Inman Harvey

Explorations in evolutionary robotics

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.

Noise and the Reality Gap: The Use of Simulation in Evolutionary Robotics

The pitfalls of naive robot simulations have been recognised for areas such as evolutionary robotics. It has been suggested that carefully validated simulations with a proper treatment of noise may overcome these problems. This paper reports the results of experiments intended to test some of these claims. A simulation was constructed of a two-wheeled Khepera robot with IR and ambient light sensors. This included detailed mathematical models of the robot-environment interaction dynamics with empirically determined parameters. Artificial evolution was used to develop recurrent dynamical network controllers for the simulated robot, for obstacle-avoidance and light-seeking tasks, using different levels of noise in the simulation. The evolved controllers were down-loaded onto the real robot and the correspondence between behaviour in simulation and in reality was tested. The level of correspondence varied according to how much noise was used in the simulation, with very good results achieved when realistic quantities were applied. It has been demonstrated that it is possible to develop successful robot controllers in simulation that generate almost identical behaviours in reality, at least for a particular class of robot-environment interaction dynamics.

Evolutionary Robotics: the Sussex Approach

We give an overview of evolutionary robotics research at Sussex over the last five years. We explain and justify our distinctive approaches to (artificial) evolution, and to the nature of robot control systems that are evolved. Results are presented from research with evolved controllers for autonomous mobile robots, simulated robots, co-evolved animats, real robots with software controllers, and a real robot with a controller directly evolved in hardware.

Evolutionary Robotics: A New Scientific Tool for Studying Cognition

We survey developments in artificial neural networks, in behavior-based robotics, and in evolutionary algorithms that set the stage for evolutionary robotics (ER) in the 1990s. We examine the motivations for using ER as a scientific tool for studying minimal models of cognition, with the advantage of being capable of generating integrated sensorimotor systems with minimal (or controllable) prejudices. These systems must act as a whole in close coupling with their environments, which is an essential aspect of real cognition that is often either bypassed or modeled poorly in other disciplines. We demonstrate with three example studies: homeostasis under visual inversion, the origins of learning, and the ontogenetic acquisition of entrainment.

Through the Labyrinth Evolution Finds a Way: A Silicon Ridge

Artificial evolution is discussed in the context of a successful experiment evolving a hardware configuration for a silicon chip (a Field Programmable Gate Array); the real chip was used to evaluate individual configurations on a tone-recognition task. The evolutionary pathway is analysed; it is shown that the population is genetically highly converged and travels far through genotype space. Species Adaptation Genetic Algorithms (SAGA) are appropriate for this type of evolution, and it is shown how an appropriate mutation rate was chosen. The role of junk on the genotype is discussed, and it is suggested that neutral networks (paths through genotype space via mutations which leave fitness unchanged) may be crucial to the effectiveness of evolution.

Unconstrained Evolution and Hard Consequences

Artificial evolution as a design methodology for hardware frees many of the simplifying constraints normally imposed to make design by humans tractable. However, this freedom comes at some cost, and a whole fresh set of issues must be considered. Standard genetic algorithms are not generally appropriate for hardware evolution when the number of components need not be predetermined. The use of simulations is problematic, and robustness in the presence of noise or hardware faults is important. We present theoretical arguments, and illustrate with a physical piece of hardware evolved in the real-world (‘intrinsically evolved’ hardware). A simple asynchronous digital circuit controls a real robot, using a minimal sensorimotor control system of 32 bits of RAM and a few flip-flops to co-ordinate sonar pulses and motor pulses with no further processing. This circuit is tolerant to single-stuck-at faults in the RAM. The methodology is applicable to many types of hardware, including Field-Programmable Gate Arrays (FPGAs).

Evolutionary Robotics: A Survey of Applications and Problems

This paper reviews evolutionary approaches to the automatic design of real robots exhibiting a given behavior in a given environment. Such a methodology has been successfully applied to various wheeled and legged robots, and to numerous behaviors including wall-following, obstacle-avoidance, light-seeking, arena cleaning and target seeking. Its potentialities and limitations are discussed in the text and directions for future work are outlined.

Artificial Evolution: A Continuing SAGA

I start with a basic tutorial on Artificial Evolution, and then show the simplest possible way of implementing this with the Microbial Genetic Algorithm. I then discuss some shortcomings in many of the basic assumptions of the orthodox Genetic Algorithm (GA) community, and give a rather different perspective. The basic principles of SAGA (Species Adaptation GAs) will be outlined, and the concept of Neutral Networks, pathways of level fitness through a fitness landscape will be introduced. A practical example will demonstrate the relevance of this.

Artificial evolution: a new path for Artificial Intelligence?

Recently there have been a number of proposals for the use of artificial evolution as a radically new approach to the development of control systems for autonomous robots. This paper explains the artificial evolution approach, using work at Sussex to illustrate it. The paper revolves around a case study on the concurrent evolution of control networks and visual sensor morphologies for a mobile robot. Wider intellectual issues surrounding the work are discussed, as is the use of more abstract evolutionary simulations as a new potentially useful tool in theoretical biology.

The use of genetic algorithms for the development of sensorimotor control systems

Provides a high-level review of current and recent work in the use of genetic algorithm-based techniques to develop sensorimotor control systems for autonomous agents. It focuses on network-based controllers and genetic encoding issues associated with them. The paper closes with a discussion of the possibility of using artificial evolutionary techniques to help tackle more specifically scientific questions about natural sensorimotor systems.