Shengxi Zhou

Biological locomotion often benefits from flex‐rigid architectures, where soft tissues coordinate with rigid skeletons or appendages to enable effective force transmission, versatile motion generation, and environmental adaptability. However, translating this paradigm to robotics remains challenging. Here, we present a bio‐inspired flex‐rigid robot (Bio‐FRR) that integrates a muscle‐like pneumatic actuating unit with a fin‐like rigid arched plate to achieve high‐performance multimodal locomotion. Unlike conventional rigid‐flexible hybrid robots in which rigid elements mainly serve as passive reinforcements or fixed appendages, the Bio‐FRR employs a deployable flex‐rigid mechanism, where a multi‐chamber pneumatic actuator actively drives rigid arched plates to fold and unfold, thereby converting local soft deformation into body‐level posture reconfiguration and locomotion‐relevant mechanical output. Such a design endows the Bio‐FRR with versatile and adaptive locomotion across various environments, realizing five high‐performance locomotion modes on land and in water without structural reconfiguration, including upright hopping, leapfrogging, terrestrial crawling, diving, and swimming. The Bio‐FRR reaches a maximum terrestrial speed of 1.23 body lengths per second (BL/s), a horizontal swimming speed of 0.42 BL/s, and an untethered driving speed of 48.67 mm/s. Bio‐FRR offers a robust blueprint for next‐generation amphibious soft robots, enabling versatile locomotion and enhanced adaptability.

This paper proposes a Y-shaped bifurcated beam magnetic self-coupled piezoelectric energy harvester (Y-MSPEH). By incorporating built-in permanent magnets, the harvester is capable of nonlinear stiffness adjustment and broadband energy harvesting, which in turn helps reduce its physical footprint. To this end, a dynamic model of the Y-MSPEH is developed based on Hamilton’s principle, with geometric nonlinearity and piezoelectric coupling effects taken into account. The study derives numerical solutions for the system response through theoretical analysis and validates them via experiments. For the Y-MSPEH, key parameters, including the magnet’s terminal length, branch angle, and magnetic field strength, are analyzed under both repulsive and attractive magnetically self-coupled states. The results show that adjusting the parameters of the Y-MSPEH can effectively tune the peak frequency of the output voltage response and broaden the response frequency band. Comparative studies indicate that the attractive and repulsive magnetically self-coupled states result in wider response bandwidth and higher response amplitude, respectively. Specifically, compared with the linear Y-shaped bifurcated beam harvester and the geometrically nonlinear Y-shaped bifurcated beam harvester, the Y-MSPEH achieves an approximately 8.26% increase in effective bandwidth in the magnetically self-coupled attractive state, thus demonstrating superior energy harvesting performance.

Piezoelectric energy harvesting (PEH) interface circuit serves as the core power supply solution for self-powered sensing devices operating under harsh conditions. However, existing technologies are confronted with prominent performance bottlenecks: the conventional synchronized electric charge extraction (SECE) circuit has limited potential for improving output power, while the parallel synchronized switch harvesting on inductor (P-SSHI) circuit suffers from poor load impedance adaptability. Neither of them can simultaneously achieve the dual performance goals of high output power and strong load impedance adaptability. To address the aforementioned technical issues, this paper proposes a high-efficiency self-powered synchronized charge extraction and inversion (ESP-SCEI) interface circuit for PEH. The circuit employs an optimally designed peak detection switch to effectively reduce energy loss, and innovatively introduces a voltage inversion mechanism analogous to that of the P-SSHI topology during the energy extraction phase of the SECE circuit. Unlike the conventional SECE circuit, which fully extracts the internal charge of the piezoelectric transducer (PZT) in a single energy extraction operation, the ESP-SCEI circuit only extracts a portion of the energy stored in the PZT during each operating cycle, with the remaining charge participating in the voltage inversion process. Notably, energy extraction and charge inversion are realized synchronously in this process. This design not only elevates the peak output voltage of the PZT but also achieves effective decoupling between the PZT and the load. Experimental results demonstrate that the ESP-SCEI circuit simultaneously exhibits the high peak output power characteristic of the P-SSHI circuit and the strong load independence characteristic of the SECE circuit. Furthermore, under the conditions of an open-circuit voltage ( V _oc ) of 8 V and a resonant frequency of 23.4 Hz, its output power reaches 1.96 times as much as that of the SECE circuit, placing it at the forefront of state-of-the-art rectifier technologies.

With the widespread adoption of sensors, research into vibration control (VC), vibration energy harvesting (VEH), and structural health monitoring (SHM) has garnered growing attention. However, conventional methods in these fields struggle with high-dimensional, multi-source data processing, while machine learning (ML) has emerged as a promising solution for handling nonlinear, complex vibration-related data and enabling real-time data analysis. Existing reviews typically focus on ML applications in a single domain, lacking a cross-field synthesis. This review comprehensively summarizes the latest ML advances in VC, VEH, and SHM: for VC, it addresses the challenge of adapting to dynamic operational conditions and outlines prospects of adaptive control via reinforcement learning; for VEH, it targets unstable energy conversion under random ambient vibrations and highlights optimized harvesting strategies using deep learning; for SHM, it tackles micro-damage feature masking by environmental interference and envisions improved detection via computer vision models. It further analyzes domain-specific ML challenges and emerging research directions, providing readers with a unified cross-field framework, clear technical bottleneck insights, and actionable guidance for advancing intelligent vibration-related engineering systems.

The increasing demand for sustainable energy solutions necessitates replacing the traditional energy sources with renewable energy, particularly for realizing self-powered wireless sensor networks (WSNs) in railway transportation. This paper proposes a double-sided rack rail vibration energy harvesting system (DVHS) to harvest vibration energy generated by trains when they pass over the rail, providing electricity to the WSNs. The input mechanism, mechanical rectifier, power generation module, and adaptive interface circuit are designed for the proposed DVHS. The mechanical rectifier, composed of a double-sided rack and one-way bearing, is employed to change the railway vibration into rotation. Rotational energy is subsequently transmitted to the power generation module via a mechanical transmission device, converting the mechanical energy into electrical energy. Experimental results show that the DVHS generates an average power output of 854 mW with the vibration amplitude of 5 mm at 5 Hz. An adaptive interface circuit with the maximum energy conversion efficiency of 93.5% is designed to efficiently extract energy from the vibration energy harvester and provide a stable output voltage. A 1.5 F supercapacitor can be fully charged by the DVHS within only 162 s, after which it powers the wireless sensing network for up to 300 s of continuous data transfer. The proposed DVHS presents a promising solution for sustainable condition monitoring in railway systems, reducing dependence on conventional power supplies.