• Two recent keywords in the 5G wireless world are O-RAN and the Open Air Interface (OAI). This project aims to implement the Split7.x architecture in OAI code base following the O-RAN principles. This creates a mutually beneficial opensource platform that can facilitate both research and commercialization in 5G area. The current OAI codebase maintains split-8 architecture where the entire signal processing is performed on a host PC exchanges I/Q samples with the radio unit for transmission and reception. The OAI system supports different radio front-end units. more The O-RAN offers a framework to manage and orchestrate the network using the concepts of near-real-time RIC and non-real-time RIC. This O-RAN fundamentally supports a split-7 architecture, where the distributed unit, called O-DU, contains Layer2 and Layer3 software and partial functionality of physical layer functionality (called HIGH-PHY). One or more remote radio unit(s) called O-RU(s), contains the physical medium level radio signal preparation processing, called LOW-PHY functionality. A control unit called O-CU will provide external interfaces to either the Mobility Management Entity (MME) in case of a non-stand-alone mode of deployment or to the 5G Core in case of stand-alone mode of deployment. less
  • Link Adaptation, Scheduling, and Feedback Techniques in Practical Wireless Systems Group: 5G Mobility PI: K V S Hari & Neelesh B. Mehta

    The availability of channel state information (CSI) at the base station (BS) plays a crucial role in determining the spectral efficiency and capacity of a cellular system. Acquiring CSI entails feedback from the user equipment (UE) to the BS. The accuracy of the CSI available at the BS affects the achievable spectral efficiency and capacity more Due to its importance, a significant amount of attention is paid to in reducing the feedback overhead in 4G and 5G standards.
    We adopt a two-pronged approach in this project. First, we investigate data-drive methods to predict the CSI from previously reported CSI and other measurements available at the BS. The data-driven approach enables the methods to be applied to and adapted to general deployment scenarios and enables the exploitation of correlation in time, frequency, and space. It also entails generating a rich set of channel data traces, i.e., data, to train and test the methods. Second, we characterize the performance gap that exists between various the schemes adopted in 4G and 5G standards and the ideal scenario in which perfect CSI is available at the base station. less
  • Interactive AR/VR digital twin for navigation and monitoring of office spaces Group: I3D Lab PI: Pradipta Biswas

    This project will develop an interactive and immersive digital twin of BT office premises in Bangalore for real time monitoring of building and help in navigation for visitors. The solution will be deployed using state of the art virtual and augmented reality systems and follow a user centred design process on developing multimodal interaction with the AR/VR system. more The IISc team has earlier worked on similar technology for developing a mixed reality digital twin of a smart manufacturing set up and also involved with developing VR cockpits with Indian Aerospace sector.  The BT team developed and deployed VR based digital system for its offices at Adastral Park, Ipswich, UK with promising result.  This project will investigate challenges in real time data visualization and interaction with AR/ VR systems in detail and can be extended to various BT offices both in India and UK. less
  • Advanced Chatbots Group: Statistics & Machine Learning PI: Prof. Ambedkar Dukkipati

    Typically chatbots in the industrial scenarios are trained for a particular task in a particular domain. Specific datasets are required for effective training such a chatbot. An example of this would be, say, restaurant booking. Seldom, such chatbots are useful in a larger context. British telecom was facing this problem, of user intention and domain change while in chat. more A user starts the chat for a certain intent, but changes the intent, maybe as well as the domain in the course. For example, a user starting the chat for flight booking, and realizing the cost of connecting flight is not economical and proceeds for hotel booking overnight. The main aim of this project is to develop for chatbot system that can deal with challenging scenarios like domain change and intent change. less
  • Conversational Speech Analytics Group: LEAP Lab PI: Prof. Sriram Ganapathy

    Automatic transcription of speech conversations is a challenging problem due to the spontaneous nature of conversational speech; presence of multiple speakers and speaker overlap.Speaker diarization and automatic speech recognition (ASR) are the two important tasks in a conversational speech transcription system. more Speaker diarization addresses the problem of "who spoke when?" in the conversation and performs a segmentation of the conversation recording into speaker homogeneous regions. The ASR system then transcribes the segmented speech, providing speaker annotated speech transcripts. This project investigates the development for speaker diarization and speech recognition solutions that are robust to various conversational artifacts. The topics of interest include conversation context aware neural architectures, accent invariance/adaptation of systems and multi speaker ASR approaches. less
  • Inter-Access Points CoordinationGroup: ECE Department PI: Prof. Chandramani Singh and Prof Neelesh Mehta

    Inter-AP (or inter-BSS) coordination is the key to efficient utilization of time, frequency and spatial resources in WiFi networks. In the state-of-the-art Wi-Fi networks, such coordination, albeit present, is limited to slow timescale functionalities such as seamless roaming between networks and channel selection. moreIEEE 802.11be task group (TGbe) has proposed several ways of tightening coordination of transmit powers, transmission schedule and antenna gain patterns. We pursue design and analysis of joint transmit power and channel access control among neighboring APs. We first focus on coordinated spatial reuse (Co-SR) in IEEE 802.11be networks. We study impact of OBSS/PD thresholds and corresponding transmit powers on network performance. Co-SR is believed to be useful when inter-BSS interference is light, but the channel state is perceived as busy. However, systematic modeling and evaluation of gains of Co-SR is warranted. We also model OFDMA based channel access, in particular, management of resource units (RUs) for random access and scheduled access. We aim to develop an RU allocation algorithm that adapts to traffic patterns and yields good throughout and delay performance. less