Reconfigurable Drones and AI for Autonomous Homes, Spaces, and Environments

figure image

The increasing prevalence of smart home hubs, sensors, and actuators is enabling greater convenience and automation within our built environments. Moreover, the advent of large language models (LLM) and foundation models (FM) seem poised to transform our environments even further, due to their human-like capacity in understanding language and sensor data. However, it is difficult to directly connect this digital intelligence to the physical world due to a lack of support in existing built environments. For example, a hungry person may receive food suggestions from their smart speaker but still needs to retrieve the food themselves.

In this line of work, we explore two ideas in this line of work. 1) First, is the concept that our future buildings, homes, and offices can dynamically discover and leverage heterogeneous compute and sensing as devices get added, removed, or moved throughout the environment.

Drones enables LLMs and FMs to zoom in to any location in the environment, providing the opportunity of fine-grained sensing and actuation anywhere compared to traditional static smart sensors and devices.
Drones enables LLMs and FMs to ‘zoom in’ to any location in the environment, providing the opportunity of fine-grained sensing and actuation anywhere compared to traditional static smart sensors and devices.

2) The second concept is the vision that one of the future devices co-existing with us in our environments will be small reconfigurable drones and robots that can act has hands that LLMs and FMs can control to interface with our environments. For example, a drone can sense any point within the environment, essentially “zooming in” to specific locations (shown above), or bring an item from one point to another due to its mobility.

Modular drone platform that can reconfigure automatically reconfigure its own sensing and actuation depending on the task at hand.
Modular drone platform that can reconfigure automatically reconfigure its own sensing and actuation depending on the task at hand.

We create a modular and reconfigurable drone platform, shown above, that allows LLMs and FMs to dynamically choose modalities to sense or perform actuation depending on the task at hand. For example, the LLM agent can attach a compartment filled with snacks that it can then fly to a person requesting something to eat.

Publications

2025

EmbodiedFly: Embodied LLM Agent with an Autonomous Reconfigurable Drone
EmbodiedFly: Embodied LLM Agent with an Autonomous Reconfigurable Drone
Minghui Zhao, Kaiyuan Hou, Junxi Xia, Yanchen Liu, Stephen Xia, Xiaofan Jiang
ACM Transactions on Internet of Things  ·  22 Oct 2025  ·  doi:10.1145/3772079
FlexiFly: Interfacing the Physical World with Foundation Models Empowered by Reconfigurable Drone Systems
FlexiFly: Interfacing the Physical World with Foundation Models Empowered by Reconfigurable Drone Systems
Minghui Zhao, Junxi Xia, Kaiyuan Hou, Yanchen Liu, Stephen Xia, Xiaofan Jiang
Proceedings of the 23rd ACM Conference on Embedded Networked Sensor Systems  ·  06 May 2025  ·  doi:10.1145/3715014.3722081
DomAIn: Towards Programless Smart Homes
DomAIn: Towards Programless Smart Homes
Yueyuan Sui, Yiting Zhang, Yanchen Liu, Minghui Zhao, Kaiyuan Hou, Jingping Nie, Xiaofan Jiang, Stephen Xia
Proceedings of the 3rd International Workshop on Human-Centered Sensing, Modeling, and Intelligent Systems  ·  06 May 2025  ·  doi:10.1145/3722570.3726888

2024

Connecting Foundation Models with the Physical World using Reconfigurable Drone Agents
Connecting Foundation Models with the Physical World using Reconfigurable Drone Agents
Minghui Zhao, Kaiyuan Hou, Junxi Xia, Stephen Xia, Xiaofan Jiang
Proceedings of the 30th Annual International Conference on Mobile Computing and Networking  ·  04 Dec 2024  ·  doi:10.1145/3636534.3698846

2021

A Drone-based System for Intelligent and Autonomous Homes
A Drone-based System for Intelligent and Autonomous Homes
Stephen Xia, Rishikanth Chandrasekaran, Yanchen Liu, Chenye Yang, Tajana Simunic Rosing, Xiaofan Jiang
Proceedings of the 19th ACM Conference on Embedded Networked Sensor Systems  ·  15 Nov 2021  ·  doi:10.1145/3485730.3492881