AI-Driven Haptic Cues for Haptic Object Reproduction

The potential of haptic technology is enormous, and haptics plays a crucial role in various fields. Recent studies on haptics have primarily focused on granting robots a sense of touch to enhance their performance in manipulation tasks and unlock barriers in the virtual world. Most haptic studies have relied on model-based approaches to interpret haptic interactions, and conventional spring-damper systems are often used. However, these methods have limitations in accurately reflecting real, nonlinear haptic interactions. Although force sensors have been widely used in these studies, they have drawbacks such as limited bandwidth, susceptibility to signal noise, complexity, non-collocation, and instability. Interestingly, force measurements based on the sensorless force control mechanism using DOB (Disturbance Observer) and RFOB (Reaction Force Observer) have proven to provide more accurate results than traditional force sensors. Haptic studies have often relied on motion parameters to detect haptic feedback, potentially failing to capture the full range of factors that can influence haptic sensation. In recent years, AI (Artificial Intelligence) approaches have demonstrated superior results in many areas. However, it is crucial to clearly understand the dataset to achieve more reliable results through AI. Therefore, this study introduces an AI-based approach to recreate vivid force sensations in a virtual model, replicating the actual environment using information abstracted through DOB and RFOB-based sensorless force control mechanism. The validity of this approach is discussed by comparing the haptic sensations produced by the AI-based model with those produced by both the conventional object model and the actual object.