Consensus-Based Distributed Information Fusion
Abstract
Statement of Work: The PIs Kevin Lynch and Randy Freeman (Northwestern University) will collaborate closely with George Jemin (Army Research Laboratory) on investigating the utility of different robust consensus algorithms for the development and implementation of distributed information fusion algorithms. Researchers would also assist the ARL researcher on developing an adaptive internal model-based robust dynamic consensus algorithm that would allow the network nodes to reach consensus on uncertain dynamic signals. These algorithms will enable the fusion of information from disparate systems and would allow shared decision-making over low-power, short lifetime sensors with limited communication capabilities. Objective: A number of sensor network consensus protocols have been proposed in the literature. In this work the PIs will investigate performance of fusion algorithms in a network of distributed sensors with limited communications range. The investigation will focus on convergence to consensus, optimality of decentralized decision-making relative to central fusion, scalability, and robustness to node failures. Approach: The PIs plan to address the following four technical challenges: (1) Lack of robustness of typical consensus-based distributed fusion approaches. They will investigate the utility of robust consensus protocols such as the proportional dynamic consensus estimator and proportional integral dynamic consensus estimator for the development and implementation of distributed fusion algorithms. (2) Lack of robust consensus protocol for uncertain dynamic signals. Based on the internal-model-based proportional-integral dynamic consensus estimator, develop an adaptive internal model based consensus protocol that could account for uncertain signal model as well as nonlinear internal model. Also investigate the possibility of extending the internal-model-based proportional integral dynamic consensus estimator such that the performance of the estimator is robust against a range of signal frequencies. (3) Lack of dynamic consensus protocols for iterative estimation problems involving nonlinear objective function. Investigate the utility of dynamic average consensus protocols for the implementation of iterative distributed fusion schemes. (4) Currently there exists no cost analysis study comparing the communication cost to the information gain associated with the consensus-based distributed fusion schemes. Theoretical analysis of communication cost associated with the distributed fusion algorithms and detailed investigation of information again associated with consensus-based fusion. Overall Merit and ONR Mission/Relevance: This work is directly relevant to ONR Information Dominance and Autonomy Focus Areas. It is expected to result insights and implementation of robust, optimal, and scalable algorithms for fusion of sensor network information.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Jun 03, 2016
- Source ID
- N000141612106
Entities
People
- Kevin M Lynch
Organizations
- Northwestern University
- Office of Naval Research
- United States Navy