Scheduler-Agnostic Adaptive-FEC for MPQUIC: Field Evaluation over Commercial Cellular Paths
For low-latency multipath transport applications in vehicular environments, this work provides a practical adaptive-FEC solution that works with existing MPQUIC stacks without scheduler modifications.
The paper proposes a scheduler-agnostic adaptive-FEC scheme for MPQUIC Datagram that uses QUIC loss detection signals and an off-the-shelf Reed-Solomon library. In field tests over commercial cellular paths, it reduced average one-way delay from 103.0 ms to 70.8 ms, 95th-percentile delay from 281.2 ms to 142.3 ms, and packet loss from 1.7% to 0.8% at a 4.0 Mbps setting.
Multipath transport has long been studied to improve robustness over heterogeneous networks, and Multipath QUIC (MPQUIC) is particularly attractive for low-latency applications because QUIC Datagram enables transmission without retransmissions. However, this advantage comes with a key drawback: QUIC Datagram is vulnerable to packet loss, especially in vehicular cellular environments with handovers and rapidly varying radio conditions. While prior studies have explored adaptive-FEC and coded multipath designs, practical deployment remains challenging when such mechanisms require substantial modifications to the transport stack or scheduler. In this paper, we present an adaptive Forward Error Correction (FEC) scheme for MPQUIC Datagram that reuses QUIC loss detection signals and an off-the-shelf Reed-Solomon (RS) coding library. The sender estimates smoothed per-path loss rates from QUIC loss signals and adaptively determines the number of parity packets for each FEC block according to the expected packet loss. We implement the proposed scheme on an existing MPQUIC stack and evaluate it through vehicular field experiments over three commercial LTE/5G paths. Under the main 4.0 Mbps setting, the proposed scheme improves both reliability and latency compared with No-FEC: the average one-way delay is reduced from 103.0 ms to 70.8 ms, the 95th-percentile delay from 281.2 ms to 142.3 ms, and the packet loss rate from 1.7 % to 0.8 %, with an average coding rate of 0.94. These results indicate that scheduler-agnostic adaptive-FEC can provide practical latency and reliability gains for MPQUIC Datagram under the measured vehicular mobile conditions.