TOWARDS PRACTICAL MOBILE CROWDSOURCING: FROM MECHANISM DESIGN TO SOFTWARE IMPLEMENTATION

The Hong Kong University of Science and Technology
Department of Computer Science and Engineering


PhD Thesis Defence

Title: "TOWARDS PRACTICAL MOBILE CROWDSOURCING: FROM MECHANISM DESIGN TO 
SOFTWARE IMPLEMENTATION"

By

Mr. Yanrong KANG


Abstract

With the development of wireless networks and the proliferation of mobile 
devices, mobile crowdsourcing (MCS) has enabled us to collect and analyze 
real-word data with unprecedented coverage and intelligence. Mobile 
crowdsourcing has inspired many applications and systems, bringing great 
convenience to research, production and people’s daily life. Yet to further 
exploit MCS, we still face many challenges ranging from mechanism design to 
system implementation. In this thesis, we address three challenging topics in 
mobile crowdsourcing. They are task management for quality control, incentive 
mechanism design and mobile apps development during MCS system implementation. 
For each topic, we focus on certain scenarios and specific problems.

In the first work, we consider the quality-aware online assignment problem for 
location-based tasks, which are typical in mobile crowdsourcing. After 
mathematically formulating the online assignment problem, a polynomial-time 
online assignment algorithm is designed to optimize tasks’ overall quality. The 
proposed algorithm is proven to approximates the offline optimal solution with 
a competitive ratio of 10/7.

Crowdsouced mobile network access (CMNA), or user-provided connectivity, 
provides more flexible and ubiquitous Internet access among mobile users. In 
the second work, we study the incentive mechanism for an operator-assisted CMNA 
model, where subscribers are incentivized by a mobile virtual network operator 
(MVNO) to operate as mobile WiFi hot spots. We characterize their equilibrium 
strategies on the basis of probabilistic analysis as a two-stage Bayesian game. 
A partial cooperation strategy (PCS) is designed for MVNO and subscribers to 
optimize their benefit, which is more practical (compared with symmetric 
strategy) and incurs much less overhead.

As most MCS systems rely on mobile apps on smartphones as terminals, mobile 
apps development is a critical part in MCS systems implementation. In the third 
work we target at the resource management problem during mobile apps 
development. We mine resource management specifications in a crowdsourced way 
and a tool called Automatic Resource Specification Miner (ARSM) is developed. 
ARSM collects the usage of resources related APIs from off-the-shelf apps, 
which come from the developer crowd. Then it mines frequent patterns from the 
gathered resource usage information. Experimental results demonstrate its 
efficiency and effectiveness. Furhter we develop an Eclipse plugin to detect 
specification violations with the patterns from ARSM. It guides software 
engineers during app development by pinpointing potential resource management 
bugs.


Date:			Wednesday, 10 August 2016

Time:			5:00pm - 7:00pm

Venue:			Room 5564
 			Lifts 27/28

Chairman:		Prof. Limin Zhang (CIVL)

Committee Members:	Prof. Lei Chen (Supervisor)
 			Prof. Gary Chan
 			Prof. Huamin Qu
 			Prof. Xiangtong Qi (IELM)
 			Prof. Jianping Wang (CityU)


**** ALL are Welcome ****