Unified semantic model for IoT data and its applications. Used in: Fiesta-IoT Platform
A Plethora of data is being generated and made available via sensing technologies. Attaching semantics to this data provides meaning to the data and helps in (a) achieving common understanding and (b) performing analysis and reasoning. A lot of ontologies such as SSN, iot-lite, and SAO, etc. have been developed to provide semantics to IoT data. However, there is no comprehensive semantic model to deal with semantic interoperability of data produced by various sensors. Thus, my key research contributions are:
Identification of semantic model [C9] that follow 4W1H (what, when, where, who and how) methodology for IoT semantics [J5, J8] and facilitate senamtic interoperability and federation between IoT testbeds [J7]
Development of proof of concept large scale crowdsensing experiment can utilize semantic model and the data and produce meaningful results
This contribution finds its implementation in the EU H2020 FIESTA-IoT project (Feb 2015 - July 2018) [B1] where, besides above, I have also led the task of architecting hetrogeneous IoT data and developing tools for performing experimentation on the federated data [J6]. My tasks were not only limited to leading and fulfilling technical duties and research but it also required supporting open call candidates. These tasks not only required robust and scalable tools but also required precise system level documentation, interaction with partners with multidisciplinary background, meeting project objectives, and writing deliverables and reports to be completed on schedule.
Large scale data acquisition and its application in transit domain. Used in: Sarathi Middleware
Public transportation is essential for the sustainable and economic development of cities. In order to understand the transportation needs that could sustain, it is also essential to understand the transit behavior of the commuters and then recommend personalizations based on the needs. This understanding can be achieved by analyzing either data acquired via Mobile CrowdSensing (MCS) applications or via getting the transit data from transport service providers. Nevertheless, acquiring data via such methods is a time-consuming activity, and further, the deployment of MCS applications lack large-scale participation of users. This makes it hard to collect the transit needs and thus limits the database required to make correct recommendations. This is referred to as cold start problem in the recommendation systems. To overcome such issues, my this research activity is focused on how large-scale data can be generated using data from the representative set and once this data is available how it can be used to generate recommendations for commuters. Further, as this data might come from different sources, how semantically this data be federated and stored to achieve efficiency. Thus, my key research contributions are:
Devlopment of middleware [C8] and prototype "MetroCognition" that aims to collect the convenience of the commuter (comfort, seat availability and waiting time).
Identification of suitable recommendation techniques that are applicable to the gathered data to provide better recommendations to the commuters
Enabling semantic interoperability of the collected data
This contribution finds its implementation in the Inria DRI/DST-CEFIPRA Associate Team Sarathi (March 2014 - September 2016) of which I was the Co-PI from October 2015 until September 2016 (end of the project). Sarathi team consists of members from IIIT Delhi.
Understanding human behavior for personalization of city wide data. Used in: 3cixty Platform
City Traveller aim to have apriori knowledge about the city the want to visit so as to optimize their travel. However, to undertand and know about a city travellers have to do exhaustive search that some time is not very useful. Leverging this fact, integration of city wide data from various sources and personalization based on own preference and behavior (such as mobility patterns) would help provide a personalized view about a city to the travellers. Thus, my key research contributions are:
Identification of semantic model [J4, C7] leveraging behavioral aspects
Development of complete proof of concept platform
user data reconciliation and investigation of privacy issues related to the reconciliation of user data.
This research if applied in EIT Digital 3cixty project (Jan 2014 - Dec 2015) which "is a platform that aims to integrate city data and offers a service platform that developers can use to create applications that enable people to instantaneously access reconciled information (including crowdsourced information) about a city with a personalized view. As a partner in the project, my involvement extends from Jan 2014 until December 2015 under the capacity of the Task Contributor in 2014 and Task Leader in 2015.
Enhancing Information Dissemination in Wireless Networks. Used in: Thesis
Information dissemination across the network is a critical aspect. Dissemination of information should be context dependent and, if needed, should reach as many devices in the network in a short time. In communication networks, those based on the device to device interactions, dissemination of the information has lately picked up a lot of interest and the need for self-organization of the network has been brought up. Self-organization leads to local behaviors and interactions that have global effects and helps in addressing scaling issues. The use of self-organized features allows autonomous behavior with low memory usage. In order to provide self-organized features to communication networks, insights from such naturally occurring phenomena are used.
Nonetheless, achieving small world properties is an attractive way to enhance information dissemination across the network. To achieve small world properties, rewiring of links in the network is performed by altering the length and the direction of the existing links. In an autonomous wireless environment, such re-organization can be achieved using self-organized phenomenon like Lateral Inhibition and Flocking, and beamforming (a concept in communication). Further, due to slow and evolutionary nature of Lateral inhibition and Flocking, it is not feasible to apply the algorithm in a mobile scenario. This leads to study of other algorithms that perform best and utilizes no extra energy and had low computational cost. Thus, in this contribution, the key questions addressed were:
how can information dissemination be enhanced across the network, either static or mobile, where device are autonomus?
what insights from nature be utilized?
The main outcomes of this contribution were: i) study of nature-inspired algorithms to increase connectivity and achieve Small World properties [J2], ii) a distributed algorithm for nodes to decide on their beamforming behavior, iii) a simulation based analysis of the achievable performance benefits of randomized beamforming and identify the challenges involved, iv) a new measure of betweenness centrality WFB (Wireless Flow Betweenness), which is used to identify beamforming nodes [J1], v) significant performance benefits can be achieved over randomized beamforming using algorithm developed in the contribution, vi) the introduction of latent states to account for variable density, vii) study of how mobility in a network with community structure could be applied to achieve large-scale dissemination in a dynamic device to device based communication network [J3], and viii), study of information dissemination in the context of Ivory Coast.
This contribution paved the way to my Ph.D. dissertation.
Development of an application platform for citywide and countrywide transport information management relying on mobile social crowd-sensing while taking into account the particular context and constraints in a region [C6].
Design of parallel processor for Natural Language Processing applications
The Project was funded by Ministry of Communications and Information Technology, India. The PI for the project was Prof. Ajai Jain of IIT Kanpur. My research includes studying the NLP Applications such as Machine Translation, Question Answering System, Text
Summarizer and Information Retrieval. The project had three phases: Research on the NLP Applications and Development of desktop application for them, development of web service and development of the processor. In the first phase, the implementation of the project was done using SystemC, C, and Java. The web service, the second phase, was implemented using Java and the front-end was developed using C sharp. The third phase used VHDL.
Specific components for Red Bull Racing Team
Developing software components essential for communication with the race car and team members during the race. Languages used include C sharp, MATLAB and SQL. The components developed became part of already existing software such as Event Request, Race Car, Fuchsia Bovine and Driver Feedback.