Context-Aware Collective Decision Making in Distributed Environments

Typically sensor networks rely on centralized information processing where the sensed information is relayed to a network hub. The network then processes the information and adjusts the network according to the perceived state of its environment. In this paper we propose a novel collective decision making approach. Instead of relaying all the sensor information to a central location, the nodes process the information locally and generate local decisions. This decision level information is then relayed to a cluster head where decision level fusion is performed to obtain a network level decision. A systems approach has been utilized here, such that all nodes in the sensor network are based on different underlying machine learning models and an overall system decision is obtained by combining the outputs of all the constituent systems. Dempster-Shafer fusion has been performed.