Real-time joint Landmark Recognition and Classifier Generation by an Evolving Fuzzy System

A new approach to real-time joint classification and classifier design is proposed in this paper. It is based on the recently developed evolving fuzzy systems (EFS) method and It applied to mobile robotics. The approach items from subtractive clustering method and its on-line evolving extension called eClustering. A new formula for data potential (spatial density) determination based on the participatory learning and data scatter concepts is introduced In the paper that Is computationally simpler and more intuitive. An EFS-based sell-organizing classifier (eClass) is designed by automatic labeling the landmarks that are detected in real-time. The proposed approach makes possible fully autonomous and unsupervised joint landmark detection and recognition without the use of absolute coordinates, any communication link or any pre-training. The proposed algorithm is recursive, non-iterative, one pass and thus computationally inexpensive and suitable for real-time applications. Extensive simulations as well as real-life tests has been carried out in an indoor environment (an office located at InfoLab21, Lancaster University) using Pioneer3 DX mobile robotic platform equipped with sonar and motion sensors and on board PC. The results indicate superior rates of recognition, flexibility, and computational demands of the proposed approach comparing with the previously published similar methods, further investigations will be directed towards development of a cooperative scheme, tests in a realistic outdoor environment, and the presence of moving obstacles.