Abstract—The rapid growth of online video traffic has intensified the need for Adaptive Bitrate (ABR) mechanisms capable of maintaining stable Quality of Experience (QoE) under diverse network conditions. While neural-based ABR approaches have shown promising results in IP-based streaming, their applicability to Named Data Networking (NDN) remains underexplored. This paper presents an experimental evaluation of a lightweight neural-based ABR scheme for NDN video streaming. Five representative ABR algorithms—Buffer-Based, Rate-Based, Simple, Hybrid, and Neural—are implemented using a customized Shaka Player integrated with the NDNts library. Experiments are conducted on an NDN-enabled testbed under low (2–3 Mbps), medium (5 Mbps), and high (10–15 Mbps) bandwidth conditions. Performance is evaluated using QoE-related metrics including startup latency, buffering behavior, dropped frames, and an overall QoE score. The results demonstrate that the proposed Neural ABR consistently achieves lower startup latency, reduced buffering, and higher QoE compared to conventional approaches across all bandwidth scenarios. These findings indicate that lightweight neural models can effectively enhance playback stability and QoE in NDN-based streaming systems, providing a practical bridge between IP-centric ABR designs and emerging content-centric networking architectures.


Keywords—Named Data Networking (NDN), Adaptive Bitrate (ABR), neural networks, video streaming, Quality of Experience (QoE)