RL Dispatch Under Information Latency
Digital-twin experiment on reinforcement-learning dispatch for autonomous haul fleets when telemetry arrives late.
Project: Real Autonomy in Mining — Digital-Twin Experiments
Notebook 1 of the Real Autonomy in Autonomous Haulage series. Studies what happens to an autonomous haul fleet when the dispatcher’s picture of the site is late (telemetry latency): whether simply correcting the stale picture with the dispatcher’s own recent actions is enough, or whether a reinforcement-learning agent adds anything on top of that. Compares three information-lag scenarios across six dispatcher policies, including a cross-deployment test of what happens when an agent (or rule) is tuned for the wrong latency. Site and operator names are anonymized.