchronization of RoboCoop. Then the field mapping application illustrates how robots can work in parallel to reduce completion time for missions that can be split into subsets. Finally, the robot soccer not only shows the use of sensor fusion in robot localization but also illustrates the cooperation of two teammates to complete a free-kick.
Date(s) - 03/07/2018
10:00 am - 11:00 am
Categories No Categories