- The paper introduces a model-based diagnostic framework that employs torsional, joint dynamics, and electrical motor constraints to isolate sensor faults.
- Simulations show the method effectively detects discrepancies like torque sensor faults beyond typical noise levels.
- The study advances robotic safety by enabling precise, model-driven sensor fault detection, enhancing collaborative operations.
Introduction
Robotics plays an increasingly pivotal role in various industrial applications such as manufacturing, assembly, and packaging. As physical barriers between humans and robots are removed, collaborative environments are established, leading to a greater need for ensuring the safety and operational integrity of robotic systems. The reliable functionality of robotics largely depends on precise feedback from an array of sensors. The failure of these sensors, responsible for monitoring position, torque, or temperature, can cause severe disruptions and even dangerous situations, particularly when humans work closely with robots.
Sensor Diagnostics Importance
Model-based sensor diagnostics serves as an essential approach to maintaining robotics safety by enabling the detection and isolation of sensor faults. It involves creating analytical models, which are equations and constraints based upon the mechanical and electrical characteristics of robotic elements. Applying model-based diagnostics to a robotic joint system, employing a series elastic actuator (SEA), this paper focuses on delineating and demonstrating three principal constraints: the Torsional Spring Constraint, the Joint Dynamics Constraint, and the Electrical Motor Constraint. These constraints are configured to detect inconsistencies in sensor readings, which if left unrecognized, may lead to detrimental outcomes.
Model Constraints
The Torsional Spring Constraint connects load position, motor position, and torque using the gear ratio of the SEA and overall stiffness of the system. The Joint Dynamics Constraint is influenced by the motor's torque generation, which is in itself, affected by multiple factors including the inertia and damping of the gearbox, stiffness of the torque-sensing spring and motor properties like torque constants. Lastly, the Electrical Motor Constraint ties the voltage applied to the motor with the current measured and the back electromotive force, reflecting the electrical behavior and operational parameters of the motor, which, if not within expected ranges, could signal sensor faults.
Simulation and Future Work
To convey the utility of this methodology, simulations were run incorporating model-based constraints. As the model assumptions may vary, the constraints also need fine-tuning to match system-specific characteristics. The simulations highlighted scenarios involving torque sensor faults, demonstrating how the methodology could detect discrepancies beyond expected noise levels. Looking forward, extending model constraints to capture thermal and electromechanical relationships could robustify the diagnostic framework. This could lead to more comprehensive monitoring and diagnostics, furthering the safety and functional assurance of robotic manipulators.
In summary, model-based sensor diagnostics offer a customizable and applicable tool on different robot components for fault diagnosis and isolation. This paper contributes valuable insights that are expected to elevate the diagnostic capabilities, essential for the seamless performance of robotic manipulators in collaborating with humans.