Study and development of distributed detectors with fast convergence

• Grant number: 18/26040-2
• Support Opportunities: Scholarships in Brazil - Doctorate
• Effective date (Start): May 01, 2019
• Effective date (End): January 31, 2022
• Field of knowledge: Engineering - Electrical Engineering - Telecommunications
• Principal Investigator: Vitor Heloiz Nascimento
• Grantee: Allan Eduardo Feitosa
• Host Institution: Escola Politécnica (EP). Universidade de São Paulo (USP). São Paulo , SP, Brazil

Abstract

Nowadays there is a large interest in developing detection techniques using distributed networks, which have received a considerable attention due to their economy of energy and resource, robustness to node and link failure, scalability, and improved tracking performance. Detection algorithms may in general be based on Neyman-Pearson (NP) detectors, which maximize the probability of detection given a maximum false alarm rate, or on Bayesian detectors, which minimize the probability of error in detection. In the case of algorithms implemented in distributed networks, using connections only with neighboring nodes, each node must decide between concurrent hypotheses about the environment state where the network is inserted.The candidate's master degree project investigated the transient performance of distributed detectors, and a binary Maximum Likelihood detector was proposed - i.e., when the prior probabilities of the Bayesian detector are equal. The decision process was based on local measurements and on estimates shared among nodes using the distributed diffusion LMS estimation algorithm. A specific initialization of the developed algorithm accelerates its performance, significantly reducing the time required for the convergence of the error probability. In this project, we propose to generalize and deepen the already achieved results to novel situations, developing a more complete analysis of transient performance of distributed detectors. We will develop solutions for NP and Bayesian detectors using different prior probabilities and for situations with more than two hypotheses, looking for theoretical generalizations of the new results, as well as consider the case of nonstationary environments (tracking). (AU)