MRG icon
Chris Adams



biological models


biological models

social robots

learning by imitation

artificial life


Contact: Bridget Hallam
Last updated: Fri Oct 12 09:00:55 2001

The RoBat project  

The RoBat project has a dual purpose: 1. to provide a platform from which to investigate the real-time information processing tasks carried out by biological systems -- i.e., echolocating bats; 2. to engender artificial navigation systems with some of the sophisticated performance and robustness of those biological systems.

RoBat consists of three main components: a biomimetic sonarhead, a 3 DOF mobile platform and a signal processing package whose operations, performed upon the received echoes, are based upon a filterbank model of the processing performed by the mammalian cochlea. These three components are all controlled and integrated into a single system by software running on a Pentium III PC (Linux).

The 6 DOF sonarhead allows panning and tilting of the neck, and independent panning and tilting of each of the two ears (receivers). The ultrasonic transducers are Polaroid electrostatic transducers. The motors driving the different axes are standard radio-control servomotors.

More information about previous work done with the bionic sonarhead can be found here.

Topics of interest:

  • Doppler-shift based navigation
  • Artificial pinnae using genetic algorithms

Current researchers:

Former researchers:

Recent MSc Projects

  • Cook, David L., 2001. A Parallel Object Model for Evolution of Echolocating-Enhancing Acoustic Structure in a Robotic Bat.

Cricket phonotaxis   The aim of this work is to use robots to model specific biological sensorimotor control systems. By taking a robotic approach (how can I get a machine to behave this way?) to a well-explored biological problem (what are the known characteristics and underlying systems for this behaviour?) we can draw on the strengths of both fields in attempting to understand how perceptual mechanisms work.

The specific system that has been examined is the phonotaxis behaviour of the cricket. Female crickets find male crickets by walking (or flying) towards a species specific song that the males produce. The robot model implements a mechanism for this task that is consistent with neuroethological evidence but substantially simpler than the mechanisms hypothesized by neuroethologists. In particular, "recognition" of the species song occurs as a side-effect of the "location" device, rather than requiring additional processing. The robot has been tested in a variety of experiments based on cricket research and has been shown to reproduce a range of behaviours, including signal preference and choice, that are usually taken as evidence for more complex mechanisms.

The robot must physically interact with a real sound field, so the problems it faces realistically represent those facing the cricket. This is an important advantage over simulation modelling, both for generating and for testing hypotheses about sensorimotor mechanisms. One line of current research is to improve the realism by implementing the sound-seeking mechanism on a much smaller (Khepera) robot. This will also permit better control of experimental parameters. We are also looking at extending the methodology to examine some of the other orientation systems in the cricket, such as negative phototaxis and escape responses.

Former researchers:

Control of locomotion  
Walking: This work is looking at the possibility of controlling a legged robot with a structure based on that used in vertebrates, i.e. Central Pattern Generators - these are neural networks which generate the basic rhythmical motor pattern for walking adapted to the simple high level control signals they receive from the brain and the sensory feedback they receive from the body. No specific animal or CPG is being directly modelled, the aim is rather to try to apply ideas gleaned from neurophysiological research into these structures to the problem of controlling legged robots.

A simple planar biped controlled by neural oscillators was originally simulated, repeating the work of Taga in 1990, and this has then been tested and extended to make it more realistic. The results of this were very promising with the robot being able to walk on a wide variety of surfaces, and further work is now being carried out to look at ways of automatically generating the control structures, rather than handcrafting them for individual robots.

To this end a general purpose mechanical simulator is being connected to an Genetic Algorithm package which will then evolve Neural Networks to control a variety of walking robots.

Swimming: This research compares naturally and artificially evolved neural networks controlling the anguilliform swimming of a lamprey. It is inspired from biological models developed by Grillner and Ekeberg from SANS . A real number Genetic Algorithm is used to evolve alternative artificial controllers composed of neurons similar to those of the biological model. Results show that artificial controllers can be obtained with other architectures than the biological one, and which, in some way, are more efficient. See Auke Ijspeert's research page for more information.

New projects:

  • Gait selection in hexapods - looking at mathematical models of how gaits change as speed increases.

Current researcher:

Former researchers:

The Halperin net   There has been a stimulating dialogue between automous robot researchers and ethologists (biologists studying animal behaviour). Our group has taken a particular interest in a model called the Halperin net.

Anton Dil worked on using the net to control a robot organising chairs.

Bridget Hallam worked on simulating animal conditioning experiments (on computer, not robot). Her results showed that the model was surprisingly good considering that it was not devised as a model of conditioning, but found several places where improvements to the model could be made.

Former researchers: