Hyperspectral Autonomous Robot

A Research Project to help build foundational models for Hyperspectral Data.

Technologies Used: Machine Learning, Pytorch, Robotics & Reinforcement Learning.

Overview

Pixxel Space envisions a constellation of hyperspectral satellites for capturing crucial data.

However, the quality of data collected is influenced by atmospheric conditions, such as light absorption and scattering. To enhance data accuracy, I designed and developed hyperspectral scanning equipment—a robotic actuator machine. This machine is set to capture data at various heights and speeds, aiding in data calibration and the creation of more accurate artificial intelligence models by providing a true value of reflectance.

Use cases of Hyperspectral Data

IMAGE (C) EPFL

Detailed Spectral Information: Captures data across narrow, contiguous spectral bands for precise analysis.
Material Identification: Enables accurate identification and classification of materials based on unique spectral signatures.
Environmental Monitoring: Tracks environmental changes, such as vegetation health and water quality.
Disease Detection in Agriculture: Aids in early detection of crop diseases and stress.
Forestry and Land Management: Supports forest health assessment, tree species identification, and land use monitoring.

Research Insights

Pixxel Space's project is founded on valuable insights:

Light Interaction in the Atmosphere: Understanding how light interacts with the atmosphere is vital for collecting accurate hyperspectral data.
Machine Requirements: Determining the speed and degrees of freedom needed at different heights to optimize data collection.

Guide: Dr. Rahul Raj (Senior Research Scientist at Pixxel Space), Tejaswi Hareesh (VP Mechanics Subsystem)

Design:

The project encompasses several design phases:

Hardware Requirements: The hardware for the autonomous robotic actuator machine is meticulously planned. This includes the design of the entire setup and the consideration of deformation analysis to ensure optimal performance.
Manufacturing: Renowned tools like 3D printing and laser cutting are employed to manufacture the equipment.
Motor Control Systems: A comparative study of motor control systems is undertaken to select the most suitable solution for the project.
Coding: The development of autonomous machine code, with software platforms such as Arduino IDE and Protuner CS-D, is essential for precise control.

Prototyping:

A robust robotic autonomous setup is constructed and tested to capture hyperspectral data in real-world scenarios. The data collected from different vegetation species is critical for calibration and speed assessment, contributing to data correction protocols.

Future Work

The Pixxel Space project is an ongoing initiative with several future goals:

Robot Usecases: The project aims to incorporate actuators into the fields to collect data and calibrate hyperspectral data received from satellites.
Artificial Intelligence Models: The machine will play a pivotal role in creating artificial intelligence models to correct and refine data, aligning with the long-term objectives of Pixxel Space's R&D department.

Reflections:

I got to work on multidisciplinary domains, and it has broadened my horizons, enriched my problem-solving skills, and allowed me to contribute to cutting-edge solutions that address complex, real-world challenges.
The project has been applied to analyze spectral signatures related to both oil leakage and agricultural data, enhancing the insights derived from climate-related information.
The project evolves, to contribute to our understanding of our planet's dynamic ecosystems and climate, and to the development of cutting-edge artificial intelligence models that will shape the future of space exploration and Earth observation.