Showing papers by "Arzad A. Kherani published in 2019"

Traffic Splitting for End-to-End Delay Jitter Control in Uplink Multi-Access Systems

Abstract: We share our approach to design and implementation of a working uplink multi-access system that caters to the use-case of a mobile smart ambulance. While attempting delivery of Ultra High Definition video from this system to a backend facility, we identified the need for mechanisms using which the end-to-end delay jitter variation across the different interfaces are controlled via intelligent traffic splitting schemes. The traffic splitting approach to achieve the target jitter performance requires an in-depth understanding of jitter performance under the control of LTE’s uplink synchronous Hybrid Automatic Repeat Request (HARQ). We provide an analysis of end-to-end delay jitter under LTE’s synchronous HARQ in uplink and use the structural results so obtained to design a traffic split mechanism. We validate our analysis, and provide traffic split performance results from our working multi-access system that uses multiple LTE modems from several operators to deliver streaming traffic to a central server.

Improved Packet Detection in LoRa-like Chirp Spread Spectrum Systems

Abstract: LoRa uses Chirp Spread Spectrum (CSS) to provide long range communications for applications with infrequent data transmission requirements We provide a new approach to detecting a known transmission in presence of another Results from an SDR implementation that processes real-world traces are provided, showing that the receiver can detect presence of a known signal even if the interfering signal received power is 6dB higher Analytical model for pure-ALOHA MAC is provided to demonstrate the improvement achievable with such mechanism when an acknowledgement is desired for each transmission Applications can be written to make use of such capability at the physical layer

Network-Coded SD-WAN in Multi-Access Systems for Delay Control

Abstract: An industrial IoT network may involve hundreds of IoT end-devices deployed over large geographical wide areas networks (WAN). These devices are expected to operate over several years and their individual communication interface may be gradually upgraded with multiple communication links in order to fit the changing application and data rate requirement. As industrial networks are expected to perform with high reliability and availability, a continuous monitoring and maintenance of these communication links is essential. A software defined WAN (SD-WAN) based solution allows to adaptively monitor and manage the connectivity links as per the application requirement. In this paper we design and implement a working SD-WAN based multi-access system for upstream communication between an IoT device and the control center. Each IoT device may have multiple cellular interfaces, and we control the end-to-end delay jitter variation across the different interfaces via an intelligent traffic splitting scheme.We provide an analysis of end-to-end delay jitter under LTEs uplink connection and use the structural results so obtained to design a traffic split mechanism. We validate our analysis, and provide traffic split performance results from our working multi-access system.

An Uplink Multi-Access system for High Definition Video Transmission

Abstract: We share our approach to implementation of a working uplink multi-access system that caters to the use-case of a mobile smart ambulance. The functional design, along with some of the implementation details are provided. A way forward to implement this setup in product form is also discussed. This setup has been used to obtain results in our related paper to appear in Comsnets main conference [1].

Multi-band multi-location RF power survey and ambient RF energy harvesting circuit

Abstract: This paper presents multi-band and multi-location radio frequency (RF) survey for ambient energy harvesting As an alternative to costly spectrum analyzers, the survey was conducted using software defined radio for FM (88–108 MHz), cellular band (880 MHz) and Wi-Fi band (24 GHz) The power levels at three different locations in a tier-II smart city were recorded and plotted to get information on the available ambient RF power levels A full wave voltage quadrupler circuit (2-stage Dickson rectifier) was implemented to act as a rectifier and a voltage booster Two different types of antennae were used for capturing the ambient RF signals – wire antenna and capacitive loop antenna No impedance matching circuitry was employed and the ambient RF signals were used to charge up a capacitor connected to the rectifier output Capacitive loop antenna helped in achieving a peak capacitor voltage of 16 V and outperformed the simple wire antenna which provided a peak capacitor voltage of 13 V The charged capacitor was also used to drive an LED

An SD-WAN Controller for Delay Jitter Minimization in Coded Multi-access Systems

Abstract: For an industrial network that is expected to perform with high reliability and availability, we consider the problem of designing a low-cost approach to continuous monitoring and maintenance of the communication links A software defined wide area network (SD-WAN) based solution is proposed to adaptively monitor and manage the connectivity links from an end device as per the application requirement Each such device may have multiple cellular interfaces, and we control the end-to-end delay jitter variation across the different interfaces via an intelligent traffic splitting scheme We propose use of interstream coding to achieve target delay jitter and also a high reliability/availability We use a decoupling approach to adapt the probabilistic traffic split into various interfaces and the extent of inter-stream coding We provide an end-to-end measurement-based traffic splitting scheme that relies on a Machine Learning algorithm We use a stochastic approximation-like algorithm (operating at a slower timescale) to obtain the right coding level The various modules developed are pluggable in a pipeline-manner and work with real interfaces as well as simulators like openairinterface We validate our analysis, and provide traffic split performance results from our working multi-access system (using cellular dongles, and also over openairinterface)

On Achieving Hidden Markov Model Estimation Freshness in Sensor Networks

Abstract: Various state estimation problems in dairy farm applications, like heat detection, feeding behavior, etc. are well studied by using Hidden Markov Models (HMM) to represent the sampled process. Our use-case of monitoring cows in remote farms in real-time and perform online state estimation/classification requires intelligent sampling of the underlying HMM. Due to the limited capacity of the sensor network at the dairy farm, and due to possible wireless errors, the central server does not always receive all the sampled data. Right value of the observation arrival rate into the parameter and state estimation algorithm running at a central server is required to ensure that a prescribed heat detection performance level is achieved. We provide a working implementation of a setup where the end devices using Bluetooth Low Energy (BLE) are provided with an online recommendation on the sampling (hence, channel access) rates so that the HMM state estimated at the central server is ( $\alpha, \beta$ )-current.