Enabling Robotics with Data and AI

This talk focuses on the critical intersection of data and Artificial Intelligence (AI) for addressing one of robotics’ core challenges: accurately modeling complex robot dynamics. Traditional analytical modeling often struggles with real-world complexities like friction, backlash, environmental interaction, and inherent nonlinearities, leading to significant discrepancies between theoretical predictions and actual robot behavior. We will explore how data-driven AI methodologies, including machine learning and deep learning, are transforming our ability to learn and represent these intricate dynamics directly from observed operational data. The discussion will cover techniques for data acquisition, intelligent feature extraction, and the development of robust, adaptive models that capture nuances otherwise intractable through first-principles approaches. Emphasis will be placed on how these learned dynamic models enhance control performance, enable more precise simulations, and facilitate advanced capabilities like predictive maintenance and safer human-robot interaction. We will also address the challenges of model generalization, computational efficiency for real-time applications, and the need for explainable AI in safety-critical robotic systems.

May 8, 2025 3:00 PM — 5:00 PM