On Risk Evaluation and Control of Distributed Multi-Agent Systems

Aray Almen,Darinka Dentcheva
2023-11-17
Abstract:In this paper, we deal with risk evaluation and risk-averse optimization of complex distributed systems with general risk functionals. We postulate a novel set of axioms for the functionals evaluating the total risk of the system. We derive a dual representation for the systemic risk measures and propose a way to construct non-trivial families of measures by using either a collection of linear scalarizations or non-linear risk aggregation. The new framework facilitates risk-averse sequential decision-making by distributed methods. The proposed approach is compared theoretically and numerically to some of the systemic risk measurements in the existing literature. We formulate a two-stage decision problem with monotropic structure and systemic measure of risk. The structure is typical for distributed systems arising in energy networks, robotics, and other practical situations. A distributed decomposition method for solving the two-stage problem is proposed and it is applied to a problem arising in communication networks. We have used this problem to compare the methods of systemic risk evaluation. We show that the proposed risk aggregation leads to less conservative risk evaluation and results in a substantially better solution of the problem at hand as compared to an aggregation of the risk of individual agents and other methods.
Optimization and Control
What problem does this paper attempt to address?
The paper attempts to address the problem of risk assessment and risk avoidance optimization in distributed multi-agent systems. Specifically, the authors focus on how to assess the overall risk of the system in complex distributed systems and propose a new method to construct systemic risk measures to achieve more effective risk avoidance decisions. ### Main Issues 1. **Assessment of Overall System Risk**: - How to comprehensively consider the risk of each agent and its contribution to the overall system risk? - In some cases, risk assessment may be based on confidential or proprietary information. How to conduct effective risk assessment in such cases? 2. **Methods of Risk Aggregation**: - How to aggregate risk factors from different sources into a loss function? This is a complex problem in many systems because the risk characteristics of different agents may be highly nonlinear and difficult to represent in analytical form. - For example, in automated robotic systems, information exchange may be limited or distorted by factors such as operating speed and spatial distance. 3. **Limitations of Existing Methods**: - Existing risk assessment methods may be effective in dealing with financial systems, but they may not be applicable to complex systems such as robotics, energy systems, business systems, and logistics problems due to the heterogeneity of risk sources and the complexity of system relationships. - Traditional risk measurement methods (such as VaR) may be overly conservative when dealing with multidimensional risks, leading to a decline in the quality of solutions. ### Main Objectives of the Paper - Propose a new risk assessment framework suitable for risk avoidance decisions in distributed multi-agent systems. - Define risk measures on random vectors by introducing a set of new axioms and derive the dual representation of these measures. - Propose several methods to construct systemic risk measures and analyze their properties. - Validate the effectiveness of the new methods through numerical experiments, especially in the application of wireless communication networks. ### Solutions - **New Axiomatic System**: The authors propose a set of new axioms to define risk measures on random vectors. These axioms include convexity, monotonicity, positive homogeneity, and translation invariance. - **Dual Representation**: Derive the dual representation of systemic risk measures, reducing assumptions on multidimensional risks. - **Risk Aggregation Methods**: Propose two methods to construct systemic risk measures, namely using linear scalarization and nonlinear risk aggregation. - **Distributed Optimization Method**: Propose a two-stage distributed optimization method to solve the systemic risk assessment problem. This method allows each subsystem to optimize its operation with minimal information exchange with other subsystems. ### Experimental Validation - The authors compare the effectiveness of different systemic risk measurement methods through numerical experiments. The results show that the newly proposed risk aggregation methods can reduce conservativeness and improve the quality of problem-solving. Overall, the paper addresses the key issues of risk assessment and optimization in distributed multi-agent systems by proposing new risk assessment frameworks and methods, providing theoretical support and technical means for practical applications.