Thesis defense — Florent Dobler — November 23rd

Direct-to-satellite IoT: analytical modeling, experimental characterization, and constellation routing #

The defense will take place on Monday 23 November 2026 at 14:00, Domaine universitaire, 38400 Saint-Martin-d’Hères.

Abstract #

Direct-to-satellite IoT (DtS-IoT) uses a low-Earth-orbit (LEO) satellite as a gateway for low-power IoT ground devices, without terrestrial infrastructure. This thesis makes three contributions: an analytical model of LR-FHSS channel capacity, an experimental characterization of LR-FHSS over long-range links, and satellite routing extended to lunar IoT. The analytical model of LR-FHSS is developed first. It models successful reception of header replicas and payload fragments under collision, then derives channel utilization and goodput. Compared to prior simulation-based work, the model predicts a three-fold improvement in goodput. The number of header replicas is optimized as a function of payload size: for a coding rate of 1/3, the default of 2 to 3 replicas is not always optimal. Two high-altitude balloon campaigns follow. The balloon serves as a proxy for a satellite link. The 2024 campaign confirmed free-space propagation and, by extrapolation, a link budget sufficient for ranges beyond 500 km. The 2025 campaign was intended to compare LR-FHSS with mioty. However, it met several equipment failures; its contribution is methodological rather than quantitative. The final contribution moves from the access link to the network layer. Satellite constellations are generated with STK, and traffic is routed over them with DtnSim using contact-graph routing. The number of orbital planes appears to play a role in delay for delay-tolerant IoT traffic, though this result is closely tied to the specific constellation and end-device density considered here. This routing approach extends to lunar orbits, a direction motivated by current programmes: Artemis, Moonlight and LCRNS. Selenet, a SPICE-based visualisation and link-availability statistics tool is presented. The feasibility discussion concludes that both terrestrial-relay and lunar IoT constellations remain small in scale. Together, these contributions provide an analytical capacity model for LR-FHSS predicting substantially higher goodput and device support than previous estimates, an experimentally confirmed link budget covering satellite-scale ranges with margin even at low transmission power, and a routing analysis pointing to orbital plane count as a factor in delay, alongside satellite count.