RFDR: a retransmission-free data reconstruction framework for emergency response networks

In emergency rescue scenarios, mobile ad hoc networks (MANETs) with dynamic topology often experience severe data missing. This compromises the reliability of observed data, which is critical for supporting real-time decision-making. Sparse inference methods exhibit significant degradation in reconstruction accuracy under high data missing rates, while retransmission-based methods introduce cascading delays and may even trigger network avalanche effects. To overcome the limitations of both methods, we propose a retransmission-free data reconstruction framework (RFDR). The framework consists of two modules: (i) a soft-sliced dynamic time warping (SS-DTW) module that identifies the most similar offline reference sample for missing value imputation; and (ii) a time-series interpolation module that captures richer contextual correlations to refine the imputation, using the reference sample and the observed sparse data as an initial estimate. To validate the effectiveness of the framework, we developed a hardware-in-the-loop (HIL) testbed that emulates dynamic wireless channel conditions. Experimental results demonstrate that RFDR achieves a 17.3% gain in transmission efficiency and a 63% reduction in median latency over standard retransmission protocols. Compared to baselines, RFDR exhibits consistently superior reconstruction accuracy across simulated random data missing rates ranging from 35% to 85%. These attributes underscore its resilience under extreme data missing conditions and indicate promising applicability in emergency settings that demand high reliability and low-latency communication.