Real-world transfer of evolved artificial immune system behaviours between small and large scale robotic platforms

Abstract

In mobile robotics, a solid test for adaptation is the ability of a control system to function not only in a diverse number of physical environments, but also on a number of different robotic platforms. This paper demonstrates that a set of behaviours evolved in simulation on a miniature robot (epuck) can be transferred to a much larger-scale platform (Pioneer), both in simulation and in the real world. The chosen architecture uses artificial evolution of epuck behaviours to obtain a genetic sequence, which is then employed to seed an idiotypic, artificial immune system (AIS) on the Pioneers. Despite numerous hardware and software differences between the platforms, navigation and target-finding experiments show that the evolved behaviours transfer very well to the larger robot when the idiotypic AIS technique is used. In contrast, transferability is poor when reinforcement learning alone is used, which validates the adaptability of the chosen architecture.

Artificial Immunology for Collective Adaptive Systems Design and Implementation

“…Various weaknesses in this work that required hand-coding of antibodies for instance have recently been addressed in [Whitbrook et al 2010b], who consider evolutionary methods for generating antibodies and reinforcement learning to connecting them in a network. This resulted in a system that has been ported successfully to real-robots [Whitbrook et al 2010a]. We extend this work in that we deal with swarms of robots rather than individuals, and that rather than considering individual actions, the robots must select a cooperative strategy to take part in.…”

An immune network approach for self-adaptive ensembles of autonomic components: a case study in swarm robotics

“…The resulting network of stimulatory and suppressive connections alters concentrations of antibodies; the one with the highest concentration applies its action. Various weaknesses in this work that required hand-coding of antibodies for instance have recently been addressed in Whitbrook et al (2010b), who consider evolutionary methods for generating antibodies and reinforcement learning to connecting them in a network, resulting in a system that has been ported successfully to real-robots Whitbrook et al (2010a). We extend this work in that we deal with swarms of robots rather than individuals, and that rather than considering individual actions, the robots must select a cooperative strategy to take There have also been previous attempts to apply idiotypic network ideas to swarms.…”