Adaptive Multicast Services for LTE/eMBMS

AMuSe Project Overview


Nokia-Bell-Labs: Yigal Bejerano, Chandu Raman, Hugo Infante, Tomas Young, Chun-Nam Yu.

Columbia University: Varun Gupta, Craig Gutterman, Gil Zussman.

Motivation

Wireless multimedia (e.g., video) content delivery in crowded venues (e.g., sport arenas and lecture halls) is challenging due to lack of spectrum. When multiple users are interested in the same content, wireless multicast provides a potential solution. However, ensuring reliable and efficient delivery requires dynamically tuning parameters such as transmission rate and error correction based on the receivers feedback about their channel conditions. Specifically, for multicast, receiving accurate feedback from thousands of receivers is infeasible. Therefore, the practicality of wireless multicast for multimedia content distribution has been limited.

In a joint project of WiMNet Lab and Bell Labs we developed the AMuSe (Adaptive Multicast Services) system for scalable, efficient and reliable content distribution to very large groups of users over WiFi or LTE networks. AMuSe combines methods for collecting accurate feedback information with low overhead and for network adaptation (e.g., transmission rate) based on this feedback. Specifically, the system includes a scheme for dynamic selection of a very small subset of the multicast receivers as feedback nodes, which periodically send channel quality information to the multicast sender. Moreover, it includes schemes for dynamic rate adaptation based on the collected feedback.

As a proof of concept, we implemented the AMuSe system in the ORBIT testbed and performed extensive experiments to evaluate its performance with 150-200 receivers. Based on our large-scale experiments on ORBIT, we developed an interactive web-based demo to show the performance of AMuSe at scale. The demo won the second prize in the NYC Media Lab 2015 Summit, from about 100 demos (for more details, see the SEAS news item).

Additional information about the project can be found at AMuSe WiMNet Lab site.