Location predication-based task assignment enhancement in spatial crowdsourcing
Ziyuan Zhao · 2015
The ubiquity of mobile devices has brought the popularity of a new problem solving mechanism - spatial crowdsourcing, which utilizes the power of crowds to accomplish location-specific tasks. Many spatial-crowdsourcing-based applications have emerged and deeply influenced our daily life, such as taxi taking, package dispatching and food delivering. Many unified and standardized crowdsourcing services adopt the server assigned tasks(SAT) mode, in which the system proactively assigns tasks to workers in proximity of requested locations. Under this task assignment mode, the travel cost between workers and tasks becomes of vital importance, less travel cost means less response time and higher task acceptance ratio. In this thesis, we formally define the minimum travel cost assignment (MTCA) problem in spatial crowdsourcing. Since we consider the total travel cost during a period of time, we explore the possible locations of future tasks to assist task assignment planning. By adopting various task-distribution prediction algorithms, we propose several effective and scalable approaches for task assignment. We conduct comprehensive experiments on real-world data to compare the effectiveness and efficiency of our proposed solutions.