Search

Farm records hold the static, temporal, and longitudinal details of the farms. For small-scale farming, the ability to accurately capture these records plays a critical role in formalizing and digitizing the agriculture industry. A trusted exchange of these records could unlock critical insights to different stakeholders across the value chain. Recently, there has been increasing attention on digitizing small scale farming with the goal of increasing farm-level transparency and visibility, access to credit, etc. using these farm records. However, most solutions proposed so far have the shortcoming of providing granular and trusted small-scale farm digitization. To address these challenges, we present a system, called AG-Wallet System(AGWS), which leverages blockchain to formalize the interactions and data flow in small-scale farming ecosystem. Utilizing instrumentation of farm tractors, we demonstrate the ability to utilize farm activities to create trusted electronic field records (EFR). Using AGWS, we processed over one hundred thousand small-scale farm-level activity events for which we also performed automated farm boundary detection of several farms.

We investigate the effect of variational autoencoder (VAE) based data anonymization and its ability to preserve anomalous subgroup properties. We present a Utility Guaranteed Deep Privacy (UGDP) system which casts existing anomalous pattern detection methods as a new utility measure for data synthesis. UGDP’s approach shows that properties of an anomalous subset of records, identified in the original data set, are preserved through the anonymization of a VAE. This is despite the newly generated records being completely synthetic. More specifically, the Bias-Scan algorithm identifies a subgroup of records that are consistently over- (or under-) risked by a black-box classifier as an area of ’poor fit’. This scanning process is applied on both pre- and post- VAE synthesized data. The areas of poor fit (i.e. anomalous records) persist in both settings. We evaluate our approach using publicly available datasets from the financial industry. Our evaluation confirmed that the approach is able to produce synthetic datasets that preserved a high level of subgroup differentiation as identified initially in the original dataset. Such a distinction was maintained while having distinctly different records between the synthetic and original dataset.

Bias in data can have unintended consequences which propagate to the design, development, and deployment of machine learning models. In the financial services sector, this can result in discrimination from certain financial instruments and services. At the same time, data privacy is of paramount importance, and recent data breaches have seen reputational damage for large institutions. Presented in this paper is a trusted model-lifecycle management platform that attempts to ensure consumer data protection, anonymization, and fairness. Specifically, we examine how datasets can be reproduced using deep learning techniques to effectively retain important statistical features in datasets whilst simultaneously protecting data privacy and enabling safe and secure sharing of sensitive personal information beyond the current state-of-practice.

Many intelligent transportation systems (ITS) in cities with developed economies are making use of mobile technology as data sources (e.g., many crowd-sourced traffic-related applications) to improve the quality and efficiency of transportation networks. Often, these data sources are used to supplement existing traffic monitoring equipment (e.g., ground-loop detectors, traffic cameras), to provide greater insights into roadway infrastructure and traffic dynamics. For cities with emerging economies where traditional traffic monitoring equipment is cost prohibitive, the rise in mobile technology presents a unique opportunity to leverage smartphone sensors as an alternative data source for ITS. There are, however, challenges to using these sensors particularly with the cost of mobile data, network consistency, and on-device resources. In this paper, we present a mobile system that instruments roads under resource constraint while a vehicle is in motion. It determines when and what data to collect and/or upload using a number of on-device valuation and optimisation functions, by prioritising data collection over uploading or vis-versa. We deployed our mobile system on a fleet of heavy-duty waste-collection trucks in Nairobi, Kenya to collect a large volume of real-word road infrastructure and mobility data. Results show that a 42 % reduction in wireless transmissions costs can be achieved with minimal impact to the time in which important data are collected, uploaded and harmonized into a frequently updated map of road infrastructure and traffic.

Driving in developing cities presents numerous challenges. Traffic congestion and traffic accidents are the most visible challenges which are caused by different underlying factors. Two chief factors are poorly planned and maintained roadway infrastructure and the decisions made by the drivers. Drivers are constantly forced to negotiate road hazards, like potholes, unlabeled speed bumps as well as moving obstacles, like pushcarts, motorcycles, and animals. Current Usage Based Insurance (UBI) models do not include the context which in many cities may be paramount to understanding driver behavior. This article presents the Context-based Driver Score (CDS) model as a unified model for scoring a driver based on a unique formulation of context that includes road quality. We demonstrate the CDS model on a real-world use case in Nairobi, Kenya, where waste-collection trucks were instrumented with smartphones in order to collect inertial and telematic data. We present an analysis of the CDS model and driver behaviors in contexts that include weather, time-of-day, and road quality. Our results show that the distribution of driving behaviors, like harsh braking and swerving, vary greatly based on the context and the definition of the CDS model. Ultimately, this work aims to extend the utility and scalability of UBI models in order to make them more suitable for deployment in developing cities.

At the peak of the West African Ebola outbreak in December 2014, the only prospect for containing the Ebola Virus Disease (EVD) was through behavior change to reduce transmission pathways. However, deeply ingrained beliefs, social systems, and cultural practices, combined with a lack of knowledge of the disease, made containment difficult. IBM Research - Africa, in collaboration with Sierra Leone Open Government Initiative (OGI), set up a call and SMS hotline to engage with citizens and understand issues on the ground using free text responses, which started on 19th Sept 2014 (during the first national lockdown). Localized versions of Centre for Disease Control (CDC) public health messages were broadcast on public radio along with a short question on Ebola, and listeners were encouraged to call or SMS the OGI hotline in reply to the question, or to report the current situation in their locality. Messages were categorized using a basic keyword search approach, based on categories that were determined manually, and the results of the analysis were communicated to the government via OGI. The initial deployment showed the potential of using a 2-way communication platform utilizing natural language responses to provide valuable feedback on behaviors and attitudes during a health emergency. Our presentation will describe the original platform, the outcomes of the analysis performed to date and the current work being undertaken to extend and scale the platform in Sierra Leone.

In this poster, we present a mobile crowdsourcing application, called Afya Ya Barabara (AYB) (which means road health in Kiswahili). The application serves 2 major roles: First is to collect and visualize data on road surface conditions in Nairobi. Second role is as a citizen engagement platform to raise awareness on the state of road conditions in Nairobi using Twitter hashtags. The focus of this work is to discuss how AYB is employed to promote citizen engagement. Nairobi is one of the most active cities in East Africa in terms of Twitter usage. This guided our decision to employ Twitter as a social engagement tool. The demand for more systematic citizen engagement and public participation in the governance of developing cities is growing. As government budgets are becoming increasingly tight, new ways of addressing road maintenance problems are required. Public governance in developing countries faces accountability issues that allow corruption and poor service delivery to seethe. Lack of information and citizen engagement perpetuates such tendencies. We propose Afya Ya Barabara as an engagement platform that will expand the capacity of municipalities and its citizens to better monitor road conditions using mobile technology.

Rapid urbanization of developing cities creates challenges for governments to update and maintain the city infrastructure. One particular challenge is to plan for road capacities that will sustain the growing demand for surface transportation on the road network. For these cities, a sophisticated road infrastructure that includes sensors (e.g., loop detectors and traffic cameras) and/or probe vehicles equipped to detect pavement roughness are ideal for traffic infrastructure management. However, the cost of such tools for developing cities is often prohibitively expensive. For these reasons, road infrastructure management strategies that are able to leverage low-cost technologies and data resources are highly sought after in these areas. The authors refer to these strategies as frugal innovations. In this paper the authors explore a frugal approach for monitoring road quality in Nairobi, Kenya. The authors aim is to use characteristics of speed distributions to detect the presence of pavement distresses (including speed bumps) for Nairobi, Kenya's road network. The authors first derive link speeds from mobile global positioning system (GPS) traces, and then perform graphical as well as quantitative distributional comparisons of the speed distributions on travel network links with speed bumps present, compared to links without speed bumps present. The authors results support the hypotheses that distributional comparisons on travel speeds can indeed serve as useful tools to detect the presence of speed bumps and other road distresses.

According to the UN-HABITAT, the city of Nairobi loses half a million USD daily due to congestion on roads designed for a city 10 times smaller. Therefore, there is a great need for traffic management and awareness solutions. Many existing solutions are unsuitable for cities like Nairobi due to economic constraints, dynamic events, uncertainty, and poor infrastructure. Recently, a novel approach called Frugal Innovation has been adopted at IBM Tokyo Research. The approach combines very low quality images (VLQI) captured by existing low-cost cameras with network flow algorithms to accurately estimate traffic flow. We extend their work to develop a mobile app, called Twende-Twende, that provides drivers with real-time traffic information and suggested routes. We incorporate locally relevant context (such as references to landmarks) to predict congestion and create traffic awareness. We deployed the app and evaluated its effectiveness, accuracy and usability. Our initial evaluation indicates that the app enhances the driving experience and can be deployed in other developing countries.

A design is presented for a nanoelectromechanical optical switch based on the horizontal deflection of an input waveguide to align with one of two output waveguides. The use of high-index (GaAs) strip waveguides surrounded by air, designed to be single-mode at 1.55 /spl mu/m, significantly decreases device dimensions as compared to previous designs. Design tradeoffs between optical and mechanical properties of the device are discussed. By means of three-dimensional numerical simulations, optical transmission is optimized for two different design strategies: butt and parallel coupling. High polarization-independent transmission (over 90%) is predicted for realistic design parameters.