3.7 Reaction-Diffusion Robots


3.7 Reaction-Diffusion Robots

Unconventional robotics closely follows unconventional computing. Evolutionary computing excited a wave of research in evolving novel robotic architectures, while smart materials opened previously unexplored routes in the design of artificial muscles and polymer-based actuators. Studies in collective intelligence pushed forward research in emergent phenomena in collectives of minimalist robots. Reactiondiffusion robots may therefore appear to naturally follow from reaction-diffusion computing.

If one were to apply theoretical advances or experimental achievements of active nonlinear medium computing to robotics, one would probably start with designing unconventional controllers for mobile robots. An unconventional controller could employ some implicit forms of computation (e.g., wave computation, in which local microprocesses of wave generation and spreading may result in certain macroscopic phenomena), which can be used as a guidance of the robot motion.

The first-ever approach to control a mobile robot by a chemical medium was presented by Ziegler, Dittrich, and Banzahf (1997), where, in this design, chemical reaction chains were employed. A light-seeking robot is under consideration (Ziegler, Dittrich, and Banzahf 1997): In this case, a controller is a chemical reactor with several reagents. There are some connections between the light sensors and the reactor, and between the reactor and the motors. Changes in the values of light sensors lead to changes in the concentration of reagents in the reactor. The chemical reactions between the metabolites result in the corresponding changes in robot behavior (some results of the simulation experiments, plus graphs of the appropriate metabolic networks can be found in Ziegler, Dittrich, and Banzahf 1997).

Another approach, developed by Adamatzky and Colleagues (Adamatzky et al. 1999; Adamatzky and Melhuish 2000, 2002; Adamatzky 2001), is to exploit spacetime excitation dynamics in nonlinear media. Obviously, if we talk about a lightseeking robot, we keep the light-sensitive Belousov-Zhabotinsky reaction in mind. Assume that the chemical medium, constituting the controller, is light sensitive. Every microvolume excites with a probability, or intensity, proportional to its relative distance from a source of light. Therefore microvolumes at those edges of a reactor that are closer to the light target are excited and generate spreading waves of excitation that travel inward in the reactor space. Velocity vectors of the wave fronts, being inverted, indicate a direction toward the source of light. The vector, indicating the position of the light target, is used as a base to rotate the robot hosting the reactor. The waves are generated continuously during the robot movement. Therefore the direction vector may be recalculated at every step of the robot movement. Thus the robot can even chase a mobile target. This idea is illustrated in figure 3.3, where a light target, contours of wave fronts, wave velocities, and the direction vector to target are indicated. Preliminary tests with an experimental Belousov-Zhabotinsky reaction coupled to a vector-to-target extracting optical system (Adamatzky et al. 2002c) have already proved the feasibility of this approach.

click to expand
Figure 3.3: A sketch of a light-sensitive chemical controller. Target-light stimulates wave generation in a reactor. Wave dynamic determines local repulsive vector field, inverted global vector guides a robot to the target.

An unconventional, wet-ware controller would greatly benefit from the coupling with wet actuators—for example, electroactive polymers (Kennedy, Melhuish, and Adamatzky 2001) or oscillating pH-sensitive gels (Yoshida, Yamaguchi, and Ichijo 1996; Tabata et al. 2002).

Looking again at Adamatzky and Colleagues' ideas (Adamatzky et al. 1999) on wave control of robots' phototactic behavior using a Belousov-Zhabotinsky medium, one realizes there is a huge disadvantage in this approach: Local vectors, and then a global vector, are calculated "externally"—patterns of excitation in a BelousovZhabotinsky thin layer are processed by a camera coupled with not-too-complicated software, then information on the global vector is passed to the motor controllers of the robot. It would be incomparably better to couple an array of "autonomous" propulsive actuators made of chemoor magnetosensitive polymers, and thus implement a membrane of artificial paramecium cilia, which would bend coherently when waves of excitation travel along it.

A breakthrough experiment—implementation of ciliary motion in an array of gel actuators controlled by a Belousov-Zhabotinsky reaction—has been reported by Tabata and Colleagues (Tabata et al. 2002). Combining X-ray lithography with a micromolding technique, Tabata and Colleagues produced an array of conical "cilia" made of isopropylacrylamide gel (figure 3.4). This gel contains a catalyst used in the Belousov-Zhabotinsky reaction (ruthenium bypyridine). The gel array of actuators was immersed in the solution, which included all other necessary components to carry out the Belousov-Zhabotinsky reaction. The reaction was initiated in the gel and waves of oxidation spread along the gel sheet, where passing wave fronts changed hydrophilic-hydrophobic properties of polymer chains and thus caused swelling/deswelling of the gel actuators. Mechanical oscillations induced by chemical waves could therefore be observed (Tabata et al. 2002).

click to expand
Figure 3.4: A ciliate sheet: a micro-array of oscillating gel actuators is coupled with Belousov-Zhabotinsky reaction. (After Tabata et al. 2002.)




Molecular Computing
Molecular Computing
ISBN: 0262693313
EAN: 2147483647
Year: 2003
Pages: 94

flylib.com © 2008-2017.
If you may any questions please contact us: flylib@qtcs.net