Incentive Contracts in Multi-agent Systems: Theory and Applications

Abstract

This thesis studies incentive contracts in multi-agent systems with applications to transportation policy. The early adoption of emerging transportation systems such as electric vehicles (EVs), peer-to-peer ridesharing, and automated vehicles (AVs) relies on governmental incentives. Those incentives help achieve a specific market share target, prevent irregular behaviors, and enhance social benefit. Yet, two challenges may impede the implementation of such incentive policies. First, the government and subsidized organizations must confront the uncertainty in a market; Second, the government has no access to the organizations' private information, and thus their strategies are unknown to it. In the face of these challenges, a command-and-control incentive policy fails.

In Chapter 2, we revisit the primary setting in which a government agency incentivizes the OEM for accelerating the widespread adoption of AVs. This work aspires to offset the negative externalities of AVs in the ``dark-age'' of AV deployment. More specifically, this chapter designs AV subsidies to shorten the early AV market penetration period and maximize the total expected efficiency benefits of AVs. It seeks a generic optimal AV subsidy structure, so-called ``two-threshold'' subsidy policy, which is proven to be more efficient than the social-welfare maximization approach.

In Chapter 3, we develop a multi-agent incentive contracts model to address the issue of stimulating a group of non-cooperating agents to act in the principal's interest over a planning horizon. We extend the single-agent incentive contract to a multi-agent setting with history-dependent terminal conditions. Our contributions include: (a) Finding sufficient conditions for the existence of optimal multi-agent incentive contracts and conditions under which they form a unique Nash Equilibrium; (b) Showing that the optimal multi-agent incentive contracts can be solved by a Hamilton-Jacobi-Bellman equation with equilibrium constraints; (c) Proposing a backward iterative algorithm to solve the problem.

In Chapter 4, we obtain the optimal EV and charging infrastructure subsidies through the multi-agent incentive contracts model. Widespread adoption of Electric Vehicles (EV) mostly depends on governmental subsidies during the early stage of deployment. The governmental incentives must strike a balance between an EV manufacturer and a charging infrastructure installer. Yet, the current supply of charging infrastructure is not nearly enough to support EV growth over the next decades. We model the joint subsidy problem as a two-agent incentive contract. The government observes two correlated processes -- the EV market penetration and the charging infrastructure expansion. It looks for an optimal policy that maximizes the cumulative social benefit in the face of uncertainty. In our case study, we find that the optimal dynamic subsidies can achieve 70% of the target EV market share in China by 2025, and also maintains the ratio of charging stations per EV.

Chapter 5 ends the thesis with conclusions and promising future research directions. In summary, this thesis provides a new approach to appraise transportation and energy policies against exogenous and endogenous risks.