Waseda University AI improves 5G multicast video delivery reliability to 87%

This development addresses the critical lack of retransmission channels in 5G broadcast standards, which previously led to frequent stream freezing. By shifting from reactive retransmission to proactive AI-driven adjustments, operators can deliver high-definition video over wireless spectrum with reliability comparable to fixed fiber. This technical shift enables the convergence of traditional broadcasting and mobile broadband on a single, spectrally efficient platform. For the broader ecosystem, it provides a viable path for local 5G networks to replace aging CATV infrastructure in underserved regions. Watch for whether 3GPP adopts these AI-native predictive models into future Release 18 or 19 specifications for global standardization.

5G multicast and broadcast services are moving from laboratory research into commercial trials across multiple markets. In South Korea, SK Telecom and KBS conducted a live 5G broadcast trial in March 2025 delivering UHD content to mobile devices using ATSC 3.0 and 5G MBS infrastructure, demonstrating that terrestrial broadcasters can reach mobile audiences without relying on unicast data channels. Japan's NHK has similarly invested in 5G broadcast research, with NHK Science & Technology Research Laboratories publishing results on 5G MBS field trials for simultaneous multi-device delivery at sports events during 2024 and 2025, showing that multicast can handle high-density viewer scenarios that would overwhelm unicast networks. These deployments underscore why Waseda University's AI-driven approach to stabilizing multicast delivery arrives at a commercially relevant moment.

The regulatory and standards landscape for 5G broadcast is also tightening. The 3GPP Release 17 specification introduced the foundational MBS framework, and 3GPP Release 18 work items include enhancements for MBS service continuity and power efficiency, with discussions ongoing about whether Release 19 will incorporate AI-native network functions for broadcast optimization. In Europe, the European Broadcasting Union has advocated for 5G broadcast as a complement to DVB-I and DVB-NIP standards, arguing that multicast delivery over cellular spectrum can extend public-service broadcasting reach without requiring every viewer to consume unicast bandwidth. The Waseda team's lightweight model, which operates at the edge rather than in core network functions, aligns with this regulatory push toward spectrally efficient delivery that does not burden operator infrastructure.

On the technical side, independent benchmarks are beginning to quantify the gap between multicast and unicast performance for live video. A 2025 study from the 5G Media Action Group found that 5G MBS reduced network load by up to 70% compared to unicast streaming for simultaneous viewers of the same live event, though the study also noted that packet loss rates in multicast mode remained significantly higher than unicast without additional error-correction mechanisms. Waseda University's 87% error-free rate represents a substantial improvement over the 32% baseline cited in their paper, but it still falls short of the near-zero loss rates that unicast TCP-based delivery achieves through retransmission. The research team, led by Professor Jiro Katto, has indicated that further model refinements targeting channel-state prediction accuracy could push error-free rates above 95%, which would bring multicast reliability into a range suitable for premium live sports and news broadcasting.