Market-based Risk Allocation for Multi-agent Systems

This paper proposes Market-based Iterative Risk Allocation (MIRA), a new market-based decentralized optimization algorithm for multi-agent systems under stochastic uncertainty, with a focus on problems with continuous action and state space. In large coordinationproblems, from power grid management to multi-vehicle missions, multiple agents act collectively in order to maximize the performance of the system, while satisfying mission constraints. Theseoptimal action plans are particularly susceptible to risk when uncertainty is introduced. We present a decentralized optimization algorithm that minimizes the system cost while ensuring that the probability of violating mission constraints is below a user-specified upper bound. We build upon the paradigm of risk allocation, in which theplanner optimizes not only the sequence of actions, but also its allocation of risk among state constraints. We extend the concept ofrisk allocation to multi-agent systems by highlighting risk as a resource that is traded in a computational market. The equilibriumprice of risk that balances the supply and demand is found by aniterative price adjustment process called t\^{a}tonnement (also knownas Walrasian auction). Our work is distinct from the classical t\^{a}tonnement approach in that we use Brent's method to provide fastguaranteed convergence to the equilibrium price. The simulationresults demonstrate the efficiency and optimality of the proposeddecentralized optimization algorithm.

Authors

• Masahiro Ono
• Brian Williams

Keywords

No keywords are indexed for this paper.