Crowdsourcing Task Assignment In Complex Scenarios

Crowdsourcing aims to use human wisdom to solve problems that are difficult to handle with machines alone.In recent years,crowdsourcing researchers have proposed the concept of complex tasks.Such tasks are often multi-skilled and require complex computing operations that cannot be completed by a single crowdsourcing worker.The assignment of complex tasks has become a research hotspot in recent years.Complex tasks can be decomposed into several sub-tasks independently so that each sub-task may be solved by a single worker.In order to decompose tasks,the task requester should have strong professional capabilities.However,for some highly coupled tasks,it is not easy for decomposing.Some researches try to assign complex tasks to workers or teams with multiple skills.The task quality cannot be guaranteed because of insufficient workers' willingness to complete complex tasks by spending much time and energy and different levels of the workers' quality.In order to solving above problems,we propose task assignment methods based on a multi-skilled worker team in complex scenarios.The main research work is as follows:(1)A new method for measuring the quality of complex tasks is proposed,which takes the sum of the effective skills of workers in the team and the team's efficiency of collaboration into quality consideration.On the one hand,it is ensured that the workers in the team have the higher skill quality required to complete the task,on the other hand,the task can be done with high quality by efficient team collaboration.(2)The heuristic algorithm HCT-GFL and the optimization algorithm OCT-GFL are proposed for assigning complex task to a team with leader.The heuristic algorithm select workers according to their availability which can meet the skill quality requirements of the task.Furthermore,the optimization algorithm narrows the search space of workers through pruning strategies,while the selected leader with a larger effective quality can make more contributions for the task.Experimental results show that the optimized OCTGFL algorithm can efficiently organize the team with leader and improve the quality of task completion.(3)The skills bucketing HB-CT-GFO algorithm and overall replacement strategy RS-CT-GFO algorithm are proposed for assigning complex task to a team without leader.The skills bucketing algorithm select workers by the single quality-price ratio and use a partial replacement strategy to control the cost of team communication;while the overall replacement strategy algorithm uses the effective quality-price ratio for worker selection,taking into account all the workers' skills that contribute to the task,and adopt an overall replacement strategy to control the cost of team communication.Experiments show that compared with the existing work,the method proposed in this paper can not only significantly improve the task completion quality of the leaderless team,but also the communication cost of the worker team is well controlled.(4)This paper introduces the incentive mechanism into the complex task assignment problem for the first time,and proposes a complex task assignment method that introduces the incentive mechanism,in which the task requester and the worker team with leader are modeled as buyer-seller auction models to optimize complex task allocation.Experimental results show that a complex task assignment algorithm with an incentive mechanism can improve the degree of workers' participation and acceptance for tasks,thereby improving the task quality.