Date：October 11，2019
Time：10:00am
Venue：Room124, Nanhai Building

Topic 1：Machine-Type Communications for Internet of Things: Roadmap to a Connected World
Speaker：Liu Liang，Assistant Professor，from Hong Kong Polytechnic University

Brief Introduction：Last decades stood witness to the remarkable achievement of the wireless technologies in terms of connecting the people all over the world. Recently, there are growing interests from both academia and industry in another direction – to provide ubiquitous connectivity among machines. Such a paradigm shift from human-type communications (HTC) toward machine-type communications (MTC) is mainly driven by the emergence of Internet of Things (IoT). To pave the way for IoT, in this talk we will focus on how to embed MTC’s unique features of massive connectivity, ultra-reliability, and low latency into the 5G networks. First, we will study a massive IoT connectivity scenario in which a massive number of IoT devices can potentially connect to the network, but at any given time only a fraction of potential devices are active due to the sporadic device traffic. Based on the state evolution of the approximate message passing (AMP) algorithm, we analytically prove that the device activity detection error probability goes down to zero as the number of antennas at the base station goes to infinity. Therefore, massive multiple-input multiple-output (MIMO) ideally suits massive IoT connectivity. Second, we will study an industry automation scenario in which the controller has to send the commands to the actuators in a reliable and timely manner. Based on the observation that tasks in the factories are generally assigned to groups of actuators working in close proximity, we investigate a two-phase protocol, in which the controller sends the commands to the carefully selected group leaders in the first phase, which relay the commands to their group members via the device-to-device (D2D) communication techniques in the second phase. Such a scheme is shown to significantly improve the reliability of the low latency communications in the factories.

Topic 2：Support Recovery From Noisy Random Measurements Via Weighted L1 Minimization
Speaker：Zhang Jun，associate professor，from Guangdong University of Technology
Brief Introduction：
The problem of estimating a high-dimensional vector from limited observations arises inmany applications. To solve this problem, a classical method in compressive sensing (CS) is the ℓ1 minimization. Herein, we analyze the sample complexity of generalweighted ℓ1 minimization in terms of support recovery from noisyunderdetermined measurements. This analysis generalizes priorwork for standard ℓ1 minimization by considering the weightingeffect. We state explicit relationship between the weights and thesample complexity such that i.i.d random Gaussian measurementmatrices used with weighted ℓ1 minimization recovers the supportof the underlying signal with high probability as the problemdimension increases. This result provides a measure that is predictiveof relative performance of different algorithms. Motivatedby the analysis, a new iterative weighted strategy is proposed.In the Reweighted Partial Support (RePS) algorithm, a sequenceof weighted ℓ1 minimization problems are solved where partialsupport recovery is used to prune the optimization; furthermore,the weights used for the next iteration are updated by the currentestimate. RePS is compared to other weighted algorithms throughthe proposed measure and numerical results, which demonstrateits superior performance for a spectrum occupancy estimationproblem motivated by cognitive radio.