Online In-Route Task Selection in Spatial Crowdsourcing

Abstract

Consider the following scenario: (a) a worker traveling on the shortest path between two locations in a city's road network, (b) he/she is willing to deviate from such path in order to complete tasks in the network, (c) tasks are associated with rewards and appear and disappear dynamically, i.e., they are not known in advance, and (d) the worker specifies a time budget which limits the total time he/she is willing to spend on his/her trip. Now assume the worker wants to minimize the detour from the original path while, at the same time, maximizing the rewards collected by completing tasks; clearly two competing criteria. We call this problem the Online In-Route Task Selection (Online-IRTS) query, and we investigate it using the paradigm of skyline queries in order to systematically explore different trade-offs between earned rewards and path deviation. Because of the online nature of the problem, i.e., irrevocable decisions about which task to perform have to be made without knowledge of future tasks, it is not possible to guarantee optimal solutions for the Online-IRTS query. Therefore, we propose two heuristic approaches, one is based on local optimizations, and the other one is based on incremental solutions, along with a method to evaluate the quality of their solutions w.r.t. the optimal offline solution. Our experiments using city-scale realistic datasets show that the first approach is more effective whereas the second is more efficient, allowing one to choose which approach to use according to his/her priorities.