Sai Srinivas Tatwik Meesala

Sai Srinivas Tatwik Meesala

Roboticist

AeroFusion: Autonomous BLIMP Navigation and Sensor Integration

Project period: Spring 2025

Role: ROS 2 integration, perception, sensor fusion, controls, and system validation

  • ROS 2 Humble
  • Raspberry Pi
  • C++
  • Python
  • OpenCV
  • YOLOv5

Project Overview

AeroFusion is an autonomous navigation and sensor-integration platform for a biologically inspired lighter-than-air robot. The system was designed for dynamic indoor environments where a BLIMP must detect and approach a visual target while maintaining stable motion with limited onboard sensing and actuation.

View project website · View ROS 2 packages

System Architecture

The platform uses a Raspberry Pi and a modular ROS 2 stack. Camera images feed a custom YOLOv5 detector, which estimates a target’s image-plane location. IMU and barometer nodes provide attitude and altitude measurements. The control stack combines these signals into body-force requests, maps the requested forces to the available thrusters with a Moore–Penrose pseudo-inverse, and converts the result to ESC commands.

Camera → Target detection ─┐
IMU ───────────────────────┼→ State and force estimation → Thruster allocation → ESCs
Barometer ─────────────────┘

Manual joystick commands and autonomous commands pass through a mode-switch node before reaching the ESC driver, preserving a direct operator-control path during testing.

Perception, Control, and Telemetry

  • A custom YOLOv5 model and quadrant logic provide target tracking.
  • PI and differential-control components generate motion commands.
  • IMU and barometer filtering stabilize attitude and altitude estimates.
  • A browser interface built with rosbridge, roslibjs, and web-video streaming displays telemetry and live imagery.
  • Explicit ROS 2 topics and launch files keep sensing, estimation, control, actuation, and user-interface functions separable.

My Contributions

  • Contributed to the ROS 2 node architecture and interface integration.
  • Supported visual perception, sensor processing, and autonomous-control development.
  • Integrated camera, IMU, barometer, joystick, and ESC workflows.
  • Contributed to documentation, launch configuration, and system-level testing.

Validation

The team characterized the propulsion system on a thrust stand and used those relationships in the control-allocation model. SolidWorks mass properties, added-mass estimates, flight-test damping observations, and Vicon motion tracking supported the dynamic model and experimental validation.

AeroFusion BLIMP model

Browser-based ROS telemetry interface

Team

Sai Srinivas Tatwik Meesala, Nihar Masurkar, and Prajjwal Dutta.