Emergence of communication in embodied agents: co-adapting communicative and non-communicative behaviours

Abstract

In this paper I discuss in which conditions

Cooperative Robotics in Robocup Soccer is Not Just Playing a Game Andrea Bonarini

“…Evolutionary Robotics (ER) is an automatic approach for synthesising robotic systems provided with desired features and behavioural abilities [Nolfi andFloreano, 2000, Harvey et al, 2005]. ER is based on Artificial Evolution [Fogel et al, 1966, Holland, 1975, Schwefel, 1981, Goldberg, 1989.…”

“…We also present in Section 3 some case studies that demonstrate the possibility to evolve robotic systems displaying coordination, self-organisation and collective decision making. Moreover, we show how ER can be a powerful tool for the study of adaptive behaviour, communication and cooperation [Harvey et al, 2005, Nolfi, 2005. Various complexity levels can be added to the basic system described above in order to evolve cooperative, cognitive behaviours in a collective.…”

“…A possible solution to the problem of designing for self-organisation is given by automatic techniques, which can synthesise the robot controller basing on some system-level utility metric defined by the user. Among the various machine learning methodologies, Evolutionary Robotics (ER) represents a viable approach to the solution of the design problem [Nolfi andFloreano, 2000, Trianni et al, 2008]. By evaluating the robotic system as a whole (i.e., by testing the global behaviour that results from the individual rules encoded into the individual genotype), ER provides an automatic process for identifying the mechanisms that produce and support the collective behaviour, and for implementing those mechanisms into the individual controller rules that regulate the robot/environment interactions.…”

Emergence of Interaction among Adaptive Agents

“…If the environment also contains other agents then the adaptation capabilities of the agents may play a role in the emergence of communication [5].…”

Evolution of Signalling in a Group of Robots Controlled by Dynamic Neural Networks

“…ER can be employed to look at the effects that the physical interactions among embodied agents and their world have on the evolution of individual behaviour and social skills (see [6]). ER also permits the co-evolution of communicative and noncommunicative behaviour, since it lets different characteristics co-adapt, only requiring an overall evaluation of the group (see [7]). Note that one of the main features of this work is that we do not explicitly reward the group for displaying signalling behaviour.…”