Smarter satellite links with AI dApps
A distributed application that runs AI inference at the network edge, co-located with the RAN, delivering decisions in under a millisecond, where traditional HARQ feedback cannot reach.
Today’s Challenge, Solution and Result
From 540 ms blind spots to 40 Āµs decisions.
ā Challenge
GEO round-trip delays of ~540 ms disable HARQ feedback, leaving the network blind to channel conditions.
ā Solution
A distributed AI app co-located with the RAN predicts optimal link parameters from real-time SNR and CRC inputs in under 1 ms.
ā Result
50% reduction in block error rate (3.04% ā 1.53%), validated end-to-end with Tier 1 partners.
10 to 100 times faster than xApps and rApps
dApps operate at the network edge with sub-millisecond control loops, far faster than near-real-time xApps or non-real-time rApps for latency-critical NTN decisions such as uplink adaptation and beam hopping.
dApps
<1 ms
Latency-critical edge AI
xApps
10 msā1 s
Near-real-time RIC
rApps
>1 s
Non-real-time RIC