A MODULAR PLATFORM FOR MULTI-AGENT ROBOTIC RESEARCH

This paper introduces a modular system to be used in single or multi-agent scenarios. Multi-agent robotic systems are studied extensively by researchers in several fields all around the world. This interest is inspired by animals or insect communities in nature. It is supported by the increasing necessity from the industry to develop robotic systems that are capable of performing various tasks as a unit or cooperate with other units to perform harder tasks. Research done in this area is not only useful for robotics application but provides biologists with a platform to study simple, yet impressive behaviours in nature. This project aims at presenting a low-cost modular platform for research and educational purposes to further investigate and analyse those behaviours. Load transportation using multiple agents is considered as an example of such behaviours. Two different methods of transportation are studied adopting distant sets of load interface modules and sensors. The first case uses a master-slave architecture, where the commands are shared among all carriers and the load is previously put onto the robots. Experiments are performed to observe the induced agents-load interaction forces. The second case introduces a more complex solution which requires decentralized control and synchronization methods to perform a successful lift and transport using limited communication resources. A model of the robot is implemented in WEBOTS simulator and target objects with various weights are studied