April 14, 2015: I used to read Bob Pease articles and he would always start with, “What’s All This Stuff, Anyhow?” Bob was an analog guru and he worked on the first reasonably, cost-effective, operational amplifier made for mass production. I start my blog with this title to pay tribute to Bob who was a legend in his own time. I bring this up because Bob died in a car accident in 2011 and the vision of the automotive industry eliminating cars from crashing into each other is starting to become a reality. It has moved from a futurist concept to real work being done in that direction in R&D labs around the world today.
In 2013, NHTSA1 reported 5.7 million vehicle crashes, 32,719 deaths and 2.3 million injuries. These are not small numbers. We have an obligation to our families, friends and children to develop the next innovations that will rocket advanced driver assistance systems (ADAS) into the future.
So what’s all this GPS stuff, anyhow? Why is it relevant to me? Why do we need to simulate GPS in the lab, and what can I do with it? I will answer all of these questions and give you some real world and lab environment examples for GPS positioning data.
These days, GPS receivers are everywhere and in everything -- our phones, our apps, our pictures, our cars, on our assets and on our soldiers. The landscape of cartography, which is the study and practice of making maps, is changing rapidly. The usage of map data is exploding in new system technology. This technology will bring us 10cm high-definition resolution maps which will help the industry develop advanced technologies to save fuel and save lives. New developments that will help you see around a corner at night easier, by adjusting your headlights before reaching a curve or knowing when a large hill is approaching to provide optimized transmission shifting.
The usage of high-definition map data, fused with GPS location information, enables these new vehicle features to be developed. Being able to also have our intersections communicate with our vehicles, and our vehicles communicating with each other, enables the next technological advancements in the mobility industry. Broadcasting your exact vehicle position to another vehicles and adding sophisticated path prediction algorithms will help ensure vehicles do not cross paths one mile at a time, saving countless lives for years to come.
Why do we need to simulate GPS in the lab? I say why not? GPS simulation technology has come down in price. Why do GPS signals have to be any different than a wheel speed or a steering angle sensor? We simulate those in a lab. Why do we do lab testing anyway? It is to ensure that systems, features and functions you’re developing are operating safely and accurately under many conditions.
Lab testing can also be used to create virtual scenarios that could be hazardous to perform on a test track. Once these scenarios are created, you can repeat those tests to accurately compare algorithm changes using the lab to better understand any limitations or undesired system features. This is a great usage of a lab. Being able to perform high crash probability tests, for example, can help reduce test track time by ensuring that when a vehicle hits the test track it has already passed lab, bench and validation testing. All of these kinds of activities help boost confidence in a system design.
The first fundamental rule in experimentation is limiting your variability from unknown factors. When validating system functionality, you should strive to limit variations in back-to-back test runs. This is to ensure that the differences you perceive at the end of your tests only contain the variation you applied during the test and nothing else. Reliability and reproducibility in the system is highly desired. GPS simulation can give you this reliability and reproducibility that is needed in the lab, versus a live GPS broadcast signal in the sky.
Here are some functions that should be tested in the lab vs. the live sky:
Here are a couple of deviation maps of a static location showing the standard deviations of GPS in the lab and from the live sky using a u-Blox EVK GNSS receiver2. The fix mode for the EVK was 3D/DGPS for the entire length of this capture plot.
The picture shows a deviation of around 1.5m of accuracy for a static location, taken from the live sky.
The picture shows a standard deviation of around 5cm for deviation using GPS simulation in a lab environment using a GPS simulator.
In my opinion, there is no better way too purposely test a system than by limiting the variability of your test scenario, which will help isolate specific aspects and outcomes from your system. This will allow easier development of algorithm comparisons during the development process.
Below is a graphical representation of the RF power output of the GPS simulator. The other picture shows satellite power and the resulting degraded navigation fix (blue) and valid 3D navigation fix/lock (green) status of each GPS satellites in view, alongside a visual indication of the navigation path, for time 5 seconds to 25 seconds.
As you can see, varying the satellite’s power will affect your navigation path output in atypical ways.
Graphical representation of the RF power output of the GPS simulator
This picture shows satellite power and the resulting degraded navigation fix (blue) and valid 3D navigation fix/lock (green) status of each GPS satellites in view, alongside a visual indication of the navigation path, for time 5 seconds to 25 seconds.