We are seeking a skilled Computer Vision / SLAM engineer to build a monocular Visual Odometry / Visual‑Inertial Odometry (VIO) solution for UAVs. We’ve already tried Kimera, OpenVINS, ORB‑SLAM3, and VINS‑Fusion, but results under real UAV flight conditions are not sufficient. You will develop and evaluate your solution in a simulator (large, realistic terrain, full telemetry, IMU + camera data, ground truth), then validate on real UAV flights.
Key Requirements & Constraints
Sensor Setup: Monocular camera + IMU (no stereo)
Flight Envelope:
Altitude: 50 m to 500 m
Speed: 100 km/h to 300 km/h
Real-time coordinates estimation
On-the-fly calibration: camera intrinsics, camera‑IMU extrinsics, time sync
Feature tracking
Robust to fast maneuvers, lighting changes, and feature-sparse scenes
Clean, modular code (Java preferred, C++, Go), minimal overhead dependencies
Logging, visualization, benchmarking, test coverage
API integration for embedding into UAV stack (Raspberry Pi on Ubuntu)
What You’ll Receive / What We Provide
Simulator environment with terrain, telemetry, and ground truth
Real flight datasets for post‑simulation testing
Hardware specs (onboard / companion compute)
Support in integration, testing, and domain knowledge
Proposal Instructions
Please include in your proposal:
High-level approach/architecture (how you’d tackle VO / VIO in this envelope)
Past work or examples in SLAM / VIO / UAV
Milestones, timeline, and budget
Questions, assumptions, or clarifications needed
Kyiv