Search for: All records

Dual-connectivity streaming is a key enabler of next generation six Degrees Of Freedom (6DOF) Virtual Reality (VR) scene immersion. Indeed, using conventional sub-6 GHz WiFi only allows to reliably stream a low-quality baseline representation of the VR content, while emerging high-frequency communication technologies allow to stream in parallel a high-quality user viewport-specific enhancement representation that synergistically integrates with the baseline representation, to deliver high-quality VR immersion. We investigate holistically as part of an entire future VR streaming system two such candidate emerging technologies, Free Space Optics (FSO) and millimeter-Wave (mmWave) that benefit from a large available spectrum to deliver unprecedented data rates. We analytically characterize the key components of the envisioned dual-connectivity 6DOF VR streaming system that integrates in addition edge computing and scalable 360° video tiling, and we formulate an optimization problem to maximize the immersion fidelity delivered by the system, given the WiFi and mmWave/FSO link rates, and the computing capabilities of the edge server and the users’ VR headsets. This optimization problem is mixed integer programming of high complexity and we formulate a geometric programming framework to compute the optimal solution at low complexity. We carry out simulation experiments to assess the performance of the proposed system using actual 6DOF navigation traces from multiple mobile VR users that we collected. Our results demonstrate that our system considerably advances the traditional state-of-the-art and enables streaming of 8K-120 frames-per-second (fps) 6DOF content at high fidelity. 

We investigate a novel communications system that integrates scalable multi-layer 360-degree video tiling, viewport-adaptive rate-distortion optimal resource allocation, and VR-centric edge computing and caching, to enable future high-quality untethered VR streaming. Our system comprises a collection of 5G small cells that can pool their communication, computing, and storage resources to collectively deliver scalable 360-degree video content to mobile VR clients at much higher quality. Our major contributions are rigorous design of multi-layer 360-degree tiling and related models of statistical user navigation, and analysis and optimization of edge-based multi-user VR streaming that integrates viewport adaptation and server cooperation. We also explore the possibility of network coded data operation and its implications for the analysis, optimization, and system performance we pursue here. We demonstrate considerable gains in delivered immersion fidelity, featuring much higher 360-degree viewport peak signal to noise ratio (PSNR) and VR video frame rates and spatial resolutions. 

We consider an LTE downlink scheduling system where a base station allocates resource blocks (RBs) to users running delay-sensitive applications. We aim to find a scheduling policy that minimizes the queuing delay experienced by the users. We formulate this problem as a Markov Decision Process (MDP) that integrates the channel quality indicator (CQI) of each user in each RB, and queue status of each user. To solve this complex problem involving high dimensional state and action spaces, we propose a Deep Reinforcement Learning based scheduling framework that utilizes the Deep Deterministic Policy Gradient (DDPG) algorithm to minimize the queuing delay experienced by the users. Our extensive experiments demonstrate that our approach outperforms state-of-the-art benchmarks in terms of average throughput, queuing delay, and fairness, achieving up to 55% lower queuing delay than the best benchmark. 

Internet of Things (IoT) sensors often operate in unknown dynamic environments comprising latency-sensitive data sources, dynamic processing loads, and communication channels of unknown statistics. Such settings represent a natural application domain of reinforcement learning (RL), which enables computing and learning decision policies online, with no a priori knowledge. In our previous work, we introduced a post-decision state (PDS) based RL framework, which considerably accelerates the rate of learning an optimal decision policy. The present paper formulates an efficient hardware architecture for the action evaluation step, which is the most computationally-intensive step in the PDS based learning framework. By leveraging the unique characteristics of PDS learning, we optimize its state value expectation and known cost computational blocks, to speed-up the overall computation. Our experiments show that the optimized circuit is 49 times faster than its software implementation counterpart, and six times faster than a Q-learning hardware accelerator.