Semester of Graduation
Spring 2026
Degree Type
Dissertation/Thesis
Degree Name
Intelligent Robotic Systems
Department
Robotics and Mechatronics Engineering
Committee Chair/First Advisor
David Guerra-Zubiaga
Second Advisor
Razvan Voicu
Third Advisor
Muhammad Hassan Tanveer
Fourth Advisor
Vladimir Kuts
Abstract
Autonomous robots play key roles in many industries, including manufacturing, search and rescue, medical, defense, and others. These robots vibrate during their kinematic operations, producing noise and wear, ultimately leading to required maintenance or system failures. Anomaly detection and fault diagnosis are major concerns for autonomous systems in busy, occupied, and dynamic environments. The ability to characterize the operations of these machines is a critical capability being sought in industry.
To address this challenge, this research developed, implemented, and validated the Vibroacoustic-Informed Twin for Analytics and Learning: VITAL, a novel digital twin (DT) framework that integrates vibroacoustic analysis, machine learning (ML), and physics modeling. This framework creates a real-time virtual representation of a robotic arm instance with enhanced physics-informed kinematics to improve system emulation, detect system anomalies, and predict operational behavior.
This dissertation makes the following contributions:
- The first exploration of a multimodal vibroacoustic ML architecture to predict robot kinematics solely based on vibroacoustic emissions.
- The first instance of a DT framework that incorporates the aforementioned model, hence is connected to the Physical Twin via a tunable range of telemetric and vibroacoustic signals.
- The first Sim-to-real Vibroacoustic domain randomization tuning framework for improved fidelity in robot manipulator simulation.
By increasing accuracy in predictive maintenance, operators can maximize operational capability while minimizing downtime. A DT-based multimodal analytical approach would improve trustworthiness in ML systems. The safety implications of robotics made more aware by advanced DT could mean reduced likelihood of crashes or incidents of injury and death in human-robot interaction.