Content caching is a promising solution to overcome the backhaul traffic delay issue by caching content at the base station (BS). However, the performance of content caching is restricted by the limited BS cache storage. In this paper, we are the first to employ artificial intelligence generated content (AIGC) to enhance the content caching performance by empowering the BS's intelligence capability. Besides caching a popular file, our AIGC-enhanced BS can alternatively cache its prompt of smaller size to reconstruct the file whenever needed by mobile terminals. We reveal the fundamental tradeoff between caching storage saving and computation delay for AIGC to decide whether to cache a file or its prompt. In this regard, this paper investigates the new problem of joint hybrid caching, computation and communication resource allocation optimization to minimize the average system latency to serve mobile terminals. As the problem is a non-convex and involves mixed integer non-linear programming, we develop a divide-and-conquer algorithm to decompose the problem into two timescale levels. We theoretically prove that our AIGC-enhanced hybrid content caching outperforms the conventional content caching once the computation capacity is non-trivial. Extensive simulation results also validate the superiority of the proposed hybrid content caching.

As a promising paradigm to support wireless connections among massive devices, the Internet of things (IoT) is facing several challenges. Among them, the energy supply to IoT devices, the weak computation ability of IoT devices, and the stringent latency requirement of IoT applications are three major challenges, which can be tackled by radio frequency energy harvesting, mobile edge computing (MEC), and short packet communications (SPC), respectively. In this paper, we investigate resource allocation in MEC-enabled wireless powered IoT (WP-IoT) networks with SPC, and formulate the problem of optimizing the computation frequency, the packet length and the packet error rate, targeting maximizing the sum effective computation throughput. Since the problem is difficult to solve in general, we first simplify the problem by analyzing its properties, then design an efficient algorithm to obtain a suboptimal solution in an iterative manner based on the bisection method, the block coordinate descent method, the successive convex approximation method, and the majorization-minimization method. Simulation results confirm the effectiveness of the proposed algorithm. Particularly, it is shown that the proposed algorithm is of low complexity and achieves the performance close to that of the optimal exhaustive search, while also significantly outperforms other benchmark algorithms in existing literature.