Vision-based tactile sensing: From performance parameters to device design

Abstract

By integrating the virtues of vision and touch, vision-based tactile sensors (VBTSs) achieve an artificial tactile capability that transcends the natural, demonstrating superior performance unattainable through either sense alone. VBTS, as an innovative sensor, boasts commendable performance metrics and has found extensive applications across various domains. Nevertheless, a comprehensive synthesis regarding the perceptual performance achievable by VBTS is currently lacking. Moreover, the performance parameter evaluation systems for VBTS are not standardized, and the strategies for enhancing these metrics remain unclear. Here, the significant advancements in VBTS over recent years are summarized, from sensing mechanisms to application scenarios. The review particularly focuses on parameters that assess performance and novel strategies in hardware design aimed at improving these performance parameters, including key performance indicators (e.g., range, spatial resolution, and sensitivity), along with error, temporal parameters, miniaturization, and stability. This review also discusses the sensing capability and application scenarios, such as item identification, grasp control, material property detection, and multimodal perception. Finally, perspectives on VBTS are provided. We expect that this review enables researchers to rapidly comprehend the capabilities and performance of VBTS, and offers references for the selection or design of its different modules.