Genetic programming in robot exploration

Exploration using mobile robots is an active research area. In general, an optimal robot exploration strategy is difficult to obtain. In this paper an investigation is conducted using genetic programming (GP) to solve this problem. GP is a form of artificial intelligence capable of automatically creating and developing computer programs to solve problems using the theory of evolution. However, like many other learning algorithms, GP is a computationally expensive and time-consuming process. This characteristic can impede its application where learning time is limited, such as in real-time robotic control applications. Therefore, this paper further investigates the possibility of developing a time-efficient GP algorithm to reduce evolution time. This is done by directly incorporating the amount of time evolved solutions take to form into the fitness function, in order to encourage time efficient problem solving. Experimental results have shown that while the time efficient aspect of the proposed GP algorithm is not conclusive, the robot exploration using GP produces promising outcomes