ISAC as a priced, prioritized, intent-driven service.

Dense-venue localization and coverage intents with arbitration under load

A sensing intent declares a localization accuracy target (e.g. 0.1m) and a resource ceiling for a geographic scope. The Near-RT RIC derives concrete sensing configuration with caps applied; a closed loop retunes toward the accuracy and coverage targets. The field-shaped pilot scenario in the InfraSense roadmap targets exactly this: dense-venue localization and coverage intents with arbitration under load.

Sensing intents for public safety — prioritized and resource-constrained

Public safety sensing intents are declared as critical priority — protected in full by the SMO Global Agent when aggregate commitments over-commit the network. Lower-priority consumers are trimmed to their declared floors before any critical sensing intent is reduced.

Vehicle-to-everything sensing over the 5G/6G RAN

V2X perception requires real-time inference with hard latency bounds — a natural fit for the RT latency class (<10ms), which admits only DU-adjacent / RT-RIC-adjacent infrastructure. The inference layer enforces tier placement and refuses requests that cannot meet the bound rather than silently degrading.

Sensing data feeding continuously calibrated digital twins

High-rate channel, reflection, and mobility data produced by ISAC must be processed close to where it is produced. InfraSense places inference at the appropriate tier and feeds the output into the AODT digital twin — keeping the simulation calibrated against live network conditions for what-if planning and xApp validation.

Beam management, interference mitigation, and integrated sensing

ML models that make or inform RAN control decisions carry hard deadlines — near-real-time loops of 10ms–1s and real-time loops below 10ms. InfraSense classifies each request and places it on the correct tier, selecting among model variants (fast / balanced / accurate) to meet accuracy within the latency bound.

Standards-aligned deployment across heterogeneous RAN estates

InfraSense attaches to A1, E2, E3, O1, and O2 as defined by O-RAN — deployment means adding a policy type, an orchestrator, and an SMO service, not replacing the stack. The roadmap includes exercising the inference layer and A1 policy flow against at least one additional non-proprietary or partner RIC implementation.