Traditional rigid robots have been widely used in various industries to replace humans in performing simple tasks. However, the inherent rigidity of rigid robots limits their ability to adapt to the external environment. In contrast, soft robots, which contain almost no rigid materials, are flexible, not only being free from environmental constraints but also capable of safe human interaction [1]. With the continuous development of soft materials [2], manufacturing technology [3] and computer control systems, soft robots have shown their ability to change our daily lives [4]. For example, soft robot mechanisms combined with traditional rigid robots can achieve object manipulation tasks in industrial environments, which is significant for handling fragile objects [[5], [6], [7]]. Compared with rigid microrobots, soft microrobots have strong adaptability, low invasiveness, and are more friendly to biological tissues [8]. Therefore, they have broad application prospects in the biomedical field [9,10]. However, the dynamic modeling of soft robots is much more complicated than that of rigid robots, which poses a huge challenge to soft robots' shape and position control. Sensing capabilities are crucial to the precise control and intelligence of soft robots. To better simulate biological systems, flexible sensors need to be integrated into soft robot systems. Although significant progress has been made, flexible sensing technology is still in its infancy [11].

Through simple open-loop control, soft robots can achieve a variety of complex movements, but it is difficult to achieve high-precision operation and error correction by open-loop control alone. Therefore, it is crucial to establish a closed-loop control system in combination with sensors to develop soft robots. In addition, in order for soft robots to perform tasks autonomously, their proprioceptive perception and the ability to detect external stimuli are essential. However, the characteristics of soft robots, such as multiple degrees of freedom, high mechanical compliance, and strong deformation ability, limit the application of many traditional sensors, including encoders, metal or semiconductor strain gauges, or inertial measurement units [12,13]. Moreover, traditional sensors are difficult to use for multimodal sensing, which further limits the multimodal perception and interaction capabilities of soft robots.

The complexity of soft robots lies in their ability to withstand continuous and multi-degree-of-freedom deformations, which poses unique challenges to sensor design. Traditional rigid sensors are not sufficient to capture the various movements of soft robots, so there is a need to develop flexible sensors that can adapt to and accurately measure these deformations. The design of such sensors must not only consider the mechanical characteristics of the robot, but also meet the needs of accurate and reliable sensing in dynamic and unstructured environments. This has prompted the exploration of innovative surface and interface structures to enhance the sensor's ability to detect subtle changes in shape and force while maintaining the flexibility and adaptability required for soft robotic applications [14]. Thanks to the development of flexible electronic technology, the progress of artificial intelligence and machine learning technologies has provided important support for the optimization and improvement of soft robot interaction algorithms, enabling soft robots to adapt to complex environments more intelligently, perceive environmental changes, and achieve more flexible and efficient interactions [[15], [16], [17], [18], [19], [20]]. Since flexible sensors are highly flexible [21] and can adapt to the surfaces of different objects, they can improve the interaction with the target system and improve the reliability and stability of detection [22]. Therefore, they also have broad application prospects in wearable smart devices [[23], [24], [25], [26], [27]], human motion monitoring [[28], [29], [30], [31]], and portable medical diagnostic equipment [[32], [33], [34]].

A large number of flexible sensors have been developed for the application of soft robots in different scenarios. Based on their basic sensing principles, flexible sensors can be divided into resistive, capacitive, piezoelectric and triboelectric types. Resistive sensors detect by converting the measured value into a change in the resistance value associated with it. Although they have a slow response speed and low sensitivity, they are easy to manufacture, convenient to integrate and low cost [35]. Capacitive sensors rely on the change in capacitance caused by external stimuli for detection [36]. They have the characteristics of fast response speed, high sensitivity and low power consumption [37]. However, under large deformation conditions, capacitive sensors are susceptible to highly nonlinear strain measurement, and surrounding conductive objects and environmental interference may also reduce their stability. Piezoelectric sensors generate electric dipole moments and convert them into electrical signals through the deformation of piezoelectric crystals under pressure [38]. They have the characteristics of fast response speed, moderate sensitivity and can work without an external power supply, but their output signals are easily affected by temperature changes, show high response drift over time, and produce electrostatic interactions with the surrounding environment [39]. Triboelectric sensors use the electrostatic charge generated by surface friction to convert signals [40]. They are characterized by fast response speed, high sensitivity and no need for external power supply, but they have high requirements for precision manufacturing processes.

In order to meet the needs of soft robots for obtaining precise information (such as curvature, strain, elongation, contact force, bending angle and proximity, etc.), the key performance of flexible sensors is also constantly improving. These key performances include sensitivity, linear sensing range, response time, resolution, hysteresis and reliability, etc. [[41], [42], [43]]. Sensitivity reflects the sensor's ability to respond to input changes [44]. The linear sensing range ensures the sensor has high accuracy and resolution within the measurement range [45]. Response time is defined as the time required for the sensor to reach a stable output after being stimulated [46], which is crucial for high-frequency signal detection. Resolution refers to the minimum change in the measured value that the sensor can detect. Hysteresis refers to the non-overlapping degree of the input-output characteristic curve of the sensor during the process of increasing and decreasing input volume [47]. In practical applications, the hysteresis phenomenon needs to be reduced as much as possible. Reliability is a measure of the sensor's ability to stably complete tasks under specified conditions and within a specified time [48], and is an important indicator that cannot be ignored in the long-term operation of the sensor.

To further improve the performance of flexible sensors, a common strategy is to design and optimize surfaces and interfaces. This is also the focus of this article. In flexible sensors, surfaces and interfaces play a crucial role in the design and construction of sensors. The interactions between the internal structures of the sensor and the interactions between the sensor and the external environment are both manifestations of surfaces and interfaces in the sensor. The morphology and geometric design of the sensing layer have a significant impact on the performance of flexible sensors, and the introduction of special structures is of great significance for the manufacture of high-performance sensors [49,50]. From simple flexible supports to active functional layers, they can be used to facilitate the transmission of various mechanical and electrical inputs [51]. The morphology design and optimization of sensing materials [52] is an important means of surface and interface construction, so the construction of surface and interface [53] can effectively improve the performance of flexible sensors and make them better adaptable to various environments and various needs.

For flexible sensors, their main functions can be divided into two categories: proprioception and external stimulus perception. Proprioception mainly involves bending deformation, torsion deformation and tensile deformation; while external stimulus perception is mainly achieved through compression deformation (As shown in Fig. 1). According to different working principles, common flexible sensors include resistive sensors, capacitive sensors, piezoelectric sensors, triboelectric sensors and other types of sensors. Among these sensors, capacitive sensors are relatively easy to use for external stimulus detection. Although resistive sensors and piezoelectric sensors can also be used for external stimulus detection, they also can detect proprioceptive deformation. Triboelectric sensors can detect external stimuli and sense proprioceptive deformation, but most of the current research is still focused on the perception of external stimuli. Therefore, this article focuses on three common strategies for constructing surfaces and interfaces of flexible sensors, including surface structure optimization, microstructure array design and void structure. These strategies are key factors in improving the performance of flexible sensors. Secondly, the preparation methods of different surfaces and interfaces are discussed in detail. Whether it is the template method, thermal wrinkling method, or technology based on chemical reaction or electrospinning, various preparation methods provide technical guarantees for the mass production and performance optimization of flexible sensors. Based on the above-mentioned surface and interface morphology optimization, this paper further introduces the application of flexible sensors in extreme environments such as human-computer interaction, deep sea and deep space, demonstrating the great potential of flexible sensors in complex environments. Finally, the future development direction of flexible sensors is discussed, aiming to inspire the development and innovation of related fields.