Vehicle Safety - Science topic

Nice question

Also, you need to findout the Hashin Damage model parameters for the axle material from the previously published experimental studies.

If the cars fulfill safety requiremenst like NCAP, also small cars can be pretty safe. In car to car crashes of course the heavier car has advantages.

Focus also on ESP - it help to save more live than a just bigger car.

It is worth checking the experience gained in large naturalistic driving studies (NDS), for example SHRP2 (https://insight.shrp2nds.us/documents/shrp2_background.pdf). The issue is not how to collect data but how to process it afterwards. One has to define and apply various (semi)automatic triggers and also link them to the environment characteristics.

Dear Mr. Altamash Jawad

Dont go for MATLAB to simulate the adhoc network scenarios especially VANETs. Best you can think about a combination of NS3 and SUMO. It is the best network simulator and traffic simulator that is available presently. 

Regards

Interesting question, which I might answer in a complete different way.

Why should a human be the teacher?

We all know physics but we do not apply them by mathematics while we are driving. We are relating to experience but I've never seen a Driver pulling out his calculus if he should accelerate or brake. The machine on the other Hand may calculate easily if it does fit into the space or not, if it knows the own velocity and the Speed of oncoming car as well as the car in front. It further knows an approximated mean acceleration of the own car. So basically it's just math which is required and math does not rely on experience otherwise we would not use it.

Certainly one thing is problematic. Friction coefficients are a good example for this. A wet or even an icy road has other coefficients than a dry one. However the Controller could be programmed in such a way that if the car is wet the worst coefficient is assumed. The detection of these Parameters may not be easy but could be communicated by V2V or X2V Communication.

In the end. It's like the autopilot for aircrafts. It is working on maths, physics, statistics, memory size and computing power. While a human may not learn everything related to driving experience until the end of his life the machine may explicitly know every situation derived from physics and statistics. No teacher actually needed because the controller does not experience emotions and age as humans do. It is just math.

If you go for the ethics. Worst case Scenario someone jumps in front of your car. you could pull away and hit another Person or drive against a wall with same injury severity for yourself. This question you should ask first yourself. If you find an appropriate answer I'll listen to it.

A recent Austroads study authored by my colleague at ARRB:

Look at the Nationalk Highway Traffic Safety Administration's regulatory impact analysis on tire pressure monitoring systems at  http://www.nhtsa.gov/DOT/NHTSA/Rulemaking/Rules/Associated%20Files/TPMS-2005-FMVSS-No138.pdf

Especially look at the analysis of impact on crashes.

100 car study have done many studies regarding to driver behaviour see link below please this is a sample you can find more studies like this related to 100-car study 

Hi Stephan,

Thank you for your Suggestion.

I think we are in same track and you understnad me. I am interested to make all safety requirements for a concrete application. I have looked in the Internet for example considering the ISO 26262 for safety requirement. Constraints is my consideration is only in the ECU part not the Overall System. Then I consider the ECUs Item for System Level consideration. The SGs has been considered for FSR and then considering the functionality TSR and HSI requirements has been creatied.

The note you have mentioned SEooC, I looked in the Standard Part 10 and we are using hte AUTOSAR Software component.

1. Do SEooC effect in the requirement list? (i think not).

2. What does Allocation of element means? (componet Name where the function has been done!)

3. How I can diferentiate the Hardware Software Interface requirement with the Hardware safety requirement and Software safety requirment?

4. How/ what is the defination of the T_lifetime ? I mean for truck/ vehicle is it  relevant to use 1 year? because i took it as Assumtion.

I found Google Scholar is helpful. Thank you again.

Best regards

Eva

Thank you all

If you look at the websites of for example Autoliv or TRW, companies which design and make car safety systems, you will find descriptions of various radar, laser and visual sensing systems which can be used to automatically brake or steer a car. Thus a car can control the headway between itself and a vehicle in front, or brake when it senses that a car, a pedestrian or some other object is too close. Completely automatically driven cars are at an advanced state of development and will be in production by 2020.

The maximum deceleration rate which can be achieved will depend upon the capability of the braking system, the traction available between the tyres and road, the mass distribution of the vehicle, and the position of the centre of mass relative to the front and rear axles. A good introduction to this subject is available in "Fundamentals of Vehicle Dynamics' by Tom Gillespie.

When you say 'safe' rate of acceleration, do you mean while maintaining control? If so then the threshold would occur when the brakes on one axle or the other lock. In that case rear lock-up is the more dangerous scenario as it will typically lead to oversteer whereas front lock-up would lead to understeer which is undesirable but more stable.

Normally all the test are done at room temperature without any preheating or cooling.

The following link have some information and a model already available in matlab..hope this will help you...

You can start research by

The Present Serviceability Rating (PSR) is a subjective, ride-based observation on the roughness of the pavement to estimate deterioration, deficiencies, and needed improvements based on early AASHO Road Tests. The higher the PSR value, the smoother the riding surface. So PSR is closely related to PSI.

Good work and good luck

In Europe, speed limiters are mandatory, and there are specific speed limits that the drivers must comply with, according to the type of roads. The itinerary must be provided to the law enforcement.

All heavy vehicles (for passengers and goods) must be equiped with a tachograph (analoge or digital) wich allows the law enforcement authorities to track the driving and rest periods, as well as the speed.

European laws are quite strict when it comes to avoid fatigue in driving so resting periods are compulsory and the company must keep the record of the data for control purposes. The driver must be able to provide the authorities the data of the last 28 days so that he can prove that he hasn´t been working more than he should.

The fines are heavy for the companies but, as in almost everything, the companies still find a way to get around the rule of law.

There are some polish available on the market, which are based on the lotus effect. These nano-structure coatings result a reduced friction coefficient on the windscreen surface, so that the rain drops are blown away by the air flow caused by the vehicle speed.

e.g.:

Kind regards,

Mario

If you are talkling about taking yearly fatality totals from say 28 countries taken at random from a UN list of countries. I do not think it can be done. The reason is that some of them are still on an upward trend (Algeria, China, Brazil...), but not all: some seem to have had their absolute maximum in fatalities.

So to build a joint estimation model (with some explanatory variables, I assume) on some that are going up and others that are going down would require knowing when and why they had their maximum. (I realize ARIMA will fit anything, but fitting is not explaining: it is self-referencing, explaining a variable by itself and various own and error lags). The problem of substance here is that we do not have the slightest idea of why the maximum occurs.

Consider only OECD countries now. About 18 of them had their absolute maximum of fatalities in 1972 or 1973 (give or take a year). The reason for this cluster is unknown, I argue in the enclosed piece (available on this site). So how can one meaningfully forecast the future if one cannot meaningfully explain obvious past maxima? Will there be an upturn for those that are on a downward trend? Will there be a minimum, an inverted Meadow? The question cannot be answered if we cannot make sense of the timing of past maxima. This would appear to deserve some work....

Finally, consider only current EU members. Last time I looked, it seemed that all of them were past their absolute maximum (but some just...barely). Some peaked in 1989 (like Spain), most before (in 1972-1973) and others at other dates (Greece, Portugal...)..  If that is so, they are arguably all past their maximum and on a downward path, but certainly with different slopes and ups and downs that are correlated, as are the residuals, on the downward trend.

Marc Gaudry (sent from Paris where I am currently working)

Using Matlab you can define it as follows:

%<<<<<<< definition of ISO 2631 limits >>>>>>>>>>>>>

%reduced physical efficiency limit (8h)

xu=[0.71 0.63 0.56 0.5 0.45 0.4 0.355 0.315 0.315 0.315 0.315 0.4 0.5 0.63 0.8 1.0 1.25 1.6 2 2.5 3.15];

%exposure limits (8h)

xsz=2*xu;

%reduced comfort limit (8h)

xk=xu/3.15;

%frequency vector

cz=[0.8 1.0 1.25 1.6 2.0 2.5 3.15 4.0 5.0 6.3 8.0 10.0 12.5 16.0 20.0 25.0 31.5 40.0 50.0 63.0 80.0];

figure;

loglog(cz,xsz,'r');

xlabel('Frequency [Hz]');

ylabel('RMS value of the vertical acceleration [m/s^2]');

hold on;

loglog(cz,xu,'m');

loglog(cz,xk,'g');

%loglog(ff,P1); here you own vector can be ploted

aac=legend('exposure limits (8h)','reduced physical efficiency limit (8h)','reduced comfort limit (8h)');

grid on;

zoom on;

That sounds complicated. I would rather use a double layered steel with air in the middle - so better cover the chassis with a thin metal sheet and some spacers to keep a distance. Rockwool will just pick up the mud and water (if you put it outside). Inside there is a certain risk that your outside steel (which covers load) will fail.

Good luck

When discussing correlations you have to keep in mind that correlations do not necessarily imply causal relations. Often, they are a consequence of a common influence by an ignored third factor such as properties of the infrastructure (many or few freeways), regional structure (predominantly urban or countryside), cultural properties,  (liberal or strict) legislation, (sloppy or stringent) control/enforcement. For example, in countries with liberal speed regulations (e.g., 75 or even 85 mph allowed on some roads) and sloppy enforcement (hardly any speed limit monitoring), few speed tickets are issued, i.e., the statistical data show little speeding. Nevertheless, in an objective sense, there may be many drivers driving too fast for the given situation resulting in comparatively many accidents.  This results in a paradoxical correlation with an unexpected sign between speeding and accidents.  This is a variant of the classical Simpson's paradox: The more firefighters are sent to the scene, the more damage the fire will  cause.

Dear Mahdi, I suppose that crash prediction models are what you are looking for. They relate crash frequency to various explanatory variables, which is exposure (traffic volume) and geometrical characteristics, such as median, etc. Since crash frequency is a count variable with non-normal distribution, they are usually developed using generalized linear models with negative binomial distribution. This is also what I am using; however for the work of real guru look for example here - https://www.researchgate.net/publication/239438865_Statistical_Road_Safety_Modeling

Dear Ioannis, you ask very important question. There are various perspectives and indicators. As you mention, crashes are the typical indicator. But are they the best one? It depends on crash frequency, time frame adopted, extent of underreporting... - how many crashes do you typically have in your conditions? Is the frequency sufficient for the analysis you need? If not, other indicators may be used, based on more frequent events, such as for example traffic conflicts.

Nevertheless AADT is usually the most important condition, this is why it is used as exposure indicator - and risk is basically count of events related to exposure. Some good reading - http://www.swov.nl/rapport/R-2002-12.pdf

Regarding section length - there are discussions about it. If you need risk related to some conditions, then these conditions should dictate how are section formed (segmentation). See for example here - http://docs.trb.org/prp/13-4372.pdf

Some sensor fusion between the camera system and radar would be necessary as camera systems can only function to a certain speed before they become unreliable (e.g. Mobileye and Eyesight by Subaru). The camera can be used for object detection (vehicle, moped, pedestrian), the radar can compensate for the short comings of cameras at higher speeds by assisting in detecting object size, and relative change in object size to determine if the time gap between the subject and lead vehicle is low enough to require a change in speed. 

Your ACC algorithm would have to have a lot of additional constraints factored into it, given the variability of driving/roadway conditions as you well know, but also where that system would be used most often and the driving behavior of individuals in those regions (e.g. lane splitting by motorcyclists and moped riders would have to be taken into account). 

I hope this helped. 

VW used these bushes in the Passat and later Golf.

Under lateral forces the curve outer side of the axle is moving to the front of the car, the inner side backwards. This reduces the oversteering behaviour.

Have a look at the picture.

using data logger

I recommend KAYABA and GPS to collect the data real time via NI and also record the vdo while collecting data

 

Hi Denis,

I believe that a driving simulator could help you.

We're using it to measure the effect of information on drivers' behaviour, and we can implement different scenarios for different level of information. Similarly, you should implement different scenarios for different situations.

Good work

Mauro

Dear Dariusz.

You can find at least two products with several options and prices (http://www.smarteye.se/ and http://www.seeingmachines.com/). I am not sure about http://www.noldus.com but you can try it.

You talk about driver fatigue detection but there are other works related with drowsiness and distractions. They are different but related problems.

In my humble opinion, there are several important facts in this problem.

Experts in traffic safety are needed to provide ground truth information.

Based on our research in automatic detection of driver distractions, to obtain a robust risky signal, you will need to take into account information about road, vehicle and driver (we combine all of them using a fusion method). We used a camera to detect driver´s hands activity. In some experiments we also considered driver eye information. At the end, driver behavior is a very complex problem and a multi-disciplinary approach is required (in our case: Traffic Safety and Technology experts).

Hope this helps,

Enrique

There's quite a threat evolving for cars in general. Right now it isn't really fiancially interesting to attack vehicles, but this may change in the future. Besides that there's always the risks of attacks that are designed to disrupt traffic (or crash a vehicle, in the extreme case). However, it should be remaked that this isn't specific to AI-controlled vehicles: this also applies to other connectivity like bluetooth, wifi, tire pressure sensors and pretty much anything that communicates through the car's CAN bus. For examples, see the following resources, which include IEEE S&P and USENIX SEC publications:

Once the attacker controls the CAN bus, it seems extremely obtruse to attack the AI system instead of the engine or the breaks, if you want to do real damage. Yes, the AI system should be protected from attacks, but it isn't exactly the most interesting target. For driverless cars, the insurance is much more interesting -- who is liable when the car crashes? In any case, I would say that current non-regulated use of driverless cars is not a good idea, also because the technology is still maturing.