Rigorous Systems Research Group (RSRG) Seminar

The Internet is plagued with congestion problems of growing severity which are worst at peak periods. In this talk, we discuss schemes that incentivize users to shift part of their usage from the peak-time to the off-peak time. We propose a raffle-based scheme which provides a random reward distributed in proportion of each user's fraction of the total shifted usage. We compare it to a more traditional time-of-day pricing scheme which gives a fixed reward per unit of shifted usage.

Using a game-theoretic model, we show that both schemes can achieve an optimal level of decongestion at a unique Nash equilibrium. We compare the schemes' sensitivity to uncertainty of the users' utilities and exhibit an interesting self-tuning property of the raffle-based scheme. Our model is based on a nonatomic game with an infinite number of players. We show that this corresponds in terms of equilibrium to the limit of an atomic game when the number of players goes to infinity.