Athena Seminar Series: Embodied AI at the Edge: Embedded Sensing and Mobile Agents for Next-Generation Intelligent Environments

We have witnessed an explosion in the capabilities of foundation models and artificial intelligence to reason and generate content in the digital domain. Despite these advancements, there still exists a fundamental disconnect between their intelligence and our everyday physical environments. To bridge this gap and create truly ubiquitous intelligent environments that are capable of supporting everything from smart living for aging populations to autonomous urban management, we need to design and deploy adaptable, context-aware AI on the edge that can perceive, act, and operate robustly in complex physical environments. Towards this vision, I will present several lines of work that 1) exploit the physical interaction between robotic agents and their environment to perform autonomous, sensorless mapping of 3D airflow fields using only the flight dynamics of low-cost micro-drones, 2) democratize the deployment of adaptable edge sensing through a modular, plug-and-play IoT sensing system that enables the rapid, accessible integration of customized hardware for AI-driven applications, and 3) demonstrate how interfacing modern foundation models with a reconfigurable drone enables embodied agents to autonomously perceive, understand, and adapt to diverse physical tasks.