Adaptive constraint satisfaction for Markov decision process congestion games: Application to transportation networks

Under the Markov decision process (MDP) congestion game framework, we study problem of enforcing population distribution constraints on a players with stochastic dynamics and coupled costs. Existing research demonstrates that players’ can be satisfied by tolls. However, computing minimum toll value for constraint satisfaction requires accurate modeling player’s Motivated settings where an cost model may unavailable (e.g. transportation networks), consider MDP unknown We assume constraint-enforcing authority repeatedly enforce tolls converges to ϵ-optimal any given toll. then construct myopic update algorithm compute while ensuring are average. analyze how sub-optimal responses impact rates convergence towards satisfaction. Finally, Uber drivers in Manhattan, New York City (NYC) using data from Taxi Limousine Commission (TLC) illustrate efficiently reduce minimizing driver earnings.