Endogenous FIFO-Batch Control for Dynamic Yard-to-Ferry Loading: A Deterministic Two-Stage Receding-Horizon Rollout

Dynamic roll-on/roll-off terminals require an executable decision between yard planning and vessel stowage: which yard lane to release, how many accessible vehicles to move, and which ferry lane to receive the ordered batch. We formulate this interface as state-aware FIFO-batch control. Each action selects one yard queue, a FIFO-prefix length, and one receiving lane/deck, while complete enumeration screens compatibility, residual capacities, discharge order, slot continuity, and a dimensionally defined normalized transverse-eccentricity envelope. The proposed beam-restricted two-stage rollout (BTR) optimizes over two batches and executes only the first before state re-observation and replanning: it expands only the W best immediate first actions and thus approximates unrestricted horizon-two minimization without forecasting future arrivals. On 180 paired synthetic paths, it loads 0.46 more vehicles and reduces weighted waiting by 50.38 units relative to myopic endogenous control, but requires 6.8 times the full-path runtime; its median, 99th-percentile, and maximum decision latencies are 0.335, 2.51, and 6.51 s. Beam and exhaustive small-instance studies show that width four preserves mean loaded count and matches the exact loaded count in seven of nine tractable cases, with a 4.56-unit mean waiting gap. Weight ablations confirm that the nominal objective is throughput-dominated, while structural tests expose rapid latency growth at 50–100 receiving lanes. The evidence therefore supports state-dependent batching more strongly than deeper lookahead. All safety statements are limited to the modeled constraints and do not constitute hydrostatic or field validation.