Search papers, labs, and topics across Lattice.
This study evaluates the effectiveness of inverse reinforcement learning (IRL) for Arctic shipping by analyzing 3,186 AIS-derived voyages from 202 vessels over nine seasons. The authors compare a linear shared reward model, a nonlinear shared reward model, and a latent-context model, finding that while the nonlinear model significantly outperforms the linear one, the addition of latent context actually reduces performance. Their analysis reveals that vessel-level behavioral variation is primarily driven by observable factors rather than hidden preferences, suggesting that reliance on latent context may be misguided in this application.
Latent context in inverse reinforcement learning may obscure more than it reveals, as it can detract from performance by failing to capture genuinely hidden preferences in Arctic shipping.
Artificial Intelligence (AI)-assisted navigation can help Arctic shipping adapt to rapidly changing sea-ice conditions, but reliable deployment requires reward models that are interpretable and robust to changing environments. Inverse reinforcement learning (IRL) provides a framework for recovering such rewards from vessel trajectories, while recent meta-IRL methods introduce latent context variables to capture behavioral heterogeneity. However, it remains unclear whether these latent representations recover genuinely hidden preferences or simply re-encode information already available in the observed state. We conduct a controlled evaluation on 3,186 AIS-derived voyages from 202 vessels across nine Arctic shipping seasons, comparing a linear shared reward, a nonlinear shared reward, and a latent-context model built on the same nonlinear architecture. The nonlinear reward improves held-out likelihood by 50.9% over the linear baseline, whereas adding vessel-specific latent context reduces performance by 16.5%. Behavioral analysis, context probes, and a pre-registered feature-hiding ablation show that apparent vessel-level variation is largely explained by observable route and environmental conditions rather than hidden vessel-specific factors. Moreover, predictive accuracy, route fidelity, and reward transfer yield different model rankings, demonstrating that no single metric is sufficient to evaluate learned rewards. These findings motivate testing whether the observed route, environmental, and vessel features already explain behavioral variation before adding per-vessel latent context. This supports more trustworthy AI deployment in safety-critical domains.
