Robot navigation using fuzzy spatial inference

We deal with the application of fuzzy concepts directly to spatial representations in order to guide inferences based on sensed spatial parameters. Our context is exploratory mobile robotics and we describe a series of experiments on a Nomad 200 robot using a software simulator. We first present a fuzzy model of distance, alignment and sensing aperture and use this to interpret ultrasonic sensor data collected with varying degrees of noise. Wall-following is used as an example exploratory task and a non-fuzzy strategy is used as a control and to collect reference data. The design and implementation of a fuzzy map follower is then described and results are discussed. The fuzzy fusion of spatial data is shown to give markedly superior performance in dealing with noise and results show the fuzzy spatial method reduces cumulative position error by 80%. Certain situations can cause occasional overshoot and these are explained in terms of the spatial resolution of the method