Skip to main content.
^ TOP

Aug. 2012 - July 2015, Research of Energy Efficiency for Crowdsensing at Scale.
Funded by National Science Council. (NSC 101-2628-E-001-004-MY3)

Crowdsensing represents an emerging genre of network applications that combine networked sensing systems with crowdsourcing techniques. It leverages the ubiquitous penetration of smart phones carried by people, and collects streams of sensing data reported by phone sensors. Crowdsensing systems have shown promise in facilitating a wide variety of applications at personal, society, and urban levels. Yet, scaling crowdsensing systems from proof of concept prototypes to real-world truly functioning systems remains extremely challenging, as it not only requires research and technique advances in multiple disciplines, but also involves incentive systems to encourage participation and contribution to crowdsensing. In this project, we tackle this research problem of crowdsensing systems at scale with a specific focus on energy efficiency. We posit that energy efficiency is the most critical factor to the success of crowdsensing systems because it needs to accommodate the trade off between the two extremes: exhaustive sensing for information accuracy of fine-granularity and conservative sensing for long lifespan. Using our deployed systems, we will investigate the adaptive duty cycle scheduling algorithm, and design an automatic tuning scheme to optimize crowdsensing systems in accordance with different environmental factors. We will study collaborative sensing techniques using heterogeneous sensors, and design context-aware duty cycle scheduling algorithms to improve energy efficiency of the overall system. Moreover, we will study compressive sensing for spatio-temporal data analysis, and apply the concept of compressive sampling to crowdsensing systems for rebuilding environment models from incomplete samples. Finally, we will investigate social persuasion techniques, and study incentive control strategies to assist and promote crowdsensing systems in real-world deployment. We will tackle the above issues by data analysis, mathematical modeling, algorithm design, implementation, experiments, and real-world deployment. The results of this work shall be beneficial for the research community of crowdsensing systems, and can go a long way in facilitating deployment and popularization of future crowdsensing systems at scale. ^ TOP

May 2011 - Apr. 2013, Pervasive Location Sensing: Data Management.
Funded by National Science Council. (NSC 100-2219-E-001-001, NSC 101-2219-E-001-002)

With the advances in wireless communications and GPS technology, pervasive location sensing applications are rapidly permeating every part of our living environments. The difference between the new genre of applications and conventional ones is that they are driven by moving objects with the location information as a function of time. With the emerging proliferation of location aware services, the amount of spatial-temporal data contributed by location sensing applications increases dramatically over time, leading to storage, transmission, and computation problems. To tackle the emerging issue of spatial-temporal data management in petabyte scale, we have proposed a novel algorithm, called Inter-Frame Coding (IFC), for trajectory compression. Using realistic spatial-temporal datasets, we demonstrate that the IFC scheme is lossless and able to achieve a high compression rate of 50%. Moreover, we propose an IFC-based range query algorithm that can reduce greatly the computational complexity of conventional range queries. We argue that the simplicity and effectiveness of the IFC scheme renders itself an ideal solution for future massive spatial-temporal data management.

In this project, we propose to research the data management aspect of emerging pervasive location sensing applications. First, based on our preliminary research results, we plan to design a set of basic spatial-temporal data query algorithms based on the IFC scheme. In addition, we design the adaptive IFC approach to optimize its performance in dynamic deployment environments, and study IFC-based data management with heterogeneous IFC configurations. Second, we plan to improve the data reliability of the IFC scheme by applying proper unequal erasure protection schemes, and enhance data security by using isomorphism functions. The two issues are essential to emerging location sensing applications, as advanced location aware applications rely on accurate data content, and volunteering participants are frustrated without privacy preservation. Finally, we investigate petabyte-scale data indexing specifically tailored to the natural of spatial-temporal data. Moreover, we study the use of MapReduce in the implementation of the IFC-based solutions, and migrate our development to existing cloud platforms.

We plan to tackle the above mentioned technical issues of massive spatial-temporal data management by performing data analysis, mathematical modeling, algorithm design, implementation, experiments, and real-world deployment. We will emphasize inter-subproject collaboration, and promote vertical and horizontal integration across the five subprojects as a whole. The results of this work shall be beneficial for the research community of spatial-temporal data management and can go a long way in facilitating future location sensing applications and enabling advanced location-based services in the near future. ^ TOP

Project Archive