Exploring different coding schemes for the evolution of anartificial insect eye

摘要

The existing literature proposes various (neuronal) architecturesfor object avoidance, which is one of the very fundamental tasks ofautonomous, mobile robots. Due to certain hardware limitations, existingresearch resorts to prespecified sensor systems that remain fixed duringall experiments, and modifications are done only in the controllers'software components. Only recent research (Lichtensteiger andEggenberger, 1999) has tried to do the opposite, i.e., prespecifying asimple neural network and evolving the sensor distribution directly inhardware. Even though first experiments have been successful in evolvingsome solutions by means of evolutionary algorithms, they have alsoindicated that systematic comparisons between different evolutionaryalgorithms and codings schemes are required in order to optimize theevolutionary process. Since these comparisons cannot be done on therobot due to experimentation time, this paper reports the result of aset of extensive simulations