Congestion-dependent pricing of network services
IEEE/ACM Transactions on Networking (TON)
Dynamic Programming and Optimal Control
Dynamic Programming and Optimal Control
Proceedings of the 33nd conference on Winter simulation
IEEE/ACM Transactions on Networking (TON)
Approximate Gradient Methods in Policy-Space Optimization of Markov Reward Processes
Discrete Event Dynamic Systems
Simplification of network dynamics in large systems
IEEE/ACM Transactions on Networking (TON)
Infinite-horizon policy-gradient estimation
Journal of Artificial Intelligence Research
Hi-index | 0.00 |
We introduce a set of algorithms for pricing calls on a multiclass loss network with unknown demand elasticity. The algorithms are design to observe the network and use real-time pricing to estimate demand elasticity and other unknown system parameters, and modify per-class prices in order to improve the long-run average revenue. The algorithms can be implemented online, have small memory and computational requirements, and are robust to parametric uncertainty. We provide sufficient conditions for the convergence of the algorithms to a local optimum, and illustrate their performance through numerous numerical examples. The paper also discusses how these algorithms can be distributed to multiple agents on a per-class basis, and provide bounds to error estimates introduced by our decoupling approach.