# Vision-Aided Velocity Estimation in GNSS Degraded or Denied Environments

## TL;DR

This paper presents a new navigation system that estimates vehicle velocity using vision data when GPS signals are unreliable or unavailable.

## Contribution

The novel architecture uses a landmark-extraction algorithm and a Sequential Kalman filter to maintain reliable velocity estimation in GPS-denied environments.

## Key findings

- The system effectively estimates velocity in GPS-degraded scenarios.
- Landmark tracking supports LiDAR odometry and reduces velocity error.
- Testing confirmed stable performance across various real-world conditions.

## Abstract

This paper introduces a novel architecture for a navigation system that is designed to estimate the position and velocity of a moving vehicle specifically for remote piloting scenarios where GPS availability is intermittent and can be lost for extended periods of time. The purpose of the navigation system is to keep velocity estimation as reliable as possible to allow the vehicle guidance and control systems to maintain close-to-nominal performance. The cornerstone of this system is a landmark-extraction algorithm, which identifies pertinent features within the environment. These features serve as landmarks, enabling continuous and precise adjustments to the vehicle’s estimated velocity. State estimations are performed by a Sequential Kalman filter, which processes camera data regarding the vehicle’s relative position to the identified landmarks. Tracking the landmarks supports a state-of-the-art LiDAR odometry segment and keeps the velocity error low. During an extensive testing phase, the system’s performance was evaluated across various real word trajectories. These tests were designed to assess the system’s capability in maintaining stable velocity estimation under different conditions. The results from these evaluations indicate that the system effectively estimates velocity, demonstrating the feasibility of its application in scenarios where GPS signals are compromised or entirely absent.