A Novel Path Prediction Strategy for Tracking Intelligent Travelers

There are various technologies for positioning and tracking of intelligent travelers such as wireless local area networks (WLAN). However, the loss of actual positioning data is a common problem due to unexpected disconnection between tracking references and the traveler. Disconnection of the mobile terminal (MT) from the access points (AP) in WLAN-based systems is the example case of the problem. While enhancement of the physical system itself can reduce the risk of disconnections, complementary algorithms provide even more robustness in localization and tracking of the traveler. This research aims to develop a novel path prediction system which could keep track of the traveler during temporary shortage of actual positioning data. The system takes the advantage of the past trajectory information to compensate for the missing information during disconnections. A novel decision support system (DSS) is devised with the ability of learning decisional as well as kinematical behaviors of intelligent travelers. The system is then used in path prediction mode for reconstructing the missing parts of the trajectory when actual positioning data is unavailable. An ActivMedia Pioneer robot navigating under fuzzy artificial potential fields (APF) and blind-folded human subjects are the two types of intelligent travelers. The reactive motion of robots and path planning strategies of the blinds are similar in that both of them locally acquire knowledge and explore the space based on route-like spatial cognition. It is proposed and shown that route-like intelligent motion is based on a combination of decisional and kinematical factors. The system is designed in such a way to integrate these two types of motion factors using causal inference mechanism of the fuzzy cognitive map (FCM). The FCM nodes are a novel selection of kinematical factors. Genetic algorithm (GA) is then used to train the FCM to be able to replicate the decisional behaviors of the intelligent traveler. Experimental works show the capabilities of the developed DSS in human path prediction using both simulated and actual WLAN-based positioning dataset. Locational error is set to be limited to 1 m which is suitable for wireless tracking of human subjects with up to 10% improvement compared to the most related works. Both simulation and actual experiments were also carried out on the Pioneer platform. The accuracy in prediction of robot trajectory was obtained about 83% with considerable improvement compared to the recent methods. Apart from the positioning algorithm of this dissertation, there are several applications of this DSS to other areas including assistive technology for the blind and human-robot interaction.

Preview PDF ABS__FK_2009_103.pdf Download (203kB)

Actions (login required)

View Item