Start date: As soon as possible
(from 20 February 2026)
Background
The maritime sector is responsible of 3% of annual C02 emissions. One promising approach is the use of
automated kite-based propulsion systems, which can significantly reduce fuel consumption without requiring major modifications to existing vessels. At Aether Swiss Kite, we aim to develop a plug-and-play, automatically controlled kite wing system that would enable large-scale decarbonization across the industry.
Overview
At Aether Swiss Kite, we are developing an automated, plug-and-play kite wing system designed to decarbonize the maritime industry. To ensure the system’s safety and reliability, we are implementing a 360° vision- based redundancy system for real-time kite tracking.
Building upon a previous semester project that achieved reliable 2D detection, trajectory mapping, and orientation tracking, this stage of the project shifts focus to 3D spatial awareness. Extensive research has identified stereovision as the most stable and effective method for obtaining depth data. Therefore, the primary objective this semester is the integration and calibration of a second camera to establish a robust stereoscopic setup.
A key secondary objective is to achieve true real-time detection. During the previous semester, real-time processing was simulated using ROS (Robot Operating System) with two nodes : a camera publisher and an analyzer as the project was conducted remotely. This semester, the goal is to move beyond this mimicked setup to implement a fully functional, high-performance real-time system integrated directly with the hardware.
Finally, once stable 3D positioning (x, y, z) is established, the project will focus on anomaly detection to provide a critical safety redundancy. By linking Computer Vision directly to control commands, the system will be able to react autonomously in case of issues (e.g., proximity to water or structural damage). A major challenge following anomaly detection will be the precise estimation of the wing’s shape, which is essential for optimizing flight performance and identifying subtle deformations.
Objectives
The student will transition the vision system from a 2D experimental phase to a functional 3D safety and diagnostic tool. A significant portion of the work will focus on anomaly detection, where the student will develop the logic to identify structural or situational risks in real-time. This involves moving beyond simple tracking to active monitoring of the kite’s health and safety status.
Key tasks include:
- Stereoscopic Calibration & Setup: Calibrating the secondary camera provided by Aether and designing a stable physical setup to ensure consistent, high-precision stereovision for 3D depth estimation.
- Real-Time System Implementation: Migrating the previous ROS-based “mimicked” real-time architecture into a robust, integrated system capable of processing live video feeds with minimal latency.
- Field Testing & Environmental Validation: Conducting rigorous testing under diverse environmental conditions such as direct sunlight, heavy cloud cover, and varying sea states to ensure the reliability of the 3D positioning system in all weather scenarios.
- Anomaly Detection & Precise Shape Estimation: Selecting and training Computer Vision models to recognize flight irregularities (tears, abnormal trajectories). A key focus will be developing techniques to detect the precise shape of the wing, allowing for a detailed analysis of its aerodynamic state.
- Safety Redundancy & Control Integration: Linking CV outputs to automated control commands. The student will develop the logic to trigger emergency protocols, such as an automatic depower or emergency release, providing a vital safety redundancy in case of kite failure.
Expected outcomes
Through this project, the student will acquire high-level expertise in Computer Vision and Robotics
within a cutting-edge environmental tech startup. Specifically, the student will:
- Gain hands-on experience in stereo-camera calibration and 3D depth estimation.
- Master the transition from simulated ROS environments to real-time hardware integration.
- Learn to interface perception algorithms with control systems for autonomous safety responses.
- Develop advanced models for complex shape detection and real-time health monitoring.
- Participate in field experiments, applying theoretical knowledge to the decarbonization of maritime transport.
Contact & Administration
- Adrien Hoffet [email protected]