0000000000225862
AUTHOR
Franco Blanchini
Robust control of production-distribution systems
A class of production-distribution problems with unknown-but-bounded uncertain demand is considered. At each time, the demand is unknown but each of its components is assumed to belong to an assigned interval. Furthermore, the system has production and transportation capacity constraints. We face the problem of finding a control strategy that keeps the storage levels bounded. We also deal with the case in which storage level bounds are assigned and the controller must keep the state within these bounds. Both discrete and continuous time models are considered. We provide basic necessary and sufficient conditions for the existence of such strategies. We propose several possible feedback contr…
Min-max control of uncertain multi-inventory systems with multiplicative uncertainties
In this note, we consider production-distribution systems with buffer and capacity constraints. For such systems, we assume that the model is not known exactly. More precisely, the entries of the matrix representing the system structure may be affine functions of some uncertain time-varying parameters that take values within assigned bounds. We give stabilizability conditions that can be checked, in principle, by solving a min-max problem on the surface of the state-space (buffer level space) unit ball. Then, we consider a special case in which each uncertain parameter affects a single column of the system matrix and is independent of all the other ones. In this case, we propose a mixed int…
Average flow constraints and stabilizability in uncertain production-distribution systems
We consider a multi-inventory system with controlled flows and uncertain demands (disturbances) bounded within assigned compact sets. The system is modelled as a first-order one integrating the discrepancy between controlled flows and demands at different sites/nodes. Thus, the buffer levels at the nodes represent the system state. Given a long-term average demand, we are interested in a control strategy that satisfies just one of two requirements: (i) meeting any possible demand at each time (worst case stability) or (ii) achieving a predefined flow in the average (average flow constraints). Necessary and sufficient conditions for the achievement of both goals have been proposed by the aut…
A saturated strategy robustly ensures stability of the cooperative equilibrium for Prisoner's dilemma
We study diffusion of cooperation in a two-population game in continuous time. At each instant, the game involves two random individuals, one from each population. The game has the structure of a Prisoner's dilemma where each player can choose either to cooperate (c) or to defect (d), and is reframed within the field of approachability in two-player repeated game with vector payoffs. We turn the game into a dynamical system, which is positive, and propose a saturated strategy that ensures local asymptotic stability of the equilibrium (c, c) for any possible choice of the payoff matrix. We show that there exists a rectangle, in the space of payoffs, which is positively invariant for the syst…
ROBUST CONTROL STRATEGIES FOR MULTI—INVENTORY SYSTEMS WITH AVERAGE FLOW CONSTRAINTS
Abstract In this paper we consider multi—inventory systems in presence of uncertain demand. We assume that i) demand is unknown but bounded in an assigned compact set and ii) the control inputs (controlled flows) are subject to assigned constraints. Given a long—term average demand, we select a nominal flow that feeds such a demand. In this context, we are interested in a control strategy that meets at each time all possible current demands and achieves the nominal flow in the average. We provide necessary and sufficient conditions for such a strategy to exist and we characterize the set of achievable flows. Such conditions are based on linear programming and thus they are constructive. In …
Optimization of Long-Run Average-Flow Cost in Networks With Time-Varying Unknown Demand
We consider continuous-time robust network flows with capacity constraints and unknown but bounded time-varying demand. The problem of interest is to design a control strategy off-line with no knowledge of the demand realization. Such a control strategy regulates the flow on-line as a function of the realized demand. We address both the case of systems without and with buffers. The main novelty in this work is that we consider a convex cost which is a function of the long-run average-flow and average-demand. We distinguish a worst-case scenario where the demand is the worst-one from a deterministic scenario where the demand has a neutral behavior. The resulting strategies are called min-max…
A decentralized solution for the constrained minimum cost flow
In this paper we propose a decentralized solution to the problem of network stabilization, under flow constraints ensuring steady—state flow optimality. We propose a stabilizing strategy for network flow control with capacity constraints which drives the buffer levels arbitrarily close to a desired reference. This is a decentralized strategy optimizing the flow via the minimization of a quadratic cost of the control. A second problem characterized by non-fully connected networks is also considered, for which an exact network equilibrium is not possible. Here, the strategy, in the absence of constraints leads to a least square decentralized problem, but, unfortunately, in the presence of con…
The linear saturated decentralized strategy for constrained flow control is asymptotically optimal
We present an algorithm for constrained network flow control in the presence of an unknown demand. Our algorithm is decentralized in the sense that it is implemented by a team of agents, each controlling just the flow on a single arc of the network based only on the buffer levels at the nodes at the extremes of the arc, while ignoring the actions of other agents and the network topology. We prove that our algorithm is also stabilizing and steady-state optimal. Specifically, we show that it asymptotically produces the minimum-norm flow. We finally generalize our algorithm to networks with a linear dynamics and we prove that certain least-square optimality properties still hold.