Intelligent Edge-Assisted Crowdcast with Deep Reinforcement Learning for Personalized QoE

Wang, Fangxin; Zhang, Cong; Feng, Wang; Liu, Jiangchuan; Zhu, Yifei; Pang, Haitian; Sun, Lifeng

doi:10.1109/infocom.2019.8737456

Cited by 74 publications

(35 citation statements)

References 11 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…We implement the DQN learning network using PyTorch. The default neuron numbers in the hidden layer are 4096 plus 2048 [46]. We consider one minute as the length of a decision epoch, and trajectories of the participants involved in this decision epoch are put into the model and trained as an episode.…”

Section: B Simulation Setupmentioning

confidence: 99%

FlexSensing: A QoI and Latency-Aware Task Allocation Scheme for Vehicle-Based Visual Crowdsourcing via Deep Q-Network

Zhu

Chiang

Xiao

et al. 2021

IEEE Internet Things J.

View full text Add to dashboard Cite

Section: B Simulation Setupmentioning

confidence: 99%

FlexSensing: A QoI and Latency-Aware Task Allocation Scheme for Vehicle-Based Visual Crowdsourcing via Deep Q-Network

Zhu

Chiang

Xiao

et al. 2021

IEEE Internet Things J.

View full text Add to dashboard Cite

“…In particular, they used an unsupervised Restricted Boltzmann Machine (RBM) [28] to capture the latent features of the input data and a supervised linear classifier to estimate the characteristics of unknown videos. Wang et al [58] designed an edge computing-assisted framework that leverages DRL to intelligently assign users to proper edge servers to achieve proper video streaming services.…”

Section: ) Adaptive Streamingmentioning

confidence: 99%

Deep Learning for Edge Computing Applications: A State-of-the-Art Survey

et al. 2020

Self Cite

View full text Add to dashboard Cite

With the booming development of Internet-of-Things (IoT) and communication technologies such as 5G, our future world is envisioned as an interconnected entity where billions of devices will provide uninterrupted service to our daily lives and the industry. Meanwhile, these devices will generate massive amounts of valuable data at the network edge, calling for not only instant data processing but also intelligent data analysis in order to fully unleash the potential of the edge big data. Both the traditional cloud computing and on-device computing cannot sufficiently address this problem due to the high latency and the limited computation capacity, respectively. Fortunately, the emerging edge computing sheds a light on the issue by pushing the data processing from the remote network core to the local network edge, remarkably reducing the latency and improving the efficiency. Besides, the recent breakthroughs in deep learning have greatly facilitated the data processing capacity, enabling a thrilling development of novel applications, such as video surveillance and autonomous driving. The convergence of edge computing and deep learning is believed to bring new possibilities to both interdisciplinary researches and industrial applications. In this article, we provide a comprehensive survey of the latest efforts on the deep-learning-enabled edge computing applications and particularly offer insights on how to leverage the deep learning advances to facilitate edge applications from four domains, i.e., smart multimedia, smart transportation, smart city, and smart industry. We also highlight the key research challenges and promising research directions therein. We believe this survey will inspire more researches and contributions in this promising field. INDEX TERMS Internet of Things, edge computing, deep learning, intelligent edge applications.

show abstract

“…In contrast, the viewer scheduling in HeteroCast addresses this concern by building up a precise QoE model. Meanwhile, some work [8,21,26] shares with us the idea of providing services based on the QoE of users. Nevertheless, these methods either result in too much computation overhead as they need to be run over each individual viewer, or generate their policies based still on some simple QoE models (e.g., the weighted sum of different QoS metrics).…”

Section: Related Workmentioning

confidence: 99%

“…Such large-scale viewers may bring significant burdens to the CDN performances( §3). Second, given the various service requirements, different viewers may have their individual preference [21,26]. For example, the viewers who enjoy interacting with broadcasters are much more sensitive to stalling event 3 , while for those who only want to take a glance and frequently switch among different channels, startup latency is much more important.…”

Section: Introductionmentioning

confidence: 99%

“…Most of them have no considerations of viewer heterogeneity [15,16,19,30,31], which separates them from efficiently scheduling livecast viewers. Meanwhile, some work indeed pays attention to the viewer heterogeneity [21,26,33], but due to their ignorance of the computation overhead, it is hard for them to be practically used for the livecast scenario with large-scale viewers.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Leveraging QoE Heterogenity for Large-Scale Livecaset Scheduling

Zhang

Huang

et al. 2020

Proceedings of the 28th ACM International Conference on Multimedia

Self Cite

View full text Add to dashboard Cite

Livecast streaming has received great success in recent years. Although many prior efforts have suggested that dynamic viewer scheduling according to the quality of service (QoS) can improve user engagement, they may suffer inefficiency due to their ignorance of viewer heterogeneity in how the QoS impact quality of experience (QoE). In this paper, we conduct measurement studies over large-scale data provided by a top livecast platform in China. We observe that QoE is influenced by a lot of QoS and non-QoS factors, and most importantly, the QoE sensitivity to QoS metrics can vary significantly among viewers. Inspired by the above insights, we propose HeteroCast, a novel livecast scheduling framework for intelligent viewer scheduling based on viewer heterogeneity. In detail, Hetero-Cast addresses this concern by solving two sub-problems. For the first sub-problem (i.e., the QoE modeling problem), we use the deep factorization machine (DeepFM) based method to precisely map complicated factors (QoS and non-QoS factors) to QoE and build the QoE model. For the second sub-problem (i.e., the QoE-aware scheduling problem), we use a graph-matching method to generate the best viewer allocation policy for each CDN provider. Specifically, by using some pruning techniques, HeteroCast only introduces slight overhead and can well adapt to the large-scale livecast scenario. Through extensive evaluation on real-world traces, HeteroCast is demonstrated to increase the average QoE by 8.87%-10.09%. CCS CONCEPTS • Information systems → Multimedia streaming; • Computing methodologies → Neural networks.

show abstract

Intelligent Edge-Assisted Crowdcast with Deep Reinforcement Learning for Personalized QoE

Cited by 74 publications

References 11 publications

FlexSensing: A QoI and Latency-Aware Task Allocation Scheme for Vehicle-Based Visual Crowdsourcing via Deep Q-Network

FlexSensing: A QoI and Latency-Aware Task Allocation Scheme for Vehicle-Based Visual Crowdsourcing via Deep Q-Network

Deep Learning for Edge Computing Applications: A State-of-the-Art Survey

Leveraging QoE Heterogenity for Large-Scale Livecaset Scheduling

Contact Info

Product

Resources

About