An example of a user interface for simulation of an insect. Click on the image to see a full sized, animated version.

The simulation system we have developed is able to simulate anything from a single leg segment to an entire walking insect, including muscles, sense organs, and the nervous control system. The key to this flexibility is our unique scripting language, which is interpreted at run time. Configuration files written in this language completely specify the physical elements that are to be simulated. Hence, the user can enter the physical dimensions of a single leg of an insect, place muscles and sense organs where they are known to be located on the leg of a living insect, and study how the nervous system might use these components to generate reflex movements produced by a stimulus applied to one of the sense organs. By entering the physical dimensions of the entire body as well as all six legs, it is possible to study the way in which the nervous system might generate coordinated walking under various circumstances. The nervous system can also be emulated flexibly -- as a single "controller" or as a hierarchical chain of controllers, as desired.

An example of a user interface for simulation of a robot arm. Click on the image to see a full sized version.

A key issue in the field of robotics and artificial intelligence is how to achieve adaptive and flexible behavior in a man-made system such as a walking robot. Although a number of walking robots have been constructed, the problem of providing adaptive control for them is not yet solved. The current generation of walking robots can handle simple obstacles, but still have difficulty in moving quickly over rough terrain or righting themselves if they fall.

Because many insects exhibit adaptable and flexible locomotion, some roboticists have recently begun to build robots that have the structural features of insects. They do this because they believe that these features will confer to the robots the same flexibility of locomotor performance exhibited by the animal system after which the robot is modeled. Insects are eminently suitable models because they are physiologically relatively simple and their walking is inherently stable. Furthermore, the physical features of insect legs and joints can readily be determined, then modeled in hardware.

It may seem that the preceding two paragraphs are misplaced, that they belong on the pages describing robotics research. However, the fact of the matter is that despite the intuitive appeal of the biomimetic approach, the design of appropriate controllers for these large-scale nonlinear systems has proved difficult. Traditional control system engineering has proven inadequate because of the large numbers of sensory and motor signals and because of the complicated nonlinearities involved.

Although not without drawbacks, simulation may offer a way around these impediments. In particular, our simulation system can be used to facilitate the rapid development and testing of control algorithms for biologically inspired walking robots. Just as it is possible to specify the physical properties of an insect or parts thereof, it is just as easy to specify the physical properties of a robot, be it a single multi-articulated arm or a multi-legged walking device. Our scripting language makes it possible to connect user-written control code in any way desired to the physical structure and hence allows control code engineers to test rapidly and efficiently particular ideas about how to control even complex movements.

In principle, our simulation system may serve not only as a tool for the study of locomotion control and the control of walking robots, it may also serve as a bridge between the two. Control approaches that are based on and developed from knowledge of biological principles of locomotor control, by being implemented via our simulation system, may well be readily translated into control schemes that could be equally effective in the control of a walking robot.

Cocatre-Zilgien, J.H., F. Delcomyn, L.V. Hall and G.J. Pijanowski. 1995. An approach to validation of a leg simulation by the comparison of two dynamic models. Comput. Biol. Med. 25, 309-319.

Reichler, J.A. and F. Delcomyn. 1998. Control algorithms for biologically inspired robots: A simulation testbed. In, R. Zobel and D. Moeller, eds, Simulation--Past, Present and Future. 12th European Simulation Multiconference, pp. 437-442.

Cocatre-Zilgien, J.H. and Delcomyn, F. 1999. Modeling stress and strain in an insect leg for simulation of campaniform sensilla responses to external forces. Biol. Cybern. 81, 149-160.

Reichler, J.A. and F. Delcomyn. 2000. A dynamics simulator and controller testbed for biologically inspired walking robots. Int. J. Robotics Res. 19, 42-58.