Institute for Systems Research Technical Reports

Permanent URI for this collectionhttp://hdl.handle.net/1903/4376

This archive contains a collection of reports generated by the faculty and students of the Institute for Systems Research (ISR), a permanent, interdisciplinary research unit in the A. James Clark School of Engineering at the University of Maryland. ISR-based projects are conducted through partnerships with industry and government, bringing together faculty and students from multiple academic departments and colleges across the university.

Browse

Filters

2012 - 20153

Settings

Author: search.filters.author.Matei, IonSubject: search.filters.subject.distributed optimization

Search Results

Now showing 1 - 3 of 3
  • Thumbnail Image
    Item
    Nonlinear Programming Methods for Distributed Optimization
    (2015-01) Matei, Ion; Baras, John
    In this paper we investigate how standard nonlinear programming algorithms can be used to solve constrained optimization problems in a distributed manner. The optimization setup consists of a set of agents interacting through a communication graph that have as common goal the minimization of a function expressed as a sum of (possibly non-convex) differentiable functions. Each function in the sum corresponds to an agent and each agent has associated an equality constraint. By re-casting the distributed optimization problem into an equivalent, augmented centralized problem, we show that distributed algorithms result naturally from applying standard nonlinear programming tech- niques. Due to the distributed formulation, the standard assumptions and convergence results no longer hold. We emphasize what changes are necessary for convergence to still be achieved for three algorithms: two algorithms based on Lagrangian methods, and an algorithm based the method of multipliers. The changes in the convergence results are necessary mainly due to the fact that the (local) minimizers of the lifted optimization problem are not regular, as a results of the distributed formulation. Unlike the standard algorithm based on the method of multipliers, for the distributed version we cannot show that the theoretical super-linear convergence rate can be achieved.
  • Thumbnail Image
    Item
    A non-heuristic distributed algorithm for non-convex constrained optimization
    (2013-02-15) Matei, Ion; Baras, John
    In this paper we introduce a discrete-time, distributed optimization algorithm executed by a set of agents whose interactions are subject to a communication graph. The algorithm can be applied to optimization problems where the cost function is expressed as a sum of functions, and where each function is associated to an agent. In addition, the agents can have equality constraints as well. The algorithm can be applied to non-convex optimization problems with equality constraints, it is not consensus-based and it is not an heuristic. We demonstrate that the distributed algorithm results naturally from applying a first order method to solve the first order necessary conditions of an augmented optimization problem with equality constraints; optimization problem whose solution embeds the solution of our original problem. We show that, provided the agents’ initial values are sufficiently close to a local minimum, and the step-size is sufficiently small, under standard conditions on the cost and constraint functions, each agent converges to the local minimum at a linear rate.
  • Thumbnail Image
    Item
    A performance comparison between two consensus-based distributed optimization algorithms
    (2012-05-04) Matei, Ion; Baras, John
    In this paper we address the problem of multi-agent optimization for convex functions expressible as sums of convex functions. Each agent has access to only one function in the sum and can use only local information to update its current estimate of the optimal solution. We consider two consensus-based iterative algorithms, based on a combination between a consensus step and a subgradient decent update. The main difference between the two algorithms is the order in which the consensus-step and the subgradient descent update are performed. We show that updating first the current estimate in the direction of a subgradient and then executing the consensus step ensures better performance than executing the steps in reversed order. In support of our analytical results, we give some numerical simulations of the algorithms as well.