When to arrive in a congested system: Achieving equilibrium via learning algorithm
P. Thaker, A. Gopalan, R. Vaze
Players sense an intermittently-available server, paying for each sample and earning less when they overlap with others, so the hard part is that no one coordinates yet the timing choices interact. We give a distributed randomized rule where each player adapts on its own and the dynamics settle on a unique fixed point that is a Nash equilibrium. The framing fits competitive WiFi sensing and contention for attention, where sensing cost and congestion both matter.
Abstract
We consider a strategic problem where multiple players compete to access a shared server platform that operates intermittently, switching between ON and OFF periods. Each player incurs costs to sample the server state and receives payoffs inversely proportional to the number of simultaneously connected players. We propose a distributed randomized learning algorithm that enables players to minimize sensing costs while converging to a unique fixed point that constitutes a Nash equilibrium. The work addresses applications in competitive WiFi sensing and competition for user attention in social networks.
Read the explainer — intuition, the key idea, and honest limitations