The increasing demand for advanced human-machine interfaces (HMIs) in robotics and virtual reality (VR) necessitates cost-effective and efficient solutions for parallel control of multi-dimensional human motions. Current exoskeleton systems, while offering assistive capabilities, often involve high costs and complex data processing. This paper introduces a novel approach using triboelectric bidirectional sensors, a low-power, cost-effective alternative, integrated within a customized exoskeleton to monitor the multiple DOFs of human upper limbs. The sensors track movements such as shoulder rotations, wrist twisting, and finger bending, enabling real-time control of virtual characters and robotic arms. The inherent structural consistency between the exoskeleton and human body allows for kinetic analysis of the sensor data to extract additional physical parameters, reducing the need for additional sensors. This innovative HMI design holds great potential to improve cost efficiency and enhance the capabilities of HMIs across diverse fields, including robotic automation, healthcare, and training.

Existing HMIs utilize various sensing technologies, including inertial sensors (high sensitivity but complex data processing), resistive sensors (strain and force sensing but temperature-dependent), and capacitive sensors. Glove-based HMIs, incorporating inertial and resistive sensors, are common but suffer from drawbacks such as power consumption and complex signal processing. Flexible wearable sensors and e-skins offer promising alternatives, with examples including capacitive tactile finger sensors, ultra-sensitive resistive strain gauges, and multifunctional sensor arrays. Triboelectric and piezoelectric sensors are gaining traction due to their low power consumption, particularly for large-scale sensory networks. While triboelectric sensors offer advantages in customization and multifunctionality, many existing bidirectional triboelectric rotation sensors are complex, using two-disk designs with multiple grating patterns and channels. These increase size and programming complexity. Current exoskeleton research mainly focuses on actuation, while few studies explore new sensors for HMI applications within exoskeletons. The proposed system addresses these limitations by utilizing a novel triboelectric sensor design and leveraging the exoskeleton's structural advantages for improved motion tracking and data interpretation.

This research involves the design, fabrication, and characterization of a novel triboelectric bidirectional (TBD) sensor and its integration into a 3D-printed exoskeleton. **Sensor Design:** Two types of TBD sensors are designed: rotational TBD (RTBD) sensors for detecting rotational movements and linear TBD (LTBD) sensors for linear movements. The RTBD sensor comprises a shaft, a fly ring with a PTFE grating pattern, and a bistable switch with two electrodes. The LTBD sensor consists of a holder, a switch, and a flexible FEP strip with a PTFE grating pattern. The sensors operate using the triboelectric sliding sensing mode, generating electric outputs proportional to movement. The design emphasizes simplicity, using a single grating pattern to reduce complexity compared to existing bidirectional sensors. **Exoskeleton Design:** A 3D-printed exoskeleton arm, comprising five components (back supporter, shoulder module, upper arm, forearm, and glove), is designed to integrate the TBD sensors for detecting multi-DOF motions. The RTBD sensors are strategically placed on the shoulder, upper arm, forearm, and wrist for comprehensive upper limb motion tracking. The LTBD sensor is incorporated into the glove for finger bending detection. **Sensor Characterization and Optimization:** The RTBD sensor's performance is optimized by varying the width and spacing of the PTFE grating pattern, and the effect of rotation speed on the signal intensity and angular resolution is investigated. The LTBD sensor is characterized by varying spacing during finger bending. Response latency of the switch for bidirectional sensing is assessed by measuring the time required for switching between directions. **Signal Processing:** A pre-processing circuit using an operational amplifier and comparator is designed to convert the triboelectric signals into square waveforms for easier peak counting and data interpretation, streamlining the signal processing for real-time control. The Arduino MEGA 2560 microprocessor is utilized for data acquisition and signal processing. **Software and Control:** The Arduino code reads signals, and the Python code performs signal processing and data visualization. Unity software manages virtual character and robotic arm control. **Demonstrations:** The system's performance is demonstrated through three key applications: controlling a virtual character in virtual space, manipulating robotic arms, and simulating a ping-pong game to demonstrate the system's capacity to accurately track and interpret complex human motions. **Kinetic Analysis:** The research also investigates the use of kinetic analysis to extract additional parameters from the sensor data. The paper illustrates how the time interval between pulses from the RTBD sensors can be used to calculate rotational velocity and further how velocity and contact time can be used to estimate force during a punching motion, demonstrating the system's ability to go beyond basic angle measurement. This analysis relies on the geometrical relationships inherent in the exoskeleton's structure and the human arm's biomechanics.

The key findings of this study include: 1. **Successful Development of a Novel TBD Sensor:** A novel bidirectional triboelectric sensor design was developed, providing a low-cost, low-power alternative to existing motion-sensing technologies. The sensor's simplicity, using a single grating pattern, reduced fabrication complexity and the need for intricate signal processing. The system was demonstrated to successfully track multiple degrees of freedom, including shoulder rotations, wrist twisting, and finger bending, in real-time. 2. **Effective Exoskeleton Integration:** The TBD sensor was successfully integrated into a 3D-printed exoskeleton, providing a comfortable and adaptable platform for motion tracking. The exoskeleton's design allowed for precise placement of sensors to capture the complex movements of the upper limb. 3. **High-Speed Performance:** The RTBD sensor showed reliable performance across a broad range of rotational speeds (10-300 RPM), demonstrating its suitability for applications requiring various levels of motion detection and resolution. 4. **Successful Real-time Control:** The exoskeleton system effectively controlled virtual characters in a virtual space and robotic arms, achieving real-time manipulation. The system used data from different sensors to make complex movements, like lifting boxes and playing ping-pong, demonstrating the intuitive nature of the interface. 5. **Force Estimation via Kinetic Analysis:** The study demonstrated the potential of using kinetic analysis of the sensor data to extract additional physical parameters, including velocity and force, without adding extra sensors. The researchers used this approach to estimate the force of a punch, showing its potential to be used in applications needing detailed motion analysis. The estimated force showed good agreement with measured values, further validating this approach.

This research successfully demonstrated a novel low-cost, low-power, and highly customizable HMI system for robotic and virtual applications. The use of triboelectric sensors and the integration into an exoskeleton overcome the limitations of existing HMIs, offering a more user-friendly and practical solution. The ability to estimate force and other kinematic parameters through kinetic analysis significantly enhances the system's capabilities, potentially eliminating the need for additional sensors and reducing complexity. The results demonstrate the potential for this technology to be broadly applied in various fields, including robotic automation, healthcare rehabilitation, and virtual/augmented reality training applications. The simplicity of the design and its low power consumption make it a particularly attractive solution for wearable applications that require long-term usage. Future research could focus on further miniaturizing the sensors using advanced fabrication techniques, such as MEMS, to increase resolution. Exploring applications beyond the upper limbs, such as lower-limb exoskeletons, presents promising future directions. Integrating advanced machine learning algorithms could also enhance the system's ability to recognize complex movements and provide more sophisticated feedback.

This paper presented a novel low-cost, low-power exoskeleton HMI system using bidirectional triboelectric sensors. The system successfully tracked multi-DOF upper limb motions and controlled virtual characters and robotic arms. Kinetic analysis enhanced its capabilities by enabling the estimation of velocity and force. This approach offers a promising cost-effective solution for various applications. Future research should focus on miniaturization and integration with advanced algorithms.

While the study demonstrates the potential of this technology, several limitations should be considered. The accuracy of force estimation relies on the accuracy of the kinematic model and may be affected by individual variations in arm dimensions and movement patterns. The current design primarily focuses on upper limb motion, and extending it to the entire body requires further development. The robustness of the sensor under extreme environmental conditions (e.g., high humidity, temperature) needs further evaluation. Finally, although a reliability test of three hours was performed, further long-term studies are needed to fully assess the durability and reliability of the system over extended periods of use.