Authenticated Robotic Teleoperation with Task Recognition

Human-in-the-loop teleoperation of robotic arms holds considerable promise for consumer electronics, particularly in immersive and interactive applications such as healthcare, gaming, augmented reality, virtual reality, smart home systems, and telemedicine. These applications demand robust user authentication to safeguard against unauthorized access, ensuring secure, user-centered control of teleoperated systems in various consumer environments. However, existing systems often lack seamless integration of security measures, such as biometric authentication, while maintaining affordability and responsiveness. We propose a low-cost teleoperation system using wearable inertial measurement units (IMUs) to securely and responsively control a robotic arm. We employ a machine-learning approach for authentication using logistic regression on time series data from the IMUs during handling tasks. Experiments with 16 operators performing three handling tasks demonstrate that random forest outperforms other classifiers in task identification, achieving a macro F1-score of 75.60%. In contrast, logistic regression performs best in user identification and authentication tasks. Our system achieves an average Equal Error Rate of approximately 8.89% in user authentication using logistic regression. The proposed system's low-cost, IMU-based design, adaptable to various end-effectors, aligns with consumer demands for affordable, intuitive, and secure teleoperation setups. This work highlights the potential for biometric-based teleoperation to advance consumer technology applications in healthcare, Internet of Things, and immersive environments, ensuring personalized and secure user experiences.