Real-Time Distributed Optimization in Networked Multi-Agent Systems

Abstract

The overarching goal of the project is to establish a systematic on-line use of optimization techniques for real-time applications in networked environments with time-varying stochastic features. Unlike off-line optimization, where one can normally afford computationally intensive methods, an on-line setting requires fast solutions for time-critical decision making, often in response to unpredictable events. This implies the need to frequently re-solve already hard problems, since conditions in the operating environment are constantly changing. The focus of the project has been on multi-agent networked systems and the framework developed rests on three conceptual cornerstones: the event-driven paradigm for optimization, data-driven methodologies for optimization algorithms, and scalable on-line solutions for optimal control problems. The three main objectives are: 1. Develop explicit on-line solutions for dynamic optimization problems in networked systems. 2. Address the question: When is decentralization possible in networked system optimization? and develop distributed algorithms for dynamic optimization in such systems. 3. Address the problem of multiple local minima in network system optimization through efficient ways of escaping such local optima.

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Jun 16, 2023
Accession Number
AD1230518

Entities

People

  • Christos Cassandras

Tags

Fields of Study

  • Computer science

Readers

  • Distributed Systems and Data Platform Development
  • Operations Research