Catálogo de publicaciones - libros

Swarm-based systems have emerged as an attractive paradigm for implementing distributed autonomous systems for various applications in commercial, military and business domains. One of the major operations in a swarm-based system is to ensure that the individual swarm units process the tasks in the environment in an efficient manner. This can be achieved using a suitable task selection mechanism that allocates the desired number of swarm units to each task while reducing inter-task latencies and communication overhead, and, ensuring adequate commitment of resources to tasks. In this paper, we describe a multi-agent based distributed task selection mechanism for swarm-based systems. We show that the distributed task selection problem is NP-complete and propose polynomial-time heuristic-based algorithms. Our simulation results show that heuristics in which each swarm unit considers both the effects of other swarm units on tasks and its own relative position to other swarm units achieve better task processing efficiency and improved distribution of swarm units over tasks.

Communication is a point of central importance in swarms of robots. This paper describes a set of simulations in which artificial evolution is used as a means to engineer robot neuro-controllers capable of guiding groups of robots in a categorisation task by producing appropriate actions. Communicative behaviour emerges, notwithstanding the absence of explicit selective pressure (coded into the fitness function) to favour signalling over non-signalling groups. Post-evaluation analyses illustrate the adaptive function of the evolved signals and show that they are tightly linked to the behavioural repertoire of the agents. Finally, our approach for developing controllers is validated by successfully porting one evolved controller on real robots.

This research is positioned in the context of controller design for (simulated) multi-robot applications. Inspired by research in survey and exploration of unknown environments where a multi-robot system is to discover features of interest given strict time and energy constraints, we defined an abstract task domain with adaptable features of interest. Additionally, we parameterized the behavioral features of the robots, so that we could classify behavioral specialization in the space of these parameters. This allowed systematic experimentation over a range of task instances and types of specialization in order to investigate the advantage of specialization. These experiments also delivered a novel neuro-evolution approach to controller design, called the collective specialization method. Results elucidated that this method derived multi-robot system controllers that outperformed a high performance heuristic and conventional neuro-evolution method.

Generally speaking, the behavioural strategies of a multi-robot system can be defined as scalable if the performance of the system does not drop by increasing the cardinality of the group. The research work presented in this paper studies the issue of scalability in artificial neural network controllers designed by evolutionary algorithms. The networks are evolved to control homogeneous group of autonomous robots required to solve a navigation task in an open arena. This work shows that, the controllers designed to solve the task, generate navigation strategies which are potentially scalable. However, through an analysis of the dynamics of the single robot controller we identify elements that significantly hinder the scalability of the system. The analysis we present in this paper helps to understand the principles underlying the concepts of scalability in this kind of multi-robot systems and to design more scalable solutions.