Budgeted Twin Synchronization over V2X Uplinks for Electric Vehicle Routing

Abstract

Connected electric vehicle (EV) fleets depend on cloud digital twins that rely on telemetry over lossy Internet of Things (IoT) and vehicle-to-everything (V2X) uplinks. When updates are lost, stale or budget-limited, twin ages and dispatch plans can fail. This paper studies two coupled decisions: when to spend an uplink message and how to plan from a stale twin. We compare periodic age-of-information (AoI) threshold and deviation triggers in a full factorial with and without a conservative twin correction based on split-conformal prediction (CP) under three communication regimes. Within each trigger pair, attempts, receptions, and AoI are identical. Across 144 confirmatory cells, raw triggers that plan on uncorrected snapshots remain at 0.0% strict feasibility. With a six attempt uplink cap and correction, AoI threshold triggering feasibility reaches 38.2%, with 4.51 mean attempts, versus a score of 36.8% for deviation+CP and 35.4% for periodic+CP. Uncapped hybrid admission does not improve feasibility, but reduces mean twin age (0.111 versus 0.166). Conservative correction is therefore a mechanism that restores feasible plans; trigger choice mainly trades messages against freshness. The routing solver is a fixed adaptive large neighborhood search (ALNS) module used only as an application key performance indicator (KPI).