Age-of-Information and Latency Minimization for Cache-enabled Mobile Crowdsensing Networks

Mobile crowdsensing (MCS) enables large-scale data collection but suffers from excessive resource consumption due to repetitive sensing operations. By introducing edge caching capabilities, MCS systems can significantly reduce redundant data collection through caching re-use. However, this introduces a fundamental challenge: how to balance sensing cost efficiency and data freshness in cache-enabled MCS systems, where stale cached data may compromise application performance while frequent re-sensing wastes resources. To address this challenge, we investigate the balance between re-using cached task results and re-sensing tasks in MCS networks, aiming to minimize the system latency while ensuring the freshness of cached task information. To this end, we formulate a weighted sum of latency and age of information (AoI) minimization problem, aiming to propose a comprehensive framework that jointly optimizes sensing decisions, user selection, data size and subchannel allocation, along with a caching strategy. To tackle this non-convex and mixed-integer programming problem, we first decompose the original long-term optimization problem into multiple one-shot sub-problems. For each one-shot problem, we reformulate it as a system latency minimization sub-problem to determine the optimal solution under the assumption of task re-sensing. The system latency is then compared with the AoI value to determine task sensing decisions. Subsequently, we develop a Bayesian caching algorithm to optimize cache utilization. Building upon these, a time-efficient algorithm is proposed for the long-term optimization problem considering temporal correlations. Rigorous theoretical analysis reveals that our proposed solution achieves a bounded optimality gap compared to the offline optimal solution. Furthermore, extensive simulation results validate that our method significantly outperforms various baseline schemes.