Time-Constrained Task Allocation and Worker Routing in Mobile Crowd-Sensing Using a Decomposition Technique and Deep Q-Learning

Mobile crowd-sensing (MCS) is a data collection paradigm, which recruits mobile users with smart devices to perform sensing tasks on city-wide scale. In MCS, key challenge task allocation, especially when MCS applications are time-sensitive, and the platform needs consider completion order (since worker may multiple different orders lead travel costs response times, i.e., times needed arrive at venues), requirements of (such as deadline required sensor) workers heterogeneity. other words, allocation problem consists problems, challenging solve due large solution space. Therefore, in this paper, we first formulate considered into two related integer linear programming problems (i.e., assignment problems) using decomposition technique reduce size enable use diverse searching strategies. Then, deep Q-learning (DQN)-based algorithm, namely DQN local search (A-DQN w/ LS), proposed determine task-worker assignments, iteratively employs an asymmetric traveling salesman (ATSP) heuristic find workers. The optimizer applied end A-DQN algorithm deal computation time optima. Simulation results show that method outperforms existing approaches under dynamics terms total cost.