Abstract. Co-evolution can give rise to the "Red Queen effect", where interacting populations alter each other's fitness landscapes. The Red Queen effect significantly complicates any measurement of co-evolutionary progress, introducing fitness ambiguities where improvements in performance of co-evolved individuals Call appear as a decline or stasis in the usual measures of evolutionary progress. Unfortunately, no appropriate measures of fitness given the Red Queen effect have been developed in artificial life, theoretical biology, population dynamics, or evolutionary genetics. We propose a set of appropriate performance measures based on both genetic and behavioral data, and illustrate their use in a simulation of co-evolution between genetically specified continuous-time noisy recurrent neural networks which generate pursuit and evasion behaviors in autonomous agents.

In this paper, we describe a novel bidding strategy that autonomous trading agents can use to participate in Continuous Double Auctions (CDAs). Our strategy is based on both short and long-term learning that allows such agents to adapt their bidding behaviour to be efficient in a wide variety of environments. For the shortterm learning, the agent updates the aggressiveness of its bidding behaviour (more aggressive means it will trade off profit to improve its chance of transacting, less aggressive that it targets more profitable transactions and is willing to trade off its chance of transacting to achieve them) based on market information observed after any bid or ask appears in the market. The long-term learning then determines how this aggressiveness factor influences an agent's choice of which bids or asks to submit in the market, and is based on market information observed after every transaction (successfully matched bid and ask). The principal motivation for the short-term learning is to enable the agent to immediately respond to market fluctuations, while for the long-term learning it is to adapt to broader trends in the way in which the market demand and supply changes over time. We benchmark our strategy against the current state of the art (ZIP and GDX) and show that it outperforms these benchmarks in both static and dynamic environments. This is true both when the population is homogeneous (where the increase in efficiency is up to 5.2%) and heterogeneous (in which case there is a 0.85 probability of our strategy being adopted in a two-population evolutionary game theoretic analysis).

We discuss the methodological foundations for our work on the development of cognitive architectures, or control systems, for situated autonomous agents. Our focus is the problems of developing sensorimotor control systems for mobile robots, but we also discuss the applicability of our approach to the study of biological systems. We argue that, for agents required to exhibit sophisticated interactions with their environments, complex sensorimotor processing is necessary, and the design, by hand, of control systems capable of such processing is likely to become prohibitively difficult as complexity increases. We propose an automatic design process involving artificial evolution, wherein the basic building blocks used for evolving cognitive architectures are noise-tolerant dynamical neural networks. These networks may be recurrent and should operate in real time. The evolution should be incremental, using an extended and modified version of a genetic algorithm.Practical constraints suggest that initial architecture evaluations should be done largely in simulation. To support our claims and proposals, we summarize results from some preliminary simulation experiments in which visually guided robots are evolved to operate in simple environments. Significantly, our results demonstrate that robust visually guided control systems evolve from evaluation functions that do not explicitly require monitoring visual input. We outline the difficulties involved in continuing with simulations and conclude by describing specialized visuorobotic equipment, designed to eliminate the need for simulated sensors and actuators.

No abstract

No abstract