Advances in Game Theory and Multi-Agent Systems

The field of game theory and multi-agent systems is witnessing significant developments, with a focus on improving the modeling and analysis of complex interactions between agents. Researchers are exploring new methods to enhance the flexibility and robustness of game theoretical models, including the integration of deep learning techniques and the development of novel optimization algorithms. Notably, the application of neural ordinary differential equations to mean-field game theory has shown promise in reducing modeling bias and improving the accuracy of predictions. Additionally, advances in inverse reinforcement learning and adversarial game playing are enabling more effective analysis and mitigation of stealthy attacks in cyber-physical systems. The development of peer-aware cost estimation frameworks is also facilitating mutual learning and intent inference in multi-agent systems. Overall, these innovations are advancing our understanding of complex strategic interactions and are expected to have significant impacts on fields such as cybersecurity, robotics, and autonomous systems. Noteworthy papers include:

  • Modelling Mean-Field Games with Neural Ordinary Differential Equations, which proposes a novel approach to modeling mean-field games using neural ordinary differential equations.
  • Recursive Deep Inverse Reinforcement Learning, which introduces an online recursive deep inverse reinforcement learning approach to recover the cost function governing an adversary's actions and goals.

Sources

Modelling Mean-Field Games with Neural Ordinary Differential Equations

Recursive Deep Inverse Reinforcement Learning

Strategic Planning of Stealthy Backdoor Attacks in Markov Decision Processes

Integrated Control and Active Perception in POMDPs for Temporal Logic Tasks and Information Acquisition

Is Learning Effective in Dynamic Strategic Interactions? Evidence from Stackelberg Games

Nash Equilibrium Learning In Large Populations With First Order Payoff Modifications

Quantitative Strategy Templates

Adversarial Knapsack for Sequential Competitive Resource Allocation

Path Integral Methods for Synthesizing and Preventing Stealthy Attacks in Nonlinear Cyber-Physical Systems

PACE: A Framework for Learning and Control in Linear Incomplete-Information Differential Games

Peer-Aware Cost Estimation in Nonlinear General-Sum Dynamic Games for Mutual Learning and Intent Inference

Built with on top of