Environment-Driven Task Allocation in Heterogeneous Spatial Crowdsourcing

Zhang, Xuan; Cui, Helei; Yu, Zhiwen; Guo, Bin

doi:10.1109/globecom48099.2022.10000860

Cited by 2 publications

(1 citation statement)

References 18 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Sun et al [13] introduced the EDPG-Assignment framework, amalgamating action embedding and neighborhood deep deterministic policy gradient to intelligently assign tasks for interactive spatial crowdsourcing applications. To address the disregarded heterogeneity of spatial crowdsourcing tasks, [20] proposed the TA-DSAC algorithm solution, training agents within spatio-temporal constraints. Ho et al [21] framed the interruption issue in microtask competition in crowdsourcing as a Markov decision process.…”

Section: Related Workmentioning

confidence: 99%

Optimizing Crowdsourcing Task Assignment Policies Using Multi-Agent Reinforcement Learning in Stochastic Games

Zhang,

He,

Chen

et al. 2023

Preprint

View full text Add to dashboard Cite

Spatio-temporal crowdsourcing as a current widely used model, how to solve the equilibrium between the task planning efficiency and the worker's revenue is a challenging task, in order to overcome these problems, we firstly define crowdsourcing task allocation as a multi-party stochastic game, and solve the equilibrium for the case with perfect recall, which gives the optimal policy for the platform and the workers. Based on the previous work, we propose a framework based on the dual actor-critic approach, which learns the collaborative optimal strategies based on actor-critic. We solve the equilibrium by introducing a Monte Carlo tree search algorithm and use a strategy gradient-based approach to optimize the actor and critic parameters. To evaluate our approach, experiments are conducted in several different crowdsourcing scenarios. The experimental results show that our method is able to achieve good performance in different scenarios and has high accuracy and stability in equilibrium solving. We will open our source code to allow other researchers to reproduce our results.

show abstract

Section: Related Workmentioning

confidence: 99%