SAGE-CAN: a slack-coupled elastic gateway for deterministic CAN-to-IoT convergence

Controller Area Network (CAN) segments and Internet-of-Things backhauls impose contradictory service disciplines: CAN is priority-arbitrated and analytically bounded, whereas Wi-Fi and MQTT transport is best-effort and subject to multi-second outages. Gateways bridging the two typically resolve this with a fixed or queue-driven aggregation window, both agnostic to the criticality of the frames they carry. This paper presents SAGE-CAN, in which the uplink aggregation window is coupled to the residual worst-case response-time slack of the CAN message set and hard-capped by a per-criticality escalation deadline, combined with tier-aware admission and delta-encoded batching. A slack estimator is derived from revised CAN response-time analysis and five algorithms are specified. Discrete-event simulation over a sixteen-message industrial rotating-machinery workload at 500 kbit/s under a Gilbert-Elliott backhaul shows tier-1 on-time delivery rising from 57.62 to 93.49 percent against the strongest baseline, mean tier-1 latency falling from 200.5 to 59.0 ms, worst-case latency bounded at 668.7 against 3299.1 ms, and uplink volume reduced 8.2-fold at 29 percent lower power. An ablation establishes that escalation capping and delta encoding dominate, while slack coupling and tiering remain dormant below approximately 85 percent bus utilisation, delimiting each mechanism's operating regime. All results are analytical or simulated; no hardware measurements are claimed and a validation protocol is specified.