Junting CHEN – Research Areas

I work on signal processing, communication, optimization, control, and statistical learning, for various applications including

Dynamic 3D Communication Networks

The recent success of commercial UAV applications motivates a new dimensino of optimizing wireless networks, where the network dynamically optimizes the node positions for enhanced commmunication capability. This can be done by introducing a flying relay to connect two geographically separated ground users. The fundamental novelty and challenge is to adapt the flying relay position to the fine-grained geographical resource landscape, such as to avoid local shadowing (caused by buildings and trees, etc.) surrounding the users and to establish good communication channels to them. In particular, there are two main challenges to be solved:

combine massive MIMO with D2D 
  • Low complexity methods to learn and predict the air-to-ground channels that capture the fine-grained shadowing details

  • Efficient algorithms to optimize the UAV position despite the complex propagation structure

Solutions to these problems require synergies between communication theory, machine learning, and optimizations. Some preliminary results can be found in:

  • J. Chen and D. Gesbert, ”Optimal Positioning of Flying Relays for Wireless Networks: A LOS Map Approach”, in Proc. IEEE Int. Conf. Commun., Paris, May 2017. [Slides][IEEE]

  • J. Chen, U. Yatnalli, and D. Gesbert, ”Learning Radio Maps for UAV-aided Wireless Networks: A Segmented Regression Approach”, in Proc. IEEE Int. Conf. Commun., Paris, May 2017. [IEEE]

End-to-end capacity map  Search trajectory for optimal UAV position 

Machine Learning for Network Localization and Communication

Source Localization without Energy-decay Models

In practice for many applications, the signal propagation characteristics are not fully known due to the complex environment, such as for an acoustic signal in water. We proposed a model to arrange a number of measurements taken at random locations into a sparse matrix, and proved that the matrix has a unimodal property that can be leveraged to localize the source. We developed a non-parametric localization method based on a smart integration of various modern techniques: sparse matrix completion, eigenspace analysis, singular vector perturbation analysis, and peak localization. The method demonstrates robustness as well as order-wise reduction in the number of sensors over the state-of-the-art baselines.

energy samples for source localization 

We are now interested in fusing multimodal data, such as energy measurement and time-of-arrival measurement, for better localization performance. We are working on a tensor observation model and developing methods to mine a common unimodal structure using tensor decomposition techniques. Our preliminary result suggests that jointly processing the data in a tensor works better and separately manipulating each data matrix.

Our preliminary results can be found in [J13, C15, C16] [Slides] [Preprint].

Learning 3D Propagation Map

In the past, statistical predictions were made based on the UAV-to-user distance (as well as elevation angle, etc.); this approach is not feasible for low altitude UAVs, where deterministic shadowing is the main factor to conquer (or exploit). We developed a segmented propagation model and reverse-engineered a hidden multi-class virtual obstacle map from a set of UAV-to-user channel gains measured at different UAV-user location pairs. I demonstrated that such a model not only captures the propagation effect of the fine-grained terrain structure (channel prediction error reduced by half compared with baseline), but also yields a nice “nested segmented propagation” property, that can be exploited for optimal UAV positioning. We are now intersted in developing tensor processing techniques to reconstruct the 3D propagation map by exploiting the sparse structure, which can be also observed from the following figures.

channel reconstruction

Preliminary results:

  • J. Chen, O. Esrafilian, D. Gesbert, and U. Mitra, “Efficient algorithms for air-to-ground channel reconstruction in UAV-aided communications,” in Proc. IEEE Global Telecomm. Conf., Dec. 2017, Wi-UAV workshop.

Massive MIMO

Massive MIMO assisted by (device-to-device) D2D communications

combine massive MIMO with D2D 

We show that if the users can share a limited amount of CSI with each other via D2D, the downlink performance can be significantly improved in a limited feedback massive MIMO system under FDD mode. Specifically, we have solved three major problems:

  • (i) For a limited amount of CSI exchange via D2D, whether it is better to feedback the quantized CSI or the precoder;

  • (ii) How the users exchange the CSI efficiently, and

  • (iii) How to design a good feedback and precoding strategy under practical D2D (finite rate and latency).

Preliminary results can be found in:

  • J. Chen, H. Yin, L. Cottatellucci, D. Gesbert, “Efficient feedback mechanisms for FDD massive MIMO under user-level cooperation,” IEEE Trans. Wireless Commun., 2017, [Preprint] [Code].

  • J. Chen, H. Yin, L. Cottatellucci, D. Gesbert, “Dual-regularized Feedback and Precoding for D2D Assisted MIMO Systems,” IEEE Trans. Wireless Commun. accepted, 2017, to appear. [Preprint] [Code]

Low complexity massive MIMO precoding algorithms

We exploited both the spatial and temporal channel correlations to reduce the computational complexity for the precoding. We formulated optimization problems and develop tracking algorithms using compensation techniques and calculus on Grassmann manifolds. The results for point-to-point MIMO and multi-cell multiuser MIMO were presented in the following papers.

  • J. Chen and V. K. N. Lau, “Multi-stream iterative SVD for massive MIMO communication systems under time varying channels,” in Proc. IEEE Int. Conf. Acoustics, Speech, and Signal Processing, Florence, May 2014, pp. 3152–3156.

  • J. Chen and V. K. Lau, “Two-tier precoding for FDD multi-cell massive MIMO time-varying interference networks,” IEEE J. Sel. Areas Commun., vol. 32, no. 6, pp. 1230–1238, Jun. 2014.

Cooperative Localization from A Game Perspective

energy samples for source localization 

Cooperation among users can significantly enhance the localization accuracy in GPS-challenging environments. The intuition is that the users can fuse the the location information by taking range measurements with each other. However, cooperation is resource consuming (power and bandwidth etc.) and the outcome is not necessarily fair among agent nodes. Our research addressed these issues by developing distributed link selection and power allocation strategies under different application scenarios. Specifically, we formulated serval resource management games according to different types of resource constraints in the network and studied their solution properties via analyzing game equilibria.

We found closed form conditions to determine whether a link should not be considered for cooperation; with that, a lot of coordination overheads among agents can be saved. In particular, we developed a novel link bargaining equilibrium concept, found sufficient conditions where the equilibrium is unique, and proposed iteration that can converge to the unique equilibrium. Such a novel game theoretic design and analysis framework may apply to other cooperation networks as well. [J9, J10, C8].

Algorithms Tracking Analysis and Enhancement via Compensation Techniques

My thesis work focuses on the design and analysis of algorithms for network resource optimization under time-varying channels. The motivation is that network resource optimization usually requires online algorithms with explicit signaling among wireless nodes. Due to signaling latency, the algorithm cannot converge fast enough, and the time-varying nature of the wireless network yields moving optimal solutions. We developed a mathematical framework to analyze the algorithm tracking performance, and based on that, we developed a class of compensation algorithms that predict the potential variation of the unknown optimal solution for enhancing the tracking performance.

As one modern application, we studied multi-timescale stochastic algorithms. Motivationally, wireless communication systems are driven by multi-timescale time-varying parameters. Traditionally, people usually treated control variables individually in different layers separated by different timescales, instead of putting them all in the same optimization framework. By contrast, we tried to derive online multi-timescale algorithms from solving a unified optimization problem in terms of control variables in different timescales. The advantage is that the slow timescale control variable (and control) can make its update before the fast timescale variable converges. However, the challenge is to understand what the convergence error will be, and how much performance to gain. Our results have been published in [J7, J8, C3]

With these general problems, we have attempted several challenging scenarios driven by various applications:

  • Parameter evolves as a Markov chain:

    • Network utility maximization (NUM) under finite state channel models [J1]

  • Parameter evolves as a stable continuous process:

    • NUM problem [J2]

  • Parameter (e.g., queue length) is coupled with control action:

    • Delay-aware resource allocation for ad hoc networks [J3]

  • Parameter evolves in two-timescale:

    • Cross-layer resource optimization for heterogeneous networks [J7, J8, C3]

  • Multiple local optima (non-convex problem):

    • Low complexity precoder tracking for massive MIMO systems [J5, C7]

End-to-end capacity map  Search trajectory for optimal UAV position 

The algorithm design and analysis framework is summarized in my thesis dissertation [Thesis].