Applying fuzzy logic and reinforcement learning to track a mobile target using a wireless sensor network

This dissertation introduces new fuzzy logic and fuzzy logic with reinforcement learning techniques that are designed to track a mobile target as it travels within a Distributed Wireless Sensor Network (DWSN). In this research, a DWSN is defined as a network of spatially distributed sensors that each have the capability to perform distributed processing and data fusion by communicating with neighboring sensor nodes in the system. The accuracy of the target position prediction process, amount of communication between distributed sensors, and the resulting power consumption are all issues that impact the performance, reliability, and survivability of such a network. The specific DWSN that is the focus of this dissertation conforms to the Deployable Autonomous Distributed System (DADS) model. In the DADS model a field of autonomous sensors is distributed along the ocean floor. These sensors are able to communicate with one another and detect the position of a target within a specified range. The communication and active sensing operations are considered to be quite costly in terms of power consumption. Also, the physical constraints that are imposed by the undersea environments limit the achievable inter-node communication rates.; Fuzzy set theory has been successfully applied in a wide range of applications in the area of complex control systems that cannot be modeled precisely even under various assumptions and approximations. Fuzzy logic technique can be combined with other techniques such as reinforcement learning. Reinforcement learning is based upon the concept of reward and punishment, where the status of a situation will promote an action and the merits of each action will be reinforced by a positive or negative reward.; In this dissertation, an existing target tracking approach is augmented using both stand-alone fuzzy logic and fuzzy logic with reinforcement learning techniques. Simulations are presented that illustrate how these approaches compare with this well documented reference approach in terms of prediction accuracy, message frequency, and projected power consumption. The simulation results presented in this research are used to support the relative effectiveness of the enhanced approaches as compared to the unaugmented tracking approach for a variety of target movement scenarios.