KYDON, a self-organized autonomous net: learning model and failure recovery

摘要

In this paper, a learning model and a failure recovery approach of an autonomous vision system multi-layer architecture, called KYDON, are presented. The KYDON architecture consists of "k" layers array processors. The lowest layers compose the KYDON's low level processing group, and the rest compose the higher level processing groups. The interconnectivity of the processors in each array is based on a full hexagonal mesh structure. The lowest layer array processors captures images from the environment by employing a 2-D photoarray. The top most layer deals with the image interpretation and understanding. The intermediate layers perform learning and pattern recognition processes to bridge the image information flow from the bottom most layer to the top most one. KYDON uses graph models to represent and process the knowledge, extracted from the image. An important feature of KYDON is that it does not need any host computer or control processor to handle I/O and other miscellaneous tasks. A novel learning model has been developed for the KYDON's distributed knowledge base.