Multi-robot exploration can be categorized mainly into two areas; centralized and decentralized. Lots of researches have been carried out with regarding centralized multi-robot systems. It can be emphasized that centralized systems achieve the optimal solution.Unfortunately, Single point of failure is one of the major disadvantages of centralized systems. Luckily, distributed systems can address this point very effectively.
Advantages of multi-robot decentralized systems
- Robustness: There can be dangerous situations such as natural disasters, explosions and etc.There for there is a high probability of robot encountering a failure.But this can be addressed very effectively in decentralized systems.Because remaining agents can adapt to the situation and distribute task according to the current situation
- Scalability: The Same task can be divided among the available agents (There can be limit exist for this. Most probably it will not be a multi-robot system if it contains hundred or two hundred agents. Swarm Robotics discuss systems with hundred more agents.)
- Reliability: Reliability of the final results can be guaranteed to be high since the malfunctioning of a single robot may not affect the final results thoroughly.
- Efficiency: It is obvious that the same task can be done using few robots after spending lesser time than a single robot.But there is an overhead.Because robots always have to take a decision after communicating with other agents.Therefore proper coordination is essential.There can be redundancies exist due to not having effective coordination.
- Flexibility: Since there is no central agent flexibility is also high.It is not necessary to build more powerful single agent instead simple agents can be occupied.Based on the implementation there can exist heterogeneous or homogeneous systems.Therefore, task can be allocated to each agent based on their capabilities
In the decentralized architecture, there are two types called distributed architecture and hierarchical architecture. In distributed architecture, each robot is treated as equal manner. In hierarchical approach, there are few local control robots. Both centralized and distributed qualities are occupied in here. In the distributed architecture each robot is making the decision based on the local information they have. The significant point is that each robot has to pick the place to discover after negotiation with the other robots in its communication range.