Adaptive Hiking-Time Prediction Through State-Dependent Energy Regulation

Accurate prediction of hiking time is important for assessing human performance, physical activity planning, and time-sensitive field operations. Most existing approaches are environment-driven and do not account for how human capacity evolves during prolonged walking. We present a human-aware hiking-time prediction model that integrates energy expenditure with a state dependent control-theoretic energy-reserve abstraction to regulate pacing throughout a hike. The reserve mechanism serves as a computational control construct that governs cumulative workload and recovery during prolonged activity, rather than direct measurement or inference of physio logical state. The model operates without subject-specific calibration and captures cumulative workload effects over extended durations. We evaluated this approach on 2228 publicly available GPS-tracked hikes and benchmarked it against widely used environment-driven hiking-time formulations and a non-adaptive human-aware baseline. In corporating state-dependent pacing significantly improved prediction accuracy and agreement with observed hiking times. The adaptive model achieved the lowest mean absolute percent error and bias, maintained near-proportional scaling with observed hiking time, and attained the highest concordance correlations. Performance gains were greatest for longer, higher-demand hikes, while static pacing remained sufficient under short, low-demand conditions.