Distributed Control Architectures for Modular and Networked Robotic Systems

In recent years application fields of robotic systems are substantially widened. Other than in manufacturing industry, a wide spectrum of different well-promising application domains (within what is conventionally called “service robotics” domain) are emerging outside the plants. On the one hand, robots are progressively entering in our homes, providing safety, fun, assistance, education, and care to our daily life. On the other hand, the engagement of robotic systems for performing professional activities is extending to an increasing number of applications, such as medical, defence, security, underwater, inspection, rescue.

In parallel with the enlargement of the application fields, the level of complexity of the required operations is continuously increasing and frequently imposes the simultaneous engagement of more than a single robotic structure to execute “difficult” tasks. As representative examples, consider the number of different scenarios in which teams of robots could be employed for performing coordinated transport and manipulation operations in not-structured hazardous environments (such as nuclear plants to be dismantled), delicate and crucial rescue activities (in the scenarios immediately following an earthquake or other calamitous events), or cooperative patrolling missions (in restricted access areas for detecting the presence of intruders or injuring agents).

When considering a generic multi robot system, one among the fundamental factors for success is certainly represented by the coordination capability. In order to improve the efficiency of the overall system (making it sufficiently flexible to be deployable in very different scenarios), it is important to adopt control solutions characterized by high levels of flexibility, scalability and modularity. A consideration this last which seems establishing a clear bridge between a generic Networked Robotic System (NRS) and a modular robot obtained by plugging together a set of basic robotic units which are asked to cooperate in executing a common task.

In this light, the talk will approach the problem of control and coordination of complex NRSs by first presenting the results obtained during years of activities in the field of modular robotics. In particular it will be emphasized how the coordination functionality, rather than being provided by external centralized processing units, can be more conveniently obtained in a decentralized fashion, by just exploiting the computation and communication capabilities available on board of every single robot controller.
Then it will be shown how the same control techniques developed for aggregation of “micro-structures” (the robotic modules) can be adopted also to deal with “macro-structures” (like a multi-arm mobile manipulator), obtained via composition of macro-robotic units (such as a vehicle or a manipulator) separately controlled and “occasionally” put together for accomplishing tasks more complex than those executable by each stand-alone separate unit.
Finally some on-going research activities aiming to adopt the same distributed paradigm also for dealing with teams of cooperative autonomous underwater vehicles will be presented.