Recent advances in robotics have started making it feasible to deploy large numbers of inexpensive robots for tasks such as surveillance and search. However, coordination of multiple robots to accomplish such tasks remains a challenging problem. This report reviews some of the recent literature in multi-robot systems. It consists of two parts. In the first part, we reviewed the studies on the pattern formation problem, that is how can a group of robots be controlled to get into and maintain a formation. The second part reviews the studies that used adaptation strategies in controlling multirobot systems. Specifically we have investigated (1) how learning (life-long adaptation) is used to make multi-robot systems respond to changes in the environment as well in the capabilities of individual robots, and (2) how evolution is used to generate group behaviors.

The distributed control of self-assembly processes requires local behaviors that will cause initially unorganized components to form a desired goal structure. While important strides have been made in designing methods for self-assembling various geometric structures under idealized simulated conditions, many unaddressed issues remain in extending these methods to more complex environments. In this work, we discuss the self-assembly of prespecified 3D structures from blocks of different sizes. Block movements through a continuous environment are constrained by each other and simulated gravity, adding to the problem’s difficulty. We present a solution that integrates three distinct techniques from the field of swarm intelligence: stigmergic pattern recognition, force-based movement control, and coordination via local message passing and state changes. Further, we empirically demonstrate that a stochastic component in the blocks ’ acceleration can aid in preventing persistent interference, and that the use of collective, flock-like movements can be beneficial in situations of low block availability. This work provides insight into the dynamics of continuous-space self-assembly, and is a step towards the design of methods for the automated “growth ” of useful structures in real-world environments.

Abstract. This paper aims at providing a rigorous definition of selforganization, one of the most desired properties for dynamic systems (e.g., peer-to-peer systems, sensor networks, cooperative robotics, or ad-hoc networks). We characterize different classes of self-organization through liveness and safety properties that both capture information regarding the system entropy. We illustrate these classes through study cases. The first ones are two representative P2P overlays (CAN and Pastry) and the others are specific implementations of Ω (the leader oracle) and one-shot query abstractions for dynamic settings. Our study aims at understanding the limits and respective power of existing self-organized protocols and lays the basis of designing robust algorithm for dynamic systems. 1

with no coordinate agreement

Abstract A new algorithm for the control of formations of mobile robots is presented. Formations with a triangular lattice structure are created using distributed control rules, using only local information on each robot. The overall direction of movement of the formation is not pre-established but rather results from local interactions, giving all the robots a common, self-organized heading. Experiments were done to test the algorithm, yielding results in which robots behaved as expected, moving at a reasonable speed and maintaining the desired distances among themselves. Up to seven robots were used in real experiments and up to forty in simulation. 1