Distributed and Mobile Computing

The Distributed and Mobile Computing group at Xerox Research Centre India is focussed on building innovative tools and technologies that leverage the cloud infrastructure to create large-scale, optimized, and sustainable services offerings. The group works on mobile and sensing systems that positively impact multiple domains ranging from improved infrastructure, to better health and education, and increased productivity and infotainment.
The research group works in the areas of distributed systems, cloud computing, web services and SOA-enabled platforms, mobile and pervasive systems, and large-scale crowd sensing/computing. The Distributed and Mobile Computing group at XRCI works on research projects that attempt to solve challenges being addressed by Xerox Services in sectors such as banking and education services, IT Cloud services, HR services, and smart city services for government and transportation agencies.

Project Themes

Personalized Messaging Engine (PME) is the next generation messaging system to facilitate higher engagement between employees and employers, and improving employee satisfaction. PME achieves this by providing a host of configuration options to define messages in the domain of health, wealth and career. It also allows the employer to define personas which are eligible for such messages. The Messaging Engine of PME then looks at the data present within the organization as well as available external data to find the eligibility of the employees for each message defined in it. PME exposes APIs which can be consumed by web portals, mobile applications and call centre applications which in turn sends the relevant messages to the employees.

Additionally, PME uses proprietary state of the art learning engine which looks at multiple factors (Subject Matter Expert opinion, Employers’ strategy, Employees’ preferences and behaviour) to come up with a ranked list of messages for each employee. The consuming portals can then use the APIs to find out the most relevant messages for each employee. The learning process is as dynamic as the generation of new data from various sources. It is also the key towards ensuring an effective participation by the employees and thus, ensuring a highly satisfied workforce.
This is a Partner and Incubate project with Human Resource Services.


Group-sparse Embeddings in Collective Matrix Factorization
Arto Klami, Guillaume Bouchard and Abhishek Tripathi
 International Conference on Learning Representations 2014. , Banff, Canada
April 2015

Cities in emerging markets typically suffer from a wide gamut of infrastructure-related issues (e.g., waterlogging, dysfunctional streetlights, uncleared garbage etc.) as well as traffic-related issues (e.g., traffic congestion, illegal parking, etc.). Existing sensor-based solutions typically incur high installation costs as well as considerable maintenance overheads, thereby making them prohibitively expensive for city governments in emerging markets (such as India). More importantly, sensor-based solutions lack human judgment while determining context of any incidents in the city. Driven by this motivation, researchers in XRCI have created a participatory sensing (a.k.a. crowdsensing) solution called CityZen that collects reports regarding related issues from residents through an eco-system of sources, e.g., social media & other web sources, and mobile apps. The core of the CityZen is a platform, referred as the urban sensing platform, that analyzes resident reports from these heterogeneous sources into a set of actionable insights for city agencies such as municipalities, transportation agencies etc. Although motivated by the smarter-cities requirements in emerging economies, CityZen has strong alignment towards an ongoing industry trend of using crowdsourcing for transportation. Specifically, CityZen can enable customer-focused services based on real-time information from commuters, which allows proactive decision making while providing transportation solutions. Example services, where such a capability can be invaluable, include dynamic demand analysis & scheduling, dynamic trip planning, dynamic congestion control, etc. CityZen is being piloted with the Bangalore Metropolitan Transport Corporation (BMTC) to generate actionable insights from the commuter feedbacks coming from the dedicated channels of BMTC such as call centers, emails, and mobile app, as well as from online public discussions in social media (twitter) and other public web pages.


Efficient and Scalable Spatial Retrieval of Information about mobile user Involvement for Events associated with City Management
Anirban Mondal, Tridib Mukherjee, Amandeep Chugh, Atul Singh, Deepthi Chander
16th IEEE International Conference on Mobile Data Management (MDM 2015), Pittsburgh, PA, USA
June 2015
Performance Characterization and Scalable Design of Sensing-as-a-Service Platform
Mukherjee T., Kumar A., Chander D., Dasgupta K., Chugh A., Mondal A
ACM/SIGAPP Symposium of Applied Computing (SAC), Salamanca, Spain
Apr 13-17, 2015
CitR: A Scalable System for Quick Spatial Retrieval of Resident Involvement Information for Smarter Cities
Mondal A., Singh A., Rangaswamy N., Mukherjee T., Chander D
ACM/Usenix/IFIP Middleware Conference, (Poster) , Bordeaux, France
December 8-12, 2014

Forms processing is one of the major lines of businesses of Xerox Services. Various forms pertaining to many clients in different domains, viz. healthcare, transportation, etc. are digitized in offshore service delivery centres to leverage on the cost arbitration possible through outsourcing. However, in recent times, cost arbitration is losing its attraction owing to the soaring costs of infrastructure and maintenance, high attrition rates in human resources and the need to optimally manage dynamically varying workloads in a resource-effective manner. As a result, accomplishing profit margins while competing with SDCs, in such circumstances has become extremely challenging.
Crowdsourcing, is the process of obtaining services by soliciting contributions from a large group of people, primarily from the online community, as opposed to traditional means of employing resources for the same. The 24X7 available and scalable crowdsourced workforce, is highly promising for SDCs to execute voluminous tasks in a cost-effective manner. However, the outsourced tasks might not be readily crowdsourceable owing to security and confidentiality concerns, the need for guaranteed quality, turn-around-time and cost in task execution - all of which are guaranteed in off-shore delivery centres albeit at a high cost.
In this project, we have developed an end to end automated Crowdsourcing System which we call as a Virtual Service Delivery System that enables a crowdsourced execution of the digitization task while handling all the challenges of taking enterprise tasks to the crowd. This Virtual Service Delivery System can configure itself to handle multiple form types and multiple client crowdsourcing requirements and provide for a cost-optimized crowd task execution of the digitization task. The project is close to commercialization with an actual client.
This project is in alignment with the transaction processing program, and has 2 granted and 7 filed patents. It has completed one test pilot and the production pilot is underway

List of People
Avantika Gupta, Chithralekha Balamurugan, Deepthi Chander, Shruti Kunde, Madhavi Shankar

The paradigm of consumer-centric healthcare and wellness solutions is steadily shifting towards the ability to provide healthcare as a service, whereby the wellness information of a person can be seamlessly integrated into various everyday activities. The multitude of sensors that surround us in the form of smartphones, wearable devices, and infrastructure (workstation devices, WiFi, iBeacon, etc.) is a core enabler of this paradigm shift, and allows fine-grained sensing and inference of the user’s context, physiological attributes, and needs. Such sensing and detection, in tandem with intelligent intervention and persuasion techniques provide a compelling closed-loop healthcare technology for consumers. This technology need not be confined to a single application or a device, and in fact must be a part of the cyber-physical ecosystem that surrounds the user, thus realizing the “IoT of health” vision.
Towards this front, we developed a workplace wellness monitoring system which uses mobile and contextual sensors to determine the activity level of employees. With the evolution of computers, the number of desk jobs have increased phenomenally in the last few decades world-wide. In the United States, less than 20% of private sector jobs have moderate levels of physical activity, decreasing by nearly 30% compared to the early 60s. Nearly 4 out of 5 people have desk jobs in the United Kingdom. There are many large scale research studies which show that prolonged sitting is a high risk factor for health problems such as diabetes, heart attack and stroke and increased mortality, among others. Our system, called StandUp, targets such a sedentary workplace lifestyle and motivates the users to walk around after continuous stretches of sitting. An internal pilot of StandUp is currently underway.

List of People
Abhishek Kumar, Agastya Nanda (Intern), Amandeep Singh, Kuldeep Yadav, Mridula Singh, Rakshit Wadhwa (Intern), Saurabh Srivastava, Sharanya Eswaran, Tridib Mukherjee

1. Multimodal monitoring systems for physical activity (Filed)


MESSIAH stands for managing enterprise scale services via integration, analytics and hyper-personalization. The goal of this project is to drive operational excellence (e.g., increased productivity, higher quality, reduced cost) in a service delivery organization via standardized processes, platforms and and data analytics. The project is aligned with the Xerox Services 2.0 construct on building standardized platforms and processes across the capability organizations (e.g., transaction processing, customer care, and human resource services). As part of this project, researchers from XRCI closely collaborate with the Xerox Services business, both for India specific operations as well as for global operations. Research problems addressed in this project include: (a) workflow modeling and analysis, (b) understanding the health of a service organization and root cause analysis, (c) incentive design for employees, (d) understanding of employee skills and task allocation. To address these problems, we use multidisciplinary approaches from distributed computing (e.g., service oriented architecture, web services, platforms), data analytics (e.g., process mining, optimization), and social science (e.g., ethnography, user experience modeling).

List of people
Deepthi Chander, Koustuv Dasgupta, Rahul Ghosh, Avantika Gupta, Jagadeesh Prabhakara, Ajith Ramanath, Gurulingesh Raravi, Varun Sharma, Atul Singh, Jyotirmaya Mahapatra.

External collaborator: Ansuman Banerjee (Indian Statistical Institute, Kolkata)

Mobile crowdsourcing is poised as a compelling service offering for a wide range of enterprise tasks.
For mobile crowdsourcing to be truly ubiquitous, such a service needs two logically orthogonal attributes:
• It should allow users to opportunistically complete conventional tasks using their mobile devices as part of their daily lifestyle-driven activities.
• It should leverage upon natural movement trajectories/context of individuals to support an emerging category of spatiotemporally-constrained tasks tied to specific real-world locations.

Our research draws on two active disciplines of computing science research, namely mobile sensing/analytics and decision optimization to address aspects such as context sensing, context prediction and utility maximization of the worker as well as that of the mobile crowdsourcing platform. In this multi-year research effort, we address both questions from a theoretical perspective and then validate the proposed solutions empirically. For the inference and prediction of mobile context, we investigate advances in both real-time stream analytics and large-scale data analytics algorithms, applied over sensor streams generated by mobile devices. A key challenge being addressed is to preserve the accuracy of the inference process, while minimizing the potentially prohibitive energy overheads of sensing on mobile devices. The work not only results in theoretical advances in these areas, but is also being empirically evaluated and demonstrated over a relatively large set (approximately 5000) of mobile urban participants. The large-scale experimental validation is being enabled by leveraging on the globally-unique LiveLabs Urban Lifestyle Experimentation Platform at SMU, which provides near-real time, deep context for a large set (approximately 30,000) of participants at multiple urban public spaces in Singapore.

List of people
Deepthi Chander

External Collaborator: Archan Misra (Singapore Management University), Shih-Fen Cheng (Singapore Management University), Cen Chen (Singapore Management University)


TRACCS: Trajectory-Aware Coordinated Urban Crowd-Sourcing
Chen C., Cheng S., Gunawan A.,   Misra A., Dasgupta K., and Chander D
Second AAAI Conference on Human Computation & Crowdsourcing (HCOMP), Pittsburgh, USA.
November 2014

Context Aware Retail Experience

Current trends in the retail arena focus heavily on the customer shopping experience. Retailers compete in the market for expanding their customer base and this is primarily achieved using by providing competitive prices, deals and discounts to the customers. This information can often be overwhelming for the consumer and he may have difficulty in selecting a set of items on his shopping list without violating any of his budget, time or other constraints. Additionally in the current retail scenario, it is important to also understand the consumer characteristics, which are based on the consumer profile and shopping history and also gather information based on the current consumer context. While some work in this domain exists, most of the efforts are centered on online retail outlets. We build shopper personas by identifying various characteristics of shoppers and utilize this information to provide customized “value-for-money” for each shopper based on his constraints (shopping list, budget / time constraints, context).

List of People

Shruti Kunde, Gurulingesh Raravi, Deepthi Chander, Sharanya Eswaran, Priyanka Sharma, Nimmi Rangaswamy

External collaborator: Archan Misra (Singapore Management University)