Multi-Objective Task Assignment in Crowdsourcing

Crowdsourcing coordinates a large group of online workers to perform small tasks published by requesters on crowdsourcing platforms. While task assignment problems within these platforms have been extensively studied, multi-objective task assignment remains an area with limited exploration. Specifically, existing research on multi-objective optimization problems has primarily focused on two objectives in the context of spatial crowdsourcing. In this paper, we address this gap by investigating a constrained three-objective task assignment in a general crowdsourcing environment. Our study establishes three key objectives from the perspective of requestors: maximize the amount of work finished; minimize the total cost; and minimize the variance among requestors’ finished work under the limited capabilities of workers. To solve this multi-objective problem, we first adopt the Particle Swarm Optimization (PSO) algorithm to handle the multi-objective task assignment. Next, inspired by the solution to the knapsack problem and the fair queueing algorithm in network scheduling, we design a heuristic algorithm (HA). Finally, we incorporate the results from HA into the initial swarm of PSO to create a hybrid algorithm, HYB. HYB leverages the strengths of both approaches: PSO’s exploration ability and HA’s directness. We conduct simulations to compare the performance of the three algorithms. Our results demonstrate that HYB generates the best outcomes, HA performs moderately, while PSO lags behind. We conclude that leveraging heuristic insights alongside optimization techniques enhances both PSO and HA, especially given the large combinatorial nature of the multi-objective task assignment problem.