An Optimization Framework for Resource Allocation in Multi-Tenant Communication Networks

A multi-tenant ground network must allocate resources to users with heterogeneous needs. Many real-world networks rely on manual scheduling or simple algorithms that yield sub-optimal schedules. This results in underutilization. We present a novel decision-theoretic approach that leverages utility functions and optimization techniques to efficiently compute schedules that best fulfill a portfolio of “user preferences”. These user preferences are encoded in an expressive mathematical form which is intuitive for satellite network operators allowing them to express abstract needs in a precise and robust way.Whereas manual scheduling requires a precise selection of a time window for a spacecraft to communicate with a particular ground station, our approach allows task selection to be fully automated. Existing algorithmic approaches typically offer a “low level” solution requiring users to invest resources in configuration and monitoring. Our approach presents a “high level” set of parameters which enable fast configuration and self-correction. Every possible schedule in the solution space can be scored using our weighted user preference functions, providing a deterministic and transitive rank order over all schedules.Our work also includes an optimization algorithm that computes network downlink schedules in a runtime that scales linearly in the number of user preferences in the system. This efficient algorithm can be effectively done in close to real-time, allowing for the network to dynamically reallocate and reschedule as new information arrives.Our paper will present the framework, algorithm, and simulation results demonstrating how optimal schedules can be found, especially in cases of over-subscribed networks.