Vehicle Mileage: Why GLONASS and Odometer Differ? Part 5. Data Quality and Validation

_

_

The discrepancies in satellite navigation data that we discussed in previous articles are often due not only to calculation methods, but also to the quality of the data itself.

In this article, we will analyze the remaining 3 fields from our dataset, except for the odometer readings obtained via the CAN bus. These fields contain values ​​that determine the reliability of the satellite data:

• satellites_count  – the number of visible satellites when solving a navigation problem.
• hdop –  horizontal precision reduction factor.
• valid  – the final flag of data validity.

We will analyze how these indicators are formed, what values ​​are considered acceptable and how to interpret them when analyzing data. Based on a real dataset, we will analyze the impact of data quality on the accuracy of mileage calculation.

Number of visible satellites

The number of navigation satellites is one of the key factors determining the accuracy and reliability of data received by a satellite navigation system. It directly affects the ability to solve a navigation problem and the accuracy of location determination.

At least 3 satellites are required to determine the coordinates on the plane (latitude and longitude) using the trilateration method. And at least 4 satellites to add the measurement of altitude above sea level. If the number of visible satellites falls below the minimum level, the navigation task becomes unsolvable, and the tracker loses the ability to calculate coordinates. Even with 4-6 satellites, the accuracy of the data can be reduced due to their poor geometric arrangement (for example, satellites are “concentrated” in one area of ​​the sky). 

Signal quality from the number of satellites:

• Good signal (10 or more satellites) . High coordinate accuracy, error within a few meters. Reliable altitude determination and stable data on speed and direction of movement.
• Average signal (6-8 satellites) . Coordinate error can reach tens of meters. Significant errors in altitude calculation. Strong dependence on the geometric arrangement of satellites (HDOP).
• Bad signal (less than 4 satellites) . Inability to solve the navigation problem. Interruptions in determination or sharp jumps in coordinates.

Factors affecting the number of visible satellites:

• Location : In cities, the signal can be reflected by buildings, causing “multipath”. In tunnels and underground parking lots, the number of visible satellites is sharply reduced or completely absent.
• Equipment . Modern receivers work with several systems (GPS, GLONASS, Galileo, Beidou), increasing the total number of visible satellites. Old receivers that support only one system are inferior in this.
• Weather conditions : Heavy precipitation or dense clouds may degrade the signal quality, although the number of satellites remains the same.
• Electronic warfare (EW) . GPS jammers can artificially create a situation where the number of visible satellites becomes insufficient.

In situations with a limited number of satellites or poor satellite locations, it is necessary to apply data filtering, combination with inertial measurement units (IMU) and take into account indicators such as HDOP to reduce errors in calculations. For example, the standard configuration of Navtelecom equipment rejects data received from less than 4 satellites.

Horizontal Decay of Accuracy

DOP (Dilution of Precision) is a term used in the field of global positioning systems for the parametric description of the geometrical mutual arrangement of satellites relative to the receiver antenna. The following components of DOP exist:

• HDOP (horizontal DOP)  – reduction of precision in the horizontal plane (latitude and longitude).
• VDOP (vertical DOP)  – reduction of precision in the vertical plane (height).
• PDOP (position DOP)  — positional precision dilution:  PDOP^2 = HDOP^2 + VDOP^2.
• TDOP (time DOP)  – time reduction of precision.
• GDOP (geometric DOP)  — total geometric reduction in precision in location and time:  GDOP^2 = PDOP^2 + TDOP^2.

In transport telematics, car trackers most often provide only the HDOP value, which reflects the quality of the geometric arrangement of satellites relative to the receiver for determining latitude and longitude. The lower the HDOP value, the more accurately the coordinates of the object are determined. This parameter is one of the key indicators of data accuracy in satellite navigation.

HDOP depends on how the satellites are positioned on the sky relative to the receiver. If the satellites are far apart (wide sky coverage), the HDOP will be low, indicating high accuracy. If the satellites are clustered in one area, the HDOP increases and the accuracy decreases. HDOP shows how large the coordinate error is due to the geometric arrangement of the satellites, even if the signals from each of them are received without errors.

Location accuracy depending on HDOP:

• 0-1 – Ideal accuracy . Such values ​​are typical for open spaces and high-quality equipment with excellent reception conditions and a large number of satellites. The coordinates have an error within 2–5 meters.
• 1-3 – High accuracy . Suitable for most tasks, including vehicle monitoring and mileage calculation. The error is usually in the range of 5-10 meters.
• 3-6 – Medium accuracy . Can be used in calculations, but the error can reach 20-50 meters, especially if the satellites are low above the horizon or grouped in one part of the sky.
• 6 or more – Low accuracy . The error exceeds 100 meters, the data becomes useless. This level of HDOP is observed in dense urban areas, in tunnels, or with poor satellite placement. 

Factors affecting HDOP:

• Number of visible satellites . The more satellites, the lower the HDOP. The minimum required number of satellites is 4, but for a low HDOP, 6 or more are needed.
• Satellite Geometry : The spread of satellites across the sky is a critical factor. If satellites are clustered in one part of the sky, HDOP increases.
• Receiver location : In tunnels, under bridges or in dense forests, signal quality deteriorates and HDOP increases.
• External interference : GPS signal jamming or multipath (signal reflections from buildings) also increases HDOP.

Low HDOP allows trusting the position, speed and direction data, while high HDOP requires data filtering or correction. Navtelecom equipment configuration by default discards any coordinates with HDOP higher than 2.5.

Validity flag

The validity flag is a parameter that determines how reliable and usable the data provided by the satellite receiver is. This indicator plays a key role in filtering and processing data in satellite navigation and transport telematics systems. 

It is installed at the level of navigation equipment, which analyzes the quality of incoming data from satellites and makes a decision on validity based on built-in algorithms. If the data does not meet the minimum accuracy requirements, the validity flag is removed.

The number of visible satellites and HDOP are the main parameters that determine the validity of coordinates. But you can also use speed and altitude values, as well as inertial and other sensors. You can read more about setting up validation in Navtelecom equipment  in their documentation .

Dataset

The validity flag, as well as the number of visible satellites and HDOP, are optional parameters in navigation telemetry, so not all car trackers are capable of sending them. And even if such functionality is available, the transmission of this data can simply be disabled in order to save traffic. Therefore, we will first select from our sample trackers that have the fields we need:

The statistics for the validity flag, number of satellites and HDOP on our entire dataset are as follows:

• valid  – validity flag
  • true : 189 139 560 – the tracker considers the data valid
  • false : 4 802 132 – the tracker considers the data invalid
  • null : 940 967 – tracker does not have a validation function
• sat_count  – number of satellites
  • exist : 194,882,659 – number of satellites known
  • null : 0 – in our sample all trackers send the number of satellites
• hdop  – precision reduction factor
  • exist : 113 206 680 – meaning hdop known
  • null : 81 675 979 – hdop value unknown

The “rarest” parameter is HDOP, and we will filter the tracker selection from the dataset by it. About half of the trackers are suitable for us, and we will take the leader in terms of telemetry:

We are lucky, because there are “flights” here that negatively affect the correctness of the final mileage of the vehicle:

The distribution by the number of visible satellites is close to normal around the value of 17. This is a good indicator, because with more than 10 satellites we can expect good quality of coordinates.

Next is the dynamics of the number of visible satellites over a short period. A pattern of the number of satellites falling to zero at the beginning of the next trip is observed. This indicates a cold start of the GPS module when the mass is turned on in the car and the initialization of the tracker connected directly to the on-board network.

Now the distribution of values ​​by HDOP. Almost all telemetry was recorded with the optimal location of navigation satellites in the sky. And this is not surprising, since above we were convinced of their sufficient number.

Dynamics of HDOP values ​​for the initial interval in the dataset:

Distribution of validity flag values. The tracker considered almost all telemetry valid and suitable for further processing.

The dynamics of the validity flag over the initial interval. Very similar to the pattern of the decrease in the number of satellites at the start of the car’s movement, right?

We already know that HDOP and the number of satellites are parameters for calculating the final value of the validity flag, so let’s look at the relationship between these values:

And we will calculate the linear correlation: 

There is an average positive correlation between the validity flag and the number of satellites, and an average negative correlation between the validity flag and HDOP. The correlation sign is obvious, since with an increase in the number of satellites and a decrease in HDOP, the quality of the data improves, and vice versa. The average linear relationship only indicates that the number of satellites and HDOP are not the only parameters for calculating the validity flag.

It’s interesting to look at the correlation of these values ​​across the entire dataset. Let’s run the calculation on 100+ million records and see what we get:

We didn’t see anything new here, the correlation remained in the same intervals. We didn’t get any mysterious insights here, which was generally expected. But it is important to note that a sufficient number of visible satellites does not guarantee excellent HDOP, since satellites can be concentrated in one area of ​​the sky and/or be too low above the horizon.

Software filtering

For example, let’s take the same tracker for a month, when 2 types of jamming were observed: “teleport” and “circles”. Telemetry was received from the Navtelecom tracker with the same parameters as in Figure 7. To begin with, let’s build a track for all points without exception and calculate the mileage. We get 1,758 km.

Now let’s try to calculate the mileage and build a track only for points with the correct validity flag. The final mileage has decreased slightly, to 1,745 km. But the circles and teleport are still there. Validation on the hardware side is clearly not enough here.

Let’s try to tighten the requirements for HDOP to 2 units. Teleport and circles remained, but the mileage decreased to 1,732 km.

Now let’s try to increase the minimum threshold of the number of visible satellites that participated in solving the navigation problem to 10. And here is the result, the teleporter completely disappeared and cut off some of the circles. The mileage is now 1,452 km.

Well, let’s play with the altitude. The topographic map says that in this region the altitude is no more than 125 meters above sea level, so we’ll filter out everything above 100 meters. As a result, there are almost no circles left, and the mileage has dropped to 1,407 km.

Thus, using software filtering, which essentially only partially repeats the validation algorithm inside the equipment, but with more stringent parameters, we reduced the mileage calculated by GNSS from 1,758 km to 1,407 km – by as much as 351 kilometers.

Of course, we cannot restore the data that was lost due to jamming and do not take into account the real mileage in these areas, which cannot be calculated using GNSS in jamming zones. But at least we avoid “winding up” fake mileage. 

Conclusion

The reliability and accuracy of satellite navigation data is the basis for effective transport monitoring. Data quality is determined by many factors, the most important of which are:

• Number of visible satellites.
• The level of horizontal dilution of precision.
• A validity flag that combines all quality indicators.

Analysis of the real dataset showed that the data quality can be improved by filtering by specific telemetry parameters. For example, the equipment filters settings exclude points with high HDOP or insufficient number of satellites. However, as the correlation showed, the validity of the data is not limited to these parameters alone — it is a complex indicator. To improve the accuracy of calculations, it is important to use additional data, such as IMU readings and other sensors.

The next part of the article series will be the final one. In it, we will study the odometer readings via CAN and compare the mileage via GNSS taking into account everything we already know. For example, how much mileage was “eaten” by the teleport and the circles above.