Different Genetic Algorithms and the Evolution of Specialization: A Study with Groups of Simulated Neural Robots (original) (raw)
Related papers
Artificial Life, 2013
Organisms that live in groups, from microbial symbionts to social insects and schooling fish, exhibit a number of highly efficient cooperative behaviours, often based on role taking and specialisation. These behaviours are relevant not only for the biologist but also for the engineer interested in decentralized collective robotics. We address these phenomena by carrying out experiments with groups of two simulated robots controlled by neural networks whose connection weights are evolved by using genetic algorithms. These algorithms and controllers are well suited to autonomously find solutions to decentralized collective robotic tasks based on principles of self-organization. The paper first presents a taxonomy of role taking and specialisation mechanisms related to evolved neural-network controllers. Then it introduces two cooperation tasks which can be accomplished by either role taking or specialisation and uses these tasks to compare four different genetic algorithms to evaluate their capacity to evolve a suitable behavioural strategy which depends on the task demands. Interestingly, only one of the four algorithms, which appears to have more biological plausibility, is capable of evolving role taking or specialisation when they are needed. The results are relevant for both collective robotics and biology as they can provide useful hints on the different processes that can lead to the emergence of specialisation in robots and organisms.
Strengths and synergies of evolved and designed controllers: A study within collective robotics
Artificial Intelligence, 2009
This paper analyses the strengths and weaknesses of self-organising approaches, such as evolutionary robotics, and direct design approaches, such as behaviour-based controllers, for the production of autonomous robots' controllers, and shows how the two approaches can be usefully combined. In particular, the paper proposes a method for encoding evolved neural-network based behaviours into motor schemabased controllers and then shows how these controllers can be modified and combined to produce robots capable of solving new tasks. The method has been validated in the context of a collective robotics scenario in which a group of physically assembled simulated autonomous robots are requested to produce different forms of coordinated behaviours (e.g., coordinated motion, walled-arena exiting, and light pursuing).
Comparison of genetic algorithms used to evolve specialisation in groups of robots
This paper investigates the role of genetic algorithms in determining which kind of specialisation emerges in decentralised simulated teams of robots controlled by evolved neural networks. As shown in previous works, different tasks may be better solved by robots specialized in a particular manner. However it was not clarified how much the genetic algorithm used might drive the evolution of one kind of specialisation or another: this is the goal of this paper. The study is conducted by evolving teams of robots that have to solve two different tasks that are better accomplished by using different types of specialisation (innate versus situated). Results suggest that the type of genetic algorithm employed plays a major role in determining how robots specialize and in most of the cases the algorithms used tend to always yield the same specialization. Only one of the algorithms tested led to the emergence of the most suitable kind of specialisation for each one of the two tasks.
In recent years a number of researchers have successfully applied artificial evolution approaches to the design of controllers for autonomous robots. To date, however, Evolutionary Robotics research has focussed almost exclusively on the design of single-robot systems. We are interested in the evolution of controllers for multi-robot systems that are capable of exhibiting cooperative and coordinated behaviour. We report on recent work in which we employed artificial evolution to design neural network controllers for small, homogeneous teams of mobile autonomous robots. The robots are evolved to perform a formation movement task from random starting positions, equipped only with infrared sensors. The dual constraints of homogeneity and minimal sensors make this a non-trivial task. We describe the behaviour of a successful evolved system, in which robots work as a team, adopting and maintaining distinct but interdependent roles in order to achieve the task. We believe this to be the first successful use of evolutionary robotics methodology to develop cooperative, coordinated behaviour for a real multi-robot system.
2007
This work concerns a biologically-inspired approach to self-assembly and pattern formation in multi-robot systems. In previous work the authors have recently studied two different approaches to multi-robot control, one based upon the evolution of controllers modelled as genetic regulatory networks (GRNs), and the other based upon a model of self-organisation in aggregates of biological cells mediated by cellular adhesion molecules (CAMs). In the current work, a hybrid GRN-CAM controller is introduced, which captures the advantages, and overcomes the disadvantages, or both of the original controllers; it combines the adaptability of the evolutionary process with the robustness of an underlying self-organising dynamics. The performance of the new controller is investigated and compared with the previous ones. For example, one experiment involves the evolution of a robot cluster that can stably maintain two different spatial patterns, switching between the two upon sensing an external signal. Another experiment involves the evolution of a cluster in which individual robots develop differentiated states despite having indentical controllers (which could be used as a starting point for functional specialisation of robots within the cluster). The results show that the combined GRN-CAM controller is more flexible and robust than either the GRN controller or the CAM controller by itself, and can produce more complex spatiotemporal behaviours. The GRN-CAM controllers are also potentially portable to robotic systems other than those for which they were evolved, as long as the new system implements the underlying CAM model of self-organisation. Some technical issues regarding the implementation of the GRN and joint GRN-CAM systems are also discussed, including the use of "smart mutation" operators to improve the speed of evolution of GRNs, and evolving the rate of dynamics of the GRN controller to suit the particular task in hand.
Open-ended evolution as a means to self-organize heterogeneous multi-robot systems in real time
2010
This work deals with the application of multi-robot systems to real tasks and, in particular, their coordination through interaction based control systems. Within this field, the practical solutions that have been implemented in real robots mainly use strongly coordinated architectures and assignment strategies because of reliability and fault tolerance issues when addressing problems in reality. Emergent approaches have also been proposed with limited success, basically due to the unpredictability of the behaviors obtained. Here, an emergent approach, called r-ASiCo, is presented containing a procedure to produce predictable solutions and thus avoiding the typical problems associated with these techniques. The r-ASico algorithm is the real time version of the Asynchronous Situated Co-evolution algorithm (ASiCo), which exploits natural open-ended evolution to generate emergent complex collective behaviors and deals with systems made up of a huge number of elements and nonlinear interactions. The goal of r-ASiCo is to design the global behavior desired for the robot team as a collective entity and allow the emergence of behaviors through the interaction of the team members using social rules they learn to implement. To this end, r-ASiCo manages a series of features that are inherent to natural evolution based methods such as energy exchange and mating selection procedures, together with a technique to guide the evolution towards a design objective, the principled evaluation function selection procedure. Hence, this paper presents the components and operation of r-ASiCo and illustrates its application through a collective cleaning task example. It was implemented using 8 e-puck robots in two different real scenarios and its results complemented with those of a 30 e-puck case. The results show the capabilities of r-ASiCo to create a self-organized and adaptive multi-robot system configuration that is tolerant to environmental changes and to failures within the robot team.
Self-organized coordinated motion in groups of physically connected robots
2007
Abstract An important goal of collective robotics is the design of control systems that allow groups of robots to accomplish common tasks by coordinating without a centralized control. In this paper, we study how a group of physically assembled robots can display coherent behavior on the basis of a simple neural controller that has access only to local sensory information. This controller is synthesized through artificial evolution in a simulated environment in order to let the robots display coordinated-motion behaviors.
Evolution of signalling in a group of robots controlled by dynamic neural networks
2007
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.