The survival robots: An artificial life

Abstract

To understand behaviors of any nature living system, a lot of experiments have to be done especially on the actual living system. Alternatively, artificial life is a sub field in artificial intelligence in which living behaviors are simulated and analyzed within computers. Virtually, artificial life can be roughly used to study of living behaviors and systems. In this work, a survival game is tested toward a colony of simulated robots. A number of simulated robots are put in a survival experiment in which the robots must try to save their life as long as possible. Initially, every robot is designed to have a brain as an artificial neural network. They have the same structure of neural network, but the different internal weights, which randomly generated. To survive, the robots must be capable of finding and eating food item to regain their reduced power otherwise they die. When a robot died, a new robot is born from two remaining robots by applying the idea of crossover operation in evolutionary computing techniques. The longer the robot lives, the better chance its own weights are spread over the population. Expectedly, at the end, some interesting behaviors will emerge as the good robot programs and robot performances are also increased. Initial experiments show the good sign for emergence of interesting robot programs.