###The price of anarchy in international trade: tackling trade with reinforcement learning
Project ID: 2228bd1061 (You will need this ID for your application)
Research Theme: Artificial Intelligence and Robotics
UCL Lead department: Computer Science
Lead Supervisor: Paolo Barucca
Project Summary:
Is global trade sustainable? Can economic agents identify optimal strategies to complex optimization problems, avoiding getting stuck in bad nonoptimal solutions due to path-dependency?
This price of anarchy has been demonstrated in some isolated contexts, such as urban mobility, but no study has focused on evaluating how much the global economy is losing from the lack of cooperation in trade and transportation networks, a loss that can be quantified in both carbon footprints and in gross domestic product.
The candidate will investigate how economic agents operating in trade networks can successfully utilize single- and multi-agent reinforcement learning to identify new sustainable solutions for global trade networks.
The goal of the project will be to dig deeper into RL strategies to achieve data-efficiency, robustness, and /or generalization, by combining discrete probability distributions and combinatorial optimization problems with neural network components.
In the first part, the candidate will define a set of mathematical problems and perform an exploratory analysis of available datasets on trade networks.
The candidate will consider the vehicle routing problem, given a set of supply networks for a set of economic agents, and a transportation network between them, what is the optimal strategy for a fleet of transporters with respect to a set of economically meaningful objective functions. The optimisation will use reinforcement learning, in a single-agent and multi-agent setting.
In the second part, the candidate will expand from an exploratory analysis to a model-based detailed analysis for international trade networks. The candidate will use the World Input-Output Database (WIOD) and Exiobase.
In the third part, the study will focus on the economic incentives and policies that could modify the objective functions, ranging from exchange rate fluctuations, political agreements, supply chain and demand shocks, and environmental challenges.