Co-located Antenna and IMU

We sell a significant number of IMUs. The IMU04 came out last year and it’s popular because integration with GPS gives stability to the velocity output when conditions are sub-optimal – like under trees. And we’ve worked really hard on this product so that the pitch and roll accuracies are now down to 0.06 degrees (RMS).

All well and good but there is an element in using one that can be a bit of a pain. You have to measure, very accurately, the distance between the GPS antenna and the IMU. Sounds simple enough but it can be really tricky, especially if there’s a significant curve to the car roof.

Why does this need to be measured? Well there are a couple of reasons. Firstly, the performance of the Kalman Filter is optimised when the correct distance is known. And secondly, because of Lever Arm Correction: a vehicle’s roof will initially travel faster than its centre of gravity when the brakes are applied. And where’s the antenna? On the roof… so integrating the IMU with the GPS data allows for this overshoot to be adjusted; but you’ve got to know how far away they are from each other and this measurement introduces the potential for human error.

IMU-lever-arm
Lever arm correction in braking

So here’s the solution: co-locate the IMU and antenna. This rather handy little unit allows the IMU to fit snugly inside it, with the antenna mounted on top. With the data sources coming from exactly the same place, the overshoot is eliminated.

“But” you say “with the IMU on the roof, it will pitch forward further than it would do if was mounted inside the car at the COG.” Yes, it will, but this can be translated back to where you’d normally fit it, and that’s much easier than measuring between the antenna and IMU. It also means that you can swap the setup between vehicles far more quickly.

IMU_roof-mount
Co-located GPS antenna and IMU.

Coming to a Test Track Near You…

…Well almost. MIRA and Heyford Park, to be exact.

We have got ourselves a hospitality trailer. Why? Well I detect a bit of a trend with the trade shows we do: they’re still worthwhile but slowly becoming less so, with fewer of the engineers and decision makers that we’re used to seeing. They can’t get out of the office: too much on their plate.

So we thought we’d go to them. We’re parking up at MIRA Proving Ground where a lot of our UK customers are based, on the 12th May; and then to Heyford Park in Oxfordshire two days later.

The best thing about this is that we aren’t limited to what we can show people on an exhibition stand – we can actually demonstrate things. We’ll have a braking setup, a steering robot (courtesy of ABD), and a couple of cars showing vehicle separation/ADAS. And if you’re not in one of those demos, come into the trailer for coffee, cake, and a chat.

Once we’ve completed our UK dates, the trailer heads off to Europe where it will visit the likes of Renault, Goodyear, and Continental. I suspect that in 2016 we’ll look to hire some facilities in Germany and invite a whole load of customers along for the day.

Look out for the splashy BMW on the motorway.

VBOX-roadshow

VBOX Test Suite

If there’s one thing I’ve learned in my travels over the last few years, which has seen me visiting a good variety of customers, it’s that they all want software for specific tasks.

This is not to downplay the role of all those frighteningly intelligent test and calibration engineers out there, who are all perfectly capable of driving VBOX Tools and setting up very complicated test structures. It’s just that they’re under a lot of pressure and it kind of makes sense to be able to send out a less qualified driver to carry out procedures – but that to do this they need software that is really easy to use.

So. Taking this feedback we set about designing the eventual replacement for VBOX Tools. Our next gen analysis software had to be really straightforward to use but equally capable of processing complex data, and cover the whole gamut of testing and dynamics in which we specialise. It was decided that the best way to achieve this is to have a ‘base’ program, and then augment it with plugins. That way each test department will only have to have the level that they need; so those who’re only interested in straight-line performance will have a nice simple setup in comparison to their colleagues in ADAS who require more.

Here it is – VBOX Test Suite. At the moment it’s just a relatively simple performance analysis package, but we’re working on the first range of plugins that will cover all the ADAS stuff as well as aquaplane, coastdown, NVH, and so on. The nice thing about plugins is that they’re customisable – we can make them specifically for any application. Let me know what you want.

I’ll let the video explain how it works.

 

New data analysis software – VBOX Test Suite from Racelogic on Vimeo.

Park Assist Testing

Park-assist is a brilliant thing, isn’t it? And it shouldn’t even be called “park-assist” – it should just be called “park”, because that’s what the latest cars do. On their own. I love it.

OK so most of them need you to apply the throttle but it’s still a great aid. My colleague Matthias who works in our German office had an interesting day recently, testing a whole load of different manufacturer’s park assist capabilities. He did this for AutoBild magazine which is renowned for carrying out some pretty comprehensive consumer testing and whose editorial carries a lot of weight with the car buying public.

BMW-2er-Active-Tourer-1200x800-33011827d5e1cedd

Using the VBOX ADAS test system he mapped the parking space. It’s actually the Lane Departure mode that is running when the car performs the manoeuvre – we’ve changed the firmware: now four contact points of the vehicle are input so that the car’s profile can be accurately measured against the lane edge.

Park-Assist

Then a VBOX HD was also set up to film the cars as they backed into the space with the separation data all being logged and displayed. Nine cars were tested, from VW, BMW/Mini, Hyundai, Peugeot, Audi, Mercedes, Ford, and Skoda.

BMW-2er-Active-Tourer-1200x800-f6e30be8f90413a9

BMW-2er-Active-Tourer-1200x800-4f410fde29129488

This video shows the Skoda backing into a space barely 60cm bigger than the car’s length: http://youtu.be/B0HaWBqyp9w

 

 

Dual Antenna Heading

Thought I’d post up a short bit about one of our new features recently released in the VBOX 3i firmware.

If you’re using one of our twin-antenna 3i SL units, up to now the heading you get has come from the main antenna when testing ADAS applications. Heading is really important, because it’s an integral part of the calculations for separation and contact points. The trouble with a single-point heading is that there is no output when the vehicle is stationary, and it can be a bit ‘noisy’ during low-dynamic testing.

_MGL0627_OSS
With dual antenna heading you don’t need such a large separation distance between them, so might not need one of these roof poles.

So we’ve now incorporated twin-antenna heading. This means that you get an output at zero velocity; you can use a smaller antenna separation between antennas (might not need to use a roof pole any longer); it doesn’t ‘drift’ like an IMU-derived heading can; and the more stable output works well with steering robots.

And speaking of steering robots, a quick plug for our friends at Anthony Best Dynamics. I’ve been down to see them a few times recently to work on this improved heading so that they can configure one of their steering units, for use in autonomous path-following. We’ve nearly cracked it… watch this space or send me an email.

GPS in Automotive Testing: how did we get here?

The road to acceptance of GPS as the principle method of vehicle test and validation was not an easy one

It will probably never be known whether the US government fully understood the ramifications of flicking the switch on GPS selective availability in May 2000. The descrambling of the publicly available (but not particularly useful) signal meant that the positional accuracy of GPS receivers became ten times better overnight. The number of GPS users at the turn of the century was approximately four million: now there probably isn’t anyone living in the developed world that doesn’t use it. Exponential growth, to put it mildly.

Such a dramatic change in technological infrastructure happens extremely rarely, especially one which costs nothing.

At the time, vehicle dynamics testing generally relied on fifth wheels, radar, microwaves, and optical sensors – which all had various disadvantages. A fifth wheel was often difficult to fit to a vehicle body and was easily damaged; microwave was very sensitive to ride height and suffered measurement noise; radar was accurate but only when measuring speed from a static position aimed at a specific point; optical sensors really didn’t work in the rain. It might have been headline news at the time, but the change to the GPS signal forced through by the Clinton administration went largely unnoticed within the automotive testing fraternity, where scepticism surrounding the advantages of GPS remained firmly in place.

2000

When we first came out with VBOX products, it was to a sector which wanted better testing methods and equipment, but which took some considerable persuasion that GPS was the way forward. Some of this was probably caused by the press release issued on the day that selective availability was removed:

“Even with SA turned off, GPS alone will not meet all users needs. For users with higher accuracy, availability, and integrity requirements – such as commercial airlines, ships navigating within harbors, railroads performing precise train control, precision farmers and miners, and surveyors – GPS will still need to be augmented locally with high-fidelity error correction systems based on differential GPS (DGPS) technology.”

Whilst this was and remains true in certain circumstances, the fact that speed, acceleration, and distance can be measured to quite remarkable levels of accuracy without DGPS correction was initially met with blank non acceptance.

2001-Fisrt-VBOX

A New Era in Automotive Testing

The first VBOX recorded GPS samples at 20Hz – a feature which itself raised a few eyebrows even amongst the forward thinkers, who assumed it was interpolated from 1Hz. The truth was that it contained one of the very first survey-grade GPS engines, capturing Doppler-derived velocity at a true twenty times a second. By 2004 we were selling a growing range of loggers and had launched the 100Hz VBOXIII, a big step forward in precision testing. A reliable DGPS base station solution fixed the positional accuracy, too. By this point, most test departments were finding that their fifth wheels and radar sensors were no longer required, and vehicle development using GPS was firmly established. A large step in automotive technological advancement had begun.

GPS Simulation

Well now here’s a thing. I don’t get very involved with LabSat, our GPS simulator, because I am busy enough already. But it seems that some of my automotive customers have started to put it to good use in developing their nav systems. Specifically, in places where GPS is, well, pretty useless.

This video explains the issue. Say you need to test your latest navigation unit – which these days are clever things with integrated gyros to provide for some rudimentary dead-reckoning when the satellite lock is interrupted. The Tokyo highway is one of the most extreme examples of this, because they build roads on top of roads (it’s a crowded city) so you end up driving through what is essentially an urban tunnel. How on earth do you go and validate your gyro-enhanced satnav in an environment like that? Impossible… also difficult in places that aren’t quite as built up but where skyscrapers create urban ‘canyons’ that create the same, if slightly lessened, problem.

Well my clever-boffin colleagues have come up with a turntable, connected to the LabSat GPS simulator, which replicates the rotation the nav unit would be put through on your chosen route.

 

Simulating Real-world GPS and Vehicle Data with LabSat Turntable Solution from Racelogic on Vimeo.

The Next Step in ADAS Testing

Autonomous driving is a popular topic at the moment, with various announcements from manufacturers who are confident that their cars will be doing the driving for us within a surprisingly short time frame. Some predictions state 2020 as the year by which the technology and infrastructure will be in place to make this a reality.

Whether or not this is feasible, there is no denying the leaps in technology that will, ultimately, deliver personal transport in which every vehicle occupant is a passenger. It could be argued that autonomous driving is already here: cars can now park and brake independently; Tesla have recently demonstrated their auto-pilot feature which controls speed based on traffic sign recognition; Volvo assure us that in five to ten years, their commercial vehicles will be constantly scanning every pedestrian, cyclist, and item of roadside furniture within the truck’s vicinity, then acting upon what it sees.

The combination of radar, lidar, GPS, and vision will ensure the current methods of advanced driver assistance will eventually morph into the actual drivers themselves.

Developing, then testing, and finally validating these systems has given rise to an automotive testing market that has seen dramatic upheaval in the last decade and it means that whilst we maintain a core competency in vehicle dynamics testing, recent years have seen significant development in providing test and validation solutions for ADAS. A recent update (firmware version 2.1) to the VBOX 3i SL-RTK introduces some interesting new features.

Testing Vision Enhancements

Traffic sign recognition systems are due to be greatly improved thanks to a new generation of high-definition cameras with better range than the current VGA resolution units. The new systems need to be tested for this greater range and a higher number of potential recognition markers, so the VBOX Multiple Static Points application allows for up to one hundred such targets to be surveyed, creating a GPS map of their locations.

Many_Signs

The desired minimum and maximum detection angle and distance for these targets are then set, and the vehicle driven along the route. When the points fall within the detection zone, the range, angle, and time-to-collision parameters of up to five of them are simultaneously displayed and logged, with further targets being tracked as the closer ones are passed by. The GPS data is then compared to the performance of the system under test.

 

Read the rest of this article here.

Integrated Inertial Measurements

It is largely accepted that the most accurate and cost effective way to carry out vehicle dynamics testing and development is with the use of GPS equipment. In some situations, however, it pays to also use inertial measurements to either augment the available data, or to overcome problems encountered in certain environments – such as a test track that isn’t in an entirely open area, or one which suffers from an obstructed view to open skies in some areas.

VBox3i_IMU04_RLCAB120

We’ve been supplying IMU integration with the VBOX3i data logger for a few years, but now we’ve got a new Inertial Measurement Unit which has a much higher performance due to tighter GPS synchronisation, better calibration and a higher sample rate.

Integrated Speed Measurements

Firstly, the new IMU04 has improved levels of velocity smoothing to cope with interruptions to the GPS signal, thanks to a new real time Kalman Filter which blends the inertial data with that of the GPS system. Although most test tracks are located in open areas, bridges over the test track and overhanging trees can lead to signal degradation– this is where the integration with the IMU signals ensures that GPS dropouts are eliminated.

KFspeed
In this example you can see that the GPS velocity trace (blue) drops outs underneath a bridge. The IMU integrated data (red trace) gives a much cleaner and un-interrupted signal.

For instance, carrying out manoeuvres such as a coast-down test (the results of which are critical to a manufacturer from a homologation perspective, as they form a significant proportion of their calculations of fuel consumption data released to the public) can be completed on a course that takes the vehicle under a bridge, which would not have been possible using GPS alone. IMU integration ensures a consistent velocity log is maintained during the momentary loss of satellite signal lock.

Indeed, there are some facilities that are not ideal places in which to conduct other procedures such as high-dynamic brake stops – they may be tree-lined, but are still used because they’re conveniently located. Integrated Inertial and GPS measurements allow for this convenience to be fully exploited because it can provide the same level of accuracy normally only possible from testing in an entirely open arena.

KFGoogleEarth
Interrupted GPS signal. The blue trace is GPS Position only; the red trace is the corrected position data obtained via integration of the accelerometer data from the IMU and GPS signals.

This is an extract from an article originally featured in Automotive Testing Technology – for the full content click here.

Collision Course

Crash reconstruction and investigation isn’t a nice job, but it has to be done. In many western countries, the number of fatalities is falling even as traffic volumes grow, due in part to studying the cause and effect of traffic accidents.This in turn has influenced the creation of the safety systems now routinely fitted to passenger vehicles.This experience should aid countries that are only now becoming heavily motorized.

There has been something of a revolution in the effectiveness of Collision Investigation Units in recent years, thanks to the uptake of superior apparatus for use in accident reconstruction and simulation.

31527190_m

Case Study: Collision Investigation Unit, Cumbria

Every police force in the UK has a dedicated Collision Investigation Unit (CIU) and the Cumbrian division were the first to adopt VBOX technology – principally using a four-camera, 20Hz, Video VBOX Pro – in almost every reconstruction they carry out.

A common requirement when conducting accident reconstruction is the measurement of probable reaction time. In order to ascertain if a witness account is truthful (speed being the most commonly contentious issue) the Video VBOX will be used to measure distance to impact, determining if there was sufficient space – and therefore time – to avoid a collision. By highlighting markings on the road and tracing the route taken, the evidence can be presented in a starkly clear fashion to a jury tasked with making a judgement on the case. Shaun McKeown, of the Cumbrian CIU, explains:

“It’s easy enough to measure a reaction-time distance, but it’s quite another to get a jury of ordinary people to understand the point you’re making. We always present data, but they don’t like looking at graphs and equations. So we configure the Video VBOX with the main view showing the path of the vehicle, with the picture-in-picture trained on the ground on which we have painted highlights of skid marks or gouges in the road surface. The overlaid graphics then display the distance to impact, and when this type of evidence is played they get it immediately.”

The simplicity with which complex data can be displayed is the Video VBOX’s main strength: graphically-enhanced footage is overwhelmingly more effective at conveying information than a dry explanation of diagrams – plus it actually gives the viewer a direct connection to the environment in which the collision occurred.

 

The text above is an extract from an article in Automotive Testing Technology magazine, and the full content is available here.