Markov decision theory framework for resource allocation in LEO satellite constellations

The authors address the problem of call admission and route selection for connection-oriented networks with deterministically varying topology, such as low-Earth orbit (LEO) satellite constellations. They direct their study to non-reservation routing schemes, i.e. where the use of channel reservation strategies for handover calls is omitted. The problem is formulated as a semi-Markov decision process (SMDP), where the objective is to maximise the long term network revenue, which is a combination of the revenue, earned from accepted new calls and the costs of forced terminated calls. The framework, which is novel in applications to LEO satellite networks, extends from the call admission control problem formulated as an SMDP in wireless networks to an end-to-end route allocation problem. The performance of the SMDP policy is compared with existing non-reservation-based routing algorithms. A set of results obtained from simulations in symmetric and asymmetric traffic demand regimes shows that the SMDP framework outperforms the previously proposed algorithms.