Joint Optimization for Mobile Crowdsensing Systems with Reliability Consideration

Mobile crowdsensing (MCS) uses sensor-embedded mobile devices to collect and share data. However, the unstable wireless channels and limited network resources deteriorate the reliability of data transmission in MCS networks. To tackle this issue, we propose a hybrid automatic repeat request (HARQ)aided MCS framework in this paper. Within the framework, we formulate a reward maximization problem by jointly optimizing user selection, sensing data allocation, transmit power, and rate allocation, taking into account various practical constraints and outage requirement. To address the non-convexity of the formulated problem, we begin by transforming it into a long term average throughput (LTAT) maximization problem per user, assuming that all users would have the opportunity to participate in the sensing task. Nevertheless, solving the transformed problem remains challenging due to the intricate nature of the outage probability, which arises from the inherent coupling between transmission power and rate variables. To facilitate the analysis, a divide-then-conquer method is proposed. Specifically, we decompose the LTAT optimization problem into two subproblems: transmission power allocation and rate selection. To tackle the transmission power allocation subproblem, we reveal that the outage probability is convex with respect to transmission power. This inspires us to solve the non-convex power allocation subproblem by using an alternating iteration algorithm in conjunction with Dinkelbach’s algorithm. Similar approach can be applied to rate selection. After resolving the two subproblems, an alternating optimization method is used to optimize transmission power and rate iteratively until convergence is achieved. At last, the user selection process is determined by selecting the users with the highest-ranked achievable rewards. Extensive simulation results verify the superiority of our proposed algorithm compared to various benchmark schemes.