ORCA: Underwater Robot

An Inspection Class Remotely Operated Underwater Vehicle for Autonomous and Manual Navigation
Award Winner
Computer Vision
Deep Learning
Deployment
Funding
GPU Training
Human-Computer Interaction
Machine Learning System
Monitoring
On-Device AI
Publication
Research
Robotics
Signal Processing
Author

Nivedhitha Dhanasekaran

Published

June 24, 2022

Project Overview

A 500x350x210 mm inspection-class underwater robot equipped for hybrid manual-autonomous navigation. ORCA features a 6-thruster mobility suite, real-time telemetry, and multi-sensor feedback including pressure, temperature, leak, and IMU modules. This system integrates a joystick-driven manual mode and an autonomous navigation module built on MOOS-IvP (C++ mission planning suite from MIT).

✨ Led a 7-member team to build and test ORCA’s software stack with kill-switch safety, homing, and GAN-enhanced visibility.
✨ Achieved >90% mission success across test runs with real-time diagnostics and fail-safe handling.
✨ Integrated autonomous control, camera feedback, and tethered communication using custom C++ modules.

Timeline of Contributions

1. ORCA Robot: Joystick Operation
  • Developed an autonomous + manual navigation stack on MOOS-IvP.
  • Built joystick control, kill switch, IMU fusion, and thrust mapping systems.
  • Bench + pool tested with full sensor telemetry and thrust validation.
  • Published at IEEE and won the Asia-Pacific Region IEEE R10 UGPVC.
2. KYOGRE: Autonomous Navigation Suite with Perception Sub-systems
  • Internally funded with β‚Ή30,000 (SSN IFSP).
  • Developed deep RL-based waypoint control for UWVs.
  • Integrated into ORCA system as a future planning module.
3. EyeSea: Real-Time Marine Threat Alerting Sub-system
  • Built a YOLOv3 + MobileNetV2 pipeline to detect marine species threats.
  • Arduino-based IMU fusion + tethered alert system for swimmer safety.
  • Developed for National Institute of Ocean Technology, winning Smart India Hackathon 2022 (Prelims).
4. SatVision: Satellite-Based Built-Up Detection
  • Developed a two-step segmentation and merging approach using satellite imagery.
  • Identified non-residential built-up clusters in dense Indian cities (e.g., Mumbai, Delhi).
  • πŸ… Won Smart India Hackathon 2022 – Software Edition, awarded β‚Ή1,00,000 by Ministry of Earth Sciences.
5. C-GAN for Enhanced Localization
  • Designed a Conditional GAN to refine robot localization sequences.
  • Corrected drift and improved tracking precision for mission-critical tasks.
  • Contributed to robustness of multi-agent localization under low-visibility conditions.

Awards & Recognition

  • πŸ† IEEE R10 Undergraduate Project Video Contest Winner (Aug 2021)
    – Winner from 7 countries and 50+ councils
    – Project video featured on IEEE YouTube channel
    – $300 cash prize

  • πŸ… Smart India Hackathon 2022 – Software Edition
    – Winner for PSID: GR823 (SatVision Project)
    – β‚Ή1,00,000 grant awarded by the Ministry of Earth Sciences, Government of India

  • πŸ’° Internally Funded Student Project (SSN College of Engineering)
    – β‚Ή30,000 grant for KYOGRE (autonomous navigation sub-system for ORCA)

Tools & Frameworks

System Layer Tools / Frameworks
Navigation & Control C++, MOOS-IvP, Arduino, Joystick
Computer Vision YOLOv3, MobileNetV2, OpenCV, GAN, Real-time Video Analytics
Embedded Systems IMU Fusion, Leak Sensor, Pressure Sensor, Arduino Serial Comms
Robotics 6-Thruster Dynamics, Manual-Autonomous Switch, Kill Switch Logic
Deployment & Monitoring Telemetry Monitoring, Underwater Testing, Sensor Drift Handling