Signalling and obfuscation for congestion control

Abstract

We aim to reduce the social cost of congestion in many smart city applications. In our model of congestion, agents interact over limited resources after receiving signals from a central agent that observes the state of congestion in real time. Under natural models of agent populations, we develop new signalling schemes and show that by introducing a non-trivial amount of uncertainty in the signals, we reduce the social cost of congestion, i.e., improve social welfare. The signalling schemes are efficient in terms of both communication and computation, and are consistent with past observations of the congestion. Moreover, the resulting population dynamics converge under reasonable assumptions

On the ergodic control of ensembles

“…In the remainder we will assume y i (k) and π(k) are scalars, but generalisations are easy to obtain. The randomness can be a result of the inherent randomness in the reaction of user i to the control signal π(k), or the response to a control signal that is intentionally randomized (Schlote et al, 2013(Schlote et al, , 2014Marecek et al, 2015). The aggregate resource utilisation y(k) := N i=1 y i (k) at time k is then also a random variable.…”

Resource allocation with population dynamics

“…In practice, it may be very hard for the transportation authorities to steer the system to-ward the equilibrium, while maintaining the trust of the users, which may be require the information provided to be truthful, in some sense. [17,19] show that when one considers the truthful information provision, announcing a function of c m (n m t−1 ) to all users at time t, the equilibria do not exists, in general. For example, in the simple case of truthful provision of costs c m (n m t−1 ), with users' choice based on a scalar perceived utility, it is easy to observe a limitcycle behaviour in a two-route example, where the traffic alternates between the routes, c.f., Appendix 1 of [19].…”

“…Still, one may want to consider both the the costs c m (n m t ) and their perceptions as non-stationary random processes, generated by the underlying non-linear dynamics. Mareček et al [17,19,18] have studied measure-theoretic notions of convergence within such models, where c m (n m t ) are non-stationary random processes, based on a closed-loop model with u m t and v m t being functions of the history of n m t−1 , n m t−2 , . .…”

“…Transportation authorities increasingly have information on traffic conditions throughout the road network at this level of detail, but want to broadcast much less information to the public. Mareček et al [17,18,19] have recently demonstrated that such information provision should be seen as means of modulating the demand process and that different means of modulation may lead to very different outcomes. Whereas, for example, it is natural for each user to forecast the travel time as a scalar for each route, if the transportation authorities announce two distinct scalar values for two distinct alternative routes, the route with the lower scalar value announced may become congested, and if the authorities are truthful in reporting the congestion, the congestion may alternate between the two alternative routes in a sub-optimal limit-cycle behaviour, ad infinitum.…”

Information-Theoretic Limits of Strategic Communication

“…They determine whether the optimal encoding policies are linear or based on quantification. Sender-receiver games are also investigated in [33], [34] for the problem of "strategic estimation" involving self-interested sensors; and in [35] for the "strategic congestion control" problem. In [36], [37], [38], the authors investigate the computational aspects of the Bayesian persuasion game when the signals are noisy.…”