Dynamic Contact Force Estimation via Integration of Soft Sensor Based on Fiber Bragg Grating and Series Elastic Actuator

Research on interactive force measurement in robotics follows two trends: distributed force sensing using soft tactile sensors and centered force sensing using rigid sensors. This study proposes a novel force sensing mechanism and algorithm to integrate the two approaches taking advantage of a soft tactile sensor and rigid actuator based on spring. Soft tactile sensors allow for gentle contact with humans, but have limited recovery and measurable force range. The rigidity of a spring-based actuator is utilized to address their force estimation issues. This allows for estimating a wider range of forces while maintaining the softness. The paper presents a novel approach for integrating two sensors using sophisticated algorithms. Specifically, a deep neural network is developed to estimate the contact location through the tactile sensor. Subsequently, a state-space observer is proposed based on the dynamic characteristics of the robot link, which integrates the network output and the torque measurements obtained from a spring-based actuator. This algorithm provides accurate force estimation during dynamic behavior and enables a wide measurable force range across the entire area of the robot link. The efficacy of the proposed mechanism and algorithm is validated through rigorous experimentation, demonstrating the fast recovery characteristics and accuracy.