Optimal Dynamic Incentive Contracts between a Principal and Multiple Agents in Controlled Markov Processes: A Constructive Approach

This paper investigates an optimal dynamic incentive contract between a risk-averse principal (system operator) and multiple risk-averse agents (subsystems) with independently local controllers in continuous-time controlled Markov processes, which can represent various cyber-physical systems. The principal fs incentive design and the agents f decision-makings under asymmetric information structure are known as the principal-agent (PA) problems in economic field. However, the standard framework in economics cannot be directly applied to the realistic control systems including large-scale cyber-physical systems and complex networked systems due to some unrealistic assumptions for an engineering perspective. In this paper, using a constructive approach based on the techniques of the classical stochastic control theory, we propose and solve a novel dynamic control/incentive synthesis for the PA problem under moral hazard.