In this paper, we propose a novel environment modeling method for local homing for indoor mobile navigation. An environment model is constructed from a set of omnidirectional sensor data obtained by rotating an ultrasonic sensor and a vision sensor. We develop a sensor fusion method to deal with uncertainties in the two types of sensor data. The proposed sensor fusion method uses fuzzy numbers to represent subjective knowledge and manipulates uncertain quantities using fuzzy arithmetic. Using the proposed method, we can determine a more precise geometric relation between the current location and the target location. To evaluate the proposed environment modeling and to verify the sensor fusion method, experiments are carried out in an indoor environment.Recently various types of sensors have been applied to obtain more accurate data and to improve the intelligence because a single sensor does not provide sufficient information needed for an intelligent system[7], [8]. Such multiple sensors increase the reliability of the sensory information and provide more complete information about the environment by combining partial information from each sensor. But, as various types of sensors are involved, an appropriate sensor fusion method is required to handle the uncertainties effectively. Sensor fusion, as defined in this paper, refers to any stage in the integration process where there is an actual combination( or fusion) of different sources of sensory information into one representational format. The fusion of data or information from multiple sensors or a single sensor over time can take place at different levels of representation[9], [10]. So far, the probabilistic approach has dominated much of the work on the representation and manipulation of uncertainties [ 11]. However, it is difficult to obtain probability density functions for a variety of environments [12]. Furthermore, if we know the probability density functions(p.d.fs), the calculation with p.d.fs is highly complicated [13]. However, when fuzzy arithmetic is used, we can represent the environment using sensor information with more ease and also incorporate the expert's knowledge about the properties of sensors.Inherent uncertainties in the sensory information, which can arise due to noise, changes in the environment, or changes in the system itself, can be represented in different ways. If we use fuzzy numbers, the degree of possibility can be easily represented. When a sensor provides a measured value m about a feature f of an environment, the sensor is actually providing a linguistic proposition of the form of "the feature f is about m". A possibility distribution can be determined to represent the degree of possibility that any x in the referential set X might actually be the true value f [ 14]. In general, the mean value of the fuzzy number represents the actual sensor measurement and the spread of the fuzzy number represents the uncertainty level, which reflects the physical properties of each sensor.

No abstract

No abstract