ABSTRACT
This chapter presents useful theories, which includes Bayesian network and k-NN models. Bayesian network technology is very useful for encoding probabilistic knowledge as graphical structures. The chapter discusses the proposed modelling for behavioural and collision avoidance for robots which includes perception of sensor data, learning and reasoning processes of the approaches. It also presents experimental evaluations of the approaches on number of comparative evaluations in static and dynamic environments using publicly available minimum ultrasound sensors readings to obstacles. It also discusses some useful concepts and theories: Bayesian networks modelling concepts and nearest-neighbour model. With regard to the biasness check, the chapter introduces an idea of making the training instances having equal number of robot actions since its prediction of new obstacle distances is based on majority votes. Imagine a robot being allowed to autonomously avoid obstacles in an environment where it was trained-static.
