A multi-agent system refers to a network of interacting, mobile, physical entities that collectively perform a complex task beyond their individual capabilities. The autonomy and communication aspects are crucial here, since every agent has information about their immediate (local) environment and may need to achieve their own (local) goals and tasks. If a global task needs to be achieved, these have to be communicated with other agents. General examples of multi-agent systems are software, machines, and even humans.
Recent advances in sensor technology, embedded systems, communication systems, and power storage now make it feasible to deploy such system of cooperating agents for various civilian and military applications. For instance, a group of autonomous (ground, underwater, water surface, or air) vehicles could be deployed in large disaster areas to perform search, mapping, surveillance, or environmental monitoring and clean up without putting first responders in harm’s way.
Often in formation control the main concern is convergence to the desired spatial configuration irrespective of its exact global position in space. That is, the formation is to be acquired up to rotation and translation of the whole set of agent positions. This means that only the relative positions of agents need to be known by the control algorithm. Formation control problems are relatively straightforward to solve when the agents’ absolute coordinates (i.e., with respect to an Earth-fixed coordinate frame) are available via a central planner. From this information, the relative positions can be readily calculated.
Sometimes multi-agent or multi-robot systems are difficult to separate conceptually from swarm robotics systems. Multi-agent systems, however, often rely on the distribution of or access to global information, broadcasts, sophisticated communication protocols that require reliable robot–robot communication, explicit assignments of roles that require robots to identify individual robots or to know the total swarm size.
Though, swarm robots are multi-robot or multi-agent systems. They typically consist of many simple robots which coordinate. Through this coordination the swarm exhibits a collective behavior (global task), e.g. flocking. It’s inspired by the study of swarm behaviors exhibited by birds, ants etc..