Visual Odometry

Vehicle Position Estimation uses AI to estimate a vehicle's current position and orientation from sequential camera images, enabling continued navigation when satellite positioning is unavailable, by fusing visual motion signals from image sequences, across autonomous vehicles and GPS-denied operational environments.

Navigation systems dependent on satellite signals lose accuracy or fail entirely in environments with weak or blocked GNSS coverage (such as agriculture fields, tunnels, or urban canyons) creating operational risk for autonomous or semi-autonomous vehicles.

The AI analyses sequential camera image pairs to detect visual motion, integrating displacement estimates over time to produce a continuous position and orientation estimate that substitutes for satellite positioning.

Reduces navigation failure incidents in GNSS-degraded environments; measured as positioning accuracy (error in metres) and system uptime in signal-denied conditions.

• •A calibrated stereo or monocular camera system is mounted on the vehicle and captures continuous image sequences
• •Sufficient compute is available on the vehicle to run inference at the required frame rate
• •Ground-truth positioning data from controlled environments is available for model validation

No intrinsic risk triggered.

No controls triggered.