The automation of mining haulage vehicles has great potential in terms of safety and economy. Particularly in hybrid mines, where surface, as well as underground mining, is carried out, an increase can be expected. The performance of autonomous vehicles depends to a large extent on highly accurate vehicle state information. Deep mines are especially challenging, as satellite-based localization methods are reaching their limits. Therefore, we introduce a new navigation filter concept for the precise and robust localization of the haulage fleet for surface operation as well as the transition zone between surface and underground operation. The multi-sensor navigation filter utilizes an inertial measurement unit (IMU) and is aided by signals of GPS and Galileo satellites. To cope with the challenges, we introduce a new optical speed sensor update step within the tightly-coupled Unscented Kalman Filter (UKF). The speed sensor measures continuously the slip-free two-dimensional speed above ground. As the sensor cannot be aligned perfectly by hand, a new state is introduced which represents the misalignment angle error. Our new filter approach was validated on a real-time hardware with different test drives with an articulated dumper. The new filter achieved a mean position error of 0.24 m and a heading error of 0.78° during a test drive of 152 s with a simulated GNSS outage of 80 s, with respect to the RTK reference position. Later in the project, the hardware setup is attached to a Bell e30 dump truck with a load capacity of 28 tonnes and evaluated in an open-pit mine under real mining conditions.