ENE-PPO: A Deep Reinforcement Learning Approach for Task Allocation of Participatory Mobile Crowd Sensing

In recent years, the Mobile Crowdsensing (MCS) develops rapidly, which acts as a key component of Internet of Things (IoT) has attracted much attention. In an MCS system, participants play an important role, since all the data are collected and provided by them. How to select participants who can provide high-quality data to improve global data quality is the core problem. In this article, we propose a participatory MCS task allocation model based on deep reinforcement learning (ENE-PPO), which aims to maximize the platform's profits during the mission time via MCS participation. Specifically, the ENE-PPO consists of two mechanisms, that a feature extractor based on Graph Neural Network (ENE) first extracts the feature of the road network environment of the unmanned vehicles. The feature extractor inputs the nodes and edges of the network in the sensing area into a neural network: ENE, and outputs them after passing through a fully connected layer. Furthermore, we use the Proximal Policy Optimization (PPO) algorithm to learn the optimal strategy. We formally show ENE-PPO has the desirable properties of computational efficiency, dispatch efficiency. We evaluate the proposed scheme via simulation using a real road network. The extensive experiments result well justify the effectiveness of the proposed approach in comparison with the other two methods.