Zero-Downtime Telematics: A Dual TCU and DSDA Framework for Multi-Modal Redundancy

Abstract

The transition to SAE Level 4 and 5 autonomous vehicles (AVs) demands connectivity architectures with strict deterministic behavior and extreme availability. Traditional telematics control units (TCUs) exhibit single-point-of-failure vulnerabilities, rendering them inadequate for safety-critical cooperative perception and remote teleoperation. This paper presents a fail-operational telematics framework integrating Dual TCUs with Dual SIM Dual Active (DSDA) technology. Using a Doer-Fallback node logic over a Time-Sensitive Networking (TSN) Automotive Ethernet backbone, the architecture completes a deterministic hardware failover in an average of 0.85 ms measured from the missed-heartbeat event. A Multi-Path TCP (MPTCP) scheduler is combined with an eight-layer Long Short-Term Memory (LSTM) predictive mobility model to orchestrate concurrent heterogeneous network streams (5G Advanced and NTN) and to execute pre-emptive edge-network handovers. In simulation, handover-induced service interruption falls from 45.3 ms under a reactive 3GPP Rel-15 baseline to 4.2 ms, alongside a 38.6% increase in peak uplink throughput. The contribution is a cross-layer synthesis of hardware redundancy and predictive transport-layer control, shifting telematics from reactive connection recovery toward proactive continuity for software-defined vehicles (SDVs). All results reported here are outputs of an ns-3 and SUMO simulation model; no physical vehicle testing is claimed.