Lunedý 31-5 14.30

Alessio Turetta Graal-Lab DIST, UniversitÓ di Genova.

Self-Organizing Distributed Control Algorithm for Modular Robotic Systems

The main objective of the present research area is the design of a modular, functional, and algorithmic architecture to face with the problem of controlling modular, multi-robot systems. More specifically, the research deals with multi-robot structures composed via electro-mechanical connection of different, separately controlled, sub-systems, and aims to preserve the existing controllers by establishing a self-coordinating behaviour among every controlled atomic units. No knowledge about kinematics and dynamics of the overall structure (upon electro-mechanical connection) is required. And no explicit differentiated roles assignment is necessary. All the units are regarded in the same manner: each module just knows itself and does its best to perform the assigned common task, asking for help to the other ones just if necessary. In order to achieve such noteworthy advantages, a price must however be paid in terms of required data exchange: a number of pipe-lined iterations along the chains have to be performed (within a sampling interval of the overall control scheme) before each unit is allowed to update and apply its locally computed control action. However, simulative results show that valuable performances can be achieved even with a restricted number of iterations; thus leading to a computational burden (including both calculation and communication time slots) of the same order of magnitude of that required for standard, centralized, closed loop kinemati inversions.

Meawhile, since the distributed nature of the so devised computational structure also admits the possibility for its organization within an hierarchy of nested clusters of distributed computational units, as well as memory units, this seems to open a future perspective for robotics applications, whenever seen in the framework of some recent trends, such as brain and neuro-robotics sciences.