Yuan He

Processing-in-Memory (PIM) is emerging as a practical path to overcome the limitations of traditional von Neumann architectures. At its core, PIM systems implement computing primitives such as logic operations and multiply-accumulate acceleration through compute-in-memory, near-memory processing, or hybrid designs. The role of memory cells varies widely across technologies, acting as inputs, outputs, or analog accumulators through bit-lines and sense amplifiers. This diversity creates trade-offs in precision, bandwidth, latency, and programmability, making it difficult to build a unified understanding on the progress of the field. In this survey, we organize recent advances of PIM into three areas. First, we discuss the progress on the architectural optimizations of PIM and its integration with both DRAM and emerging non-volatile memories. Second, we examine how PIM is being used to accelerate key computing domains, including generative AI workloads and high-performance kernels, along with new approaches. Third, we highlight the growing adoption of PIM in computational sciences, where it is being applied to solve interdisciplinary problems such as genome analysis, mRNA quantification, mass spectrometry, quantum circuit simulation, wave modeling, and secure computation. Finally, we synthesize the major challenges that continue to slow PIM adoption, including manufacturing constraints, power delivery, thermal reliability, data consistency, runtime and memory-management coordination, and the difficulty of building portable software abstractions without sacrificing commercial viability. This work provides an updated, structured perspective on PIM’s potential across computing and computational sciences and the barriers that must be solved for it to reach its full impact.

Extended depth-of-field (DoF) technology is widely used in three-dimensional (3D) depth measurement. However, it is hard to achieve extended DoF imaging for conventional stereoscopic vision systems in dynamic scenes. Here, we propose an extended DoF vision system based on variable-focus lenses for target motion scenes. This system operates at two synergistic levels: optical-level focal sweeping and computational-level all-in-focus synthesis dynamically. Specifically, we introduce a temporally adaptive focal calibration model that precisely predicts curvature variations of varifocal lenses under non-linear electromechanical responses, ensuring sub-millisecond focal accuracy across a depth range of 0.45∼1.2 m. In addition, an adaptive homography-based multi-plane remapping mechanism is designed to compensate for disparity drift and distortion artifacts induced by rapid focus change, followed by a sharpness-weighted fusion strategy that synthesizes an all-in-focus image with preserved high-frequency details. The depth measurement performance was improved by approximately 6.2%, the relative depth error rate is optimized to 5.82%, and the RMSE is reduced to 0.0505 compared to conventional stereo vision systems in images obtained. The proposed system has the advantage of a compact form factor and dynamic extended DoF imaging. The potential applications include biomedical imaging, industrial vision inspection, robotic vision with navigation, and AR/VR.

Head-mounted displays (HMDs) require precise measurement of virtual image distance for user comfort, but this is challenging due to dynamic variations. This paper addresses the difficulty by proposing a prototype using a variable-focus liquid lens and a calculation model for virtual image distance. We developed an experimental platform to validate the method and introduced an optimization algorithm to find the optimal focal length for maximum sharpness. Results showed a distance error of about 5 cm, confirming that our method accurately measures virtual image distance in HMDs, with potential applications in virtual and augmented reality.

Vehicle-to-grid (V2G) networks, as an emerging smart grid paradigm, can be integrated with renewable energy resources to provide power services and manage electricity demands. When accessing electricity services, an electric vehicle $\left ({{\mathcal {E}\mathcal {V}}}\right)$ typically provides authentication or/and payment information containing identifying data to a service provider, which raises privacy concerns as malicious entities might trace $\mathcal {E}\mathcal {V}$ activity or exploit personal information. Although numerous anonymous authentication and payment schemes have been presented for V2G networks, no such privacy-preserving scheme supports authentication and payment simultaneously. Therefore, this paper is the first to present a privacy-preserving authentication scheme with anonymous payment for V2G networks (PAP, for short). In addition, this scheme also supports accountability and revocability, which are practical features to prevent malicious behavior; minimal attribute disclosure, which maximizes the privacy of $\mathcal {E}\mathcal {V}$ when responding to the service provider’s flexible access policies; payment binding, which guarantees the accountability in the payment phase; user-controlled linkability, which enables $\mathcal {E}\mathcal {V}$ to decide whether different authentication sessions are linkable for continuous services. On the performance side, we implement PAP with the pairing cryptography library, then evaluate it on different hardware platforms, showing that it is essential for V2G applications.

A large open aperture in an optical system can capture high-resolution images but yields a shallow depth of field. To overcome this issue, we propose a method for retrofitting microscopy imaging systems by using a variable-focus liquid lens to achieve 3D focus scanning. Specifically, the focal length of the imaging system was changed by controlling the liquid lens, and a sequence of images was captured at different focal planes in milliseconds. The image scale and phase were corrected. Then the in-focus pixels were abstracted by employing the Laplacian operator. These pixels were marked in the index table. According to the index table, the focused parts of multiple images were merged, and an all-in-focus image was generated. At the same time, a depth map was constructed based on the image number in the index table and the extracted depth information. Additionally, we have optimized the image processing flow; the processing speed was improved to around 6.5 fps.