PATHFINDER

1

Why have you developed your own software GNSS receiver when there are good and cheap hardware receivers out there?

Powerful GPU-based compute platforms for ADAS and Autonomy (e.g. NVIDIA DRIVE) are available at suitable cost levels for mass market & series deployment end enable our solution to scalable and cost-effective replace a legacy hardware receiver with the latest functional safety oriented positioning algorithms. Unlike a black box solution, the software approach gives us full control over the entire signal processing chain, starting with the digitized antenna signal. This is the basic requirement for generating GNSS measurements with high integrity for functionally safe integration into the sensor fusion environment. In contrast to an integrated circuit the software approach enables seamless interfacing to the sensor fusion environment. Our software can easily be tailored to specific use cases and allows deeply coupled sensor fusion. Since PATHFINDER is written in C++ according to strictly controlled coding standards, it runs on all ADAS platforms with GPU‘s and is compatible with the most common middlewares. Our software can be quickly upgraded to future signals and services (e.g. Galileo HAS or other modern GNNS argumentation services like SPACORDA).

2

How many position candidates and confidence pairs come from PATHFINDER?

That currently depends on the receiver mode (single point positioning, D-GPS, RTK) and is ultimately depending on both the environment and receiver configuration. For example, under open-sky reception conditions a single candidate is usually enough to approximate the probability density of the position estimation well enough. Only under multipath and/or non-line-of-sight conditions ambiguities arise, which may be further increased by the ambiguity problem RTK has to resolve anyhow. In this case, 2-5 candidates are usually a fair enough representation.

3

How do you handle multi-path & NLOS situations?

Our GNSS receiver employs probabilistic models within the signal tracking itself to assess and mitigate the presence of multi-path and non-line-of-sight conditions. This leads to a reliable confidence estimation for the receiver measurements and increases its resistance to localization biases caused by such effects (especially during short-term signal disturbance).

4

Is there an indicator of the current integrity of the localization?

Confidence is expressed using average Normalized Estimation Error Squared (A-NEES). PATHFINDER confidence estimate maximum in deeply coupled mode is A-NEES < 10.

5

PATHFINDER uses IMU data internally, so it can output position candidates even though fewer than four satellites are in sight, right? And a low-cost IMU can meet the PATHFINDER requirements? What‘s the IMU model in your reference board?

PATHFINDER itself does not require an IMU to function, although such additional input can highly increase its performance depending on the chosen data fusion scheme. If PATHFINDER is to be integrated into a larger data-fusion based localization engine that already uses such an inertial sensor, then IMU data shall only be used to tune its signal tracking dynamics. This would still improve its tracking performance without violating the data-fusion systems presumptions (fusing a sensor twice is usually not a sound approach, especially if the caused correlations are ignored). If this is not the case, then PATHFINDER can deeply integrate IMU sensor data on its own to fully exploit its usefulness. It is designed to work with arbitrary IMU models if the noise/drift parameters of that IMU can be adequately determined. We usually do not supply an IMU on our own, PATHFINDER is software, although we can point out models we used so far. As a last note: As most modern GNSS receivers, PATHFINDER can update its position estimation using a single satellite in sight as well. The four satellites requirement is true for trivial single positioning algorithms only, that need to solve the position and clock bias estimation at each update step. Using more sophisticated assumptions on the receiver dynamics softens this requirement. However, this is only suitable to bridge short term signal interruptions like at an underpass. Understandably, precision will suffer over time (but slower with an IMU) and confidence estimation will disclose this.

6

In case of cold booting – when will PATHFINDER be “alive” again?

As described at the previous question, this depends on the information the receiver can still rely upon. If it leaves a GNSS denied area, it will regain position estimates within few seconds, typically between 5s and 15s. That’s usually called the warm or normal time-to-first-fix (TTFF). There is also a “cold TTFF” describing the time to the first position if the receiver has no valid information at all. In this case it needs to find available satellites and retrieve the required information from them. In traditional, unassisted operation, this tends to take around 40s to 60s.

7

In case of a long tunnel the IMU will take over the positioning. In most cases the performance will not be sufficient because of the drift. How can the NAVENTIK Technology cover this kind of a worst case scenario?

By principle, long term operation in a GNSS denied environment leaves the scope of our technology. Depending on its performance, this allows bridging signal gaps from several seconds up to few minutes, with ever increasing covariance estimations. PATHFINDER is precisely designed for integration into a large-scale positioning data fusion system, maximizing the use of GNSS. ADAS needs precise and reliable positioning during a broad range of conditions, something no single sensor can achieve. Optical systems like camera and Lidar based visual odometry can be helpful to reduce the drift, but can fail due to poor lighting conditions, map and feature based approaches fail in unmapped areas or if the surrounding lacks notable features and GNSS fails under poor radio reception conditions. To solve this, many different sensors and a lot of know-how must be combined to function under all conditions sufficiently, with NAVENTIK covering the GNSS part and integration support.

PATHFINDER FAQ

If you’re into GNSS and sensor fusion this guide will help you learn more about NAVENTIK's technology and its features.

1