Demonstrations

Two ROSbot Pro 2 robots demonstrate perception-aware leader-follower control using Control Barrier Functions (CBFs) to maintain visual contact. Our approach reliably keeps the leader within the follower's field of view, while conventional formation control without CBFs often loses visual contact, leading to unsafe behavior.

A perception-aware distributed formation control scheme for agents using body-fixed cameras with limited field of view (FOV). Our CBF-based approach ensures leader visibility by incorporating FOV constraints, while neural network and double bounding box estimators provide robust state estimation from real-time image data across various environments.

Our FACETS-derived model running on the MAX78000 microcontroller demonstrates efficient object detection within tight hardware constraints (432 KB memory, 32-layer limit). Despite these limitations, it achieves 45.4% less energy consumption, 29.3% lower latency, and 4.5% higher mAP compared to Analog Devices' TinierSSD baseline.

A hierarchical framework integrating natural language commands with autonomous quadrotor navigation using the ModalAI Seeker drone. Our fine-tuned LLaMA model (via LoRA) interprets high-level instructions into task goals, while the drone executes them through onboard VIO-based control, path planning, and real-time object detection with hardware-in-the-loop testing.

Computer Engineering Technology undergraduates designed, assembled, and programmed this custom drone from scratch as their senior capstone project. Working under NAIL Lab guidance, the drone features autonomous navigation using OptiTrack motion capture and real-time onboard object detection—demonstrating hands-on robotics and AI development by UH students.

Demonstration of advanced multi-robot coordination where an autonomous UAV successfully lands on a moving ground vehicle, showcasing precise real-time control and inter-robot communication.

Live demonstration of our ModalAI Seeker drone performing simultaneous localization and mapping (SLAM), autonomous motion planning, and real-time object detection in complex environments.

Complete autonomous navigation demonstration featuring our Jackal robot performing SLAM, dynamic obstacle avoidance, and object detection in both indoor and outdoor environments.