Energy-efficient task allocation in mobile crowdsensing using mobility and participation prediction

Abstract This paper presents a two-phase framework for task allocation in mobile crowdsensing (MCS), consisting of offline training and task allocation phases. In the offline training phase, user mobility patterns and participation rates in sensing blocks are trained and predicted. In the task allocation phase, tasks are assigned to users based on their predicted future locations and participation rates in the blocks. The task allocation is performed using Binary Enhanced Whale Optimization (BE-WOA), U-shaped Binary Particle Swarm Optimization (UBPSO), and Ant Colony System Optimization (ACS) to achieve the dual objectives of maximizing data quality and minimizing users' energy consumption. Two real-world datasets were utilized in the offline training phase to predict users' future movement paths, achieving an average MAE of 0.004579 and 0.00699, while the prediction of participation rates in blocks yielded an MAE of 0.2465. In the task allocation phase, 11 distinct task allocation scenarios comprising 110 simulations were conducted, resulting in an average 70% energy savings for users and maximum data quality. Additionally, the proposed approach demonstrated a 37.6% improvement in coverage rate compared to prior research. Based on the cost function, BE-WOA outperformed the random and greedy algorithms by 124% and 65%, respectively, and was identified as the best task allocation algorithm in this study. This work provides valuable insights into IoT-driven data collection systems, offering an effective and energy-efficient solution to maintain high data quality and user engagement with minimal resource consumption.