Augur

The system implements a proactive communication strategy for tracking moving objects, using uncertainty estimates from predictive models to decide when to transmit updates. It supports multiple neural backbones—LSTM and Transformer—paired with either point‑prediction or mixture‑density heads for richer uncertainty quantification. A simulation engine rolls out dual‑prediction scenarios to evaluate how early updates affect tracking accuracy under a fixed communication budget. The framework is designed for researchers and developers who need to balance bandwidth constraints with positional fidelity in real‑time applications. It provides configurable strategies, training pipelines, and evaluation tools to compare periodic, reactive, and proactive approaches. The goal is to lower cumulative tracking error while keeping the number of transmissions minimal.