# How to detect if a nearby parked car has started engine?

We have a project with a machine servicing a car parked next to it. It's comparable to a car washning machine.

Before the machine is started, we need to be 100% sure that the engine is off. Also, if we detect that the engine is turned on during the process (which takes severa minutes), the machine should withdraw its operation immediall, to avoid damages.

The machine is supposed to refuel the car without any manual action by the driver.

Electrical cars are not a part of our current plan. But it certainly is relevant that some cars stop the engine when the brake is down.

Our problem is to detect whether the car parked at the machine is running or not. We have already been thinking of these solutions:

1. Microphone listening for engine sound. Problem: In a noisy environment, we might have too many false alerts, when a big truck is parking close nearby.

2. IR camera looking for hot exhaust gas. Problem: We have tried that and can't see any significant changes whether the engine is on or not

3. Camera detecting whether the engine lights are on. Problem: Not all cars have lights on when engine is running. Not all cars require the driver to push the brake to start the engine.

What other options do you see for us?

• Do people not consent to this process? I've never seen someone run into a car wash and try to start their car mid suds. Nov 3 '17 at 4:05
• Have a key slot where the ignition key has to be entered before the machine will start ... of course someone could bring a second key... this is what the real problem is " design a system that is foolproof : but fools are so ingenious..." Nov 3 '17 at 10:21
• You were prepared to use cameras, but not to just throw down some rubber sensor pads? Baffling. :)
– user98663
Nov 3 '17 at 10:53
• If your machine has a robotic arm to reach for the fuel tank, you can add another to the exhaust. Nov 3 '17 at 14:32
• So, the safety of the operation depends on beeing "100% sure" that the engine is off, and this detection relies only on the technical device that you'll build, and not on human... 100% could be certified only for just a model of car... but there are thousands of brands/models/year... 100% sure for all cars just seems impossible. I liked @Solar Mike 's suggestion, the slot for putting the key... but, besides it's not elegant, there are cars that use RFID keys, and their engines start just with the press of a button... Nov 3 '17 at 16:19

This sounds like an XY problem.

Your problem is you need to prevent damage to your equipment. Your rabbit-hole solution is to assume that you need to retract your equipment when the vehicle begins to drive. You are myopically focused on detecting when the vehicle begins moving.

Instead, I propose you prevent the vehicle from moving. An electric or hydraulic bollard is cheap ($2000-3000) and designed from both a mechanical and visual perspective that a car needs to stay put. Give the user a flashy screen, like a car wash. Show them the status, gallons filled, estimated time remaining. Retract the bollard after your robot's sensitive bits have tucked themselves away, then change the screen to a big friendly "you're done!". • The primary safety concern is the ignition of fuel (by hot exhaust gases), according to the Q's comments, not vehicle movement. – user98663 Nov 3 '17 at 15:50 • Fine, bollard + Halon fogging. :) Nov 3 '17 at 16:19 • @Wossname "The main problem will be the car moving, which will cause damage to the mechanics." - This is the main concern according to the OP, and the one addressed in both Bryan's and my answers. Nov 3 '17 at 17:07 • Fuel ignition can be significantly reduced by vacuuming up the fumes. The fumes will spread near the filling point, and a simple hose with negative pressure will mop up a lot of it. Nov 4 '17 at 10:02 • "Even if the barrier only lets the car drive 10 cm, the damage is done already" This really doesn't make much sense, as fuel caps are at different heights, can be on either side of the car, or even the back, so your fill device must be able to move up/down, fill at 3 discrete orientations; yet you can't figure out how to allow for some play after engaging? Nov 6 '17 at 14:05 Definitely sounds like an XY problem, you don't need to detect whether the engine is running, you need the car not to move. Bryan's idea made me think of one that I think is even better, as there is no risk of damage to the car, if someone were to run into the bollard when it is up. Mount rollers in the ground (parallel to the direction of travel) that are normally locked (preventing them from rolling), allowing them to be driven on. When your apparatus is engaged, unlock the rollers, thus preventing the car from driving away even if the engine is started and the driver attempts to drive. When done, and the device is retracted, lock the rollers again. Most Petrol Cars can be heard on a sensitive AM radio .Most of the electrical noise comes from the spark plugs .You can try using a regular MW AM radio to start experimenting with .If you are close to the car like feet away you should hear something .Better would be a SW radio if you can get your hands on one .It has been said that car ignition noise has a very broad peak around 20 to 40 MHz .If you can hear something from your experiments then you could process the Audio with a DSP .Electric cars do make engine noise too .I can hear my Nissan leaf on my experimental MW AM car radios when I use an external test Antenna and plug them into the Standard 12V Cig lighter plug .The sensitivity of these test car radios is good but not excellent .This means that you wont need some really expensive comms radio to achieve this . I was told in 1974 by my father who was a Civil Engineer who worked for the city council that the traffic light signalling in Dunedin where I grew up detected cars by antenna wires buried under the road .Diesel cars were rare then but large trucks were diesel .I do not know how well thier system worked . • This would only work for petrol engine cars. Diesel enginges don't produce sparks for ignition. Nov 3 '17 at 12:31 • Regular short-wave listener here. Due to EM pollution, I often have to go to far away from buildings to listen to LW/MW/SW. My regulat listening locations are relatively close (20 m or so) from roads. In the last 3 years of regular listening, I never heard a car on shortwave and only one heard a motorcyle. For yor radios, did you test if you're picking up radiated or conducted emmisions? Nov 4 '17 at 10:48 Laser "microphone" vibration detection. Can be done invisibly with IR lasers. All you need to do is point it at a piece of bodywork and look for vibrations in the range of engine RPM with suitable magnitude. You'll need to calibrate it so the vibrations of passengers moving around or loud sound systems don't trigger it. You can also detect vibration at a distance with doppler ultrasound, but I think this is harder. http://ieeexplore.ieee.org/document/1293292/ With electric cars I think this may be impossible; but there may also be ways of wirelessly contacting the car's control electronics to monitor this. Note that existing fuelling stations handle the risk of people driving off with the hose attached - if this does happen it breaks off in a safe-ish manner and you charge the driver's insurance. • We know that the nozzles can handle it. but our robot might get damages. Therefore, we want the robot to withdraw immediately if the engine is activated in the car. Maybe we should look further into vibration detection. IR lasers is not an option since all cars are different and we don't know where to look, but only know the wheel locations and car model (and thereby location of the fuel door) Nov 4 '17 at 20:34 • You don't have to be very specific about location, the whole bodywork vibrates tangibly when the car is running. Nov 4 '17 at 22:18 Exhaust pipe detection requires a Flir or IR THERMAL camera, not an IR security camera. Thermal cameras are far more pricey, use exotic lenses (germanium, chalcogenide, etc.) and they see IR longwave, not the shorter near-visible wavelengths. There are even thermal cameras with narrow filters tuned to CO2, and can see the gas itself rather than the high-temp emissions. "Gas FindIR" product from Flir corp. Cheaper might be a handful of IR thermometers, pointed to various locations where you'd expect to find the exhaust outlet. Experiment with those$12 Harbor Freight keychain thermal IR detectors.

• I have tried with a FLIR One camera and motorized string trimmer, and saw now siginificant difference when I turned the engine on or off. I was only able to see if it was hot or not. But I will look for the CO2 detecting ones. The array of IR thermometers is not an options as we want to support a wide range of cars with very different locations of the exhaust. Nov 4 '17 at 20:31
• Use array, since a single thermometer would only see one model of car. But if FLIR can't see hot exhaust gas, then IR thermometers won't see it either. Ooo, idea: Air pump hose with a carbon Monoxide sensor (and perhaps a fan to enlarge and stir up the exhaust plumes.) Nov 6 '17 at 21:15

OBD

You should be able to find ready to use hardware and possibly software libraries too, or you could write your own.

Make it a prerequisite to have your OBD hardware plugged into your Vehicle. Only when you have established that Vehicle engine is dead, should you start remainder of service process. OBD will also allow you to sense if/when engine comes to life.

And it will cover a vast majority of current vehicles.

I would go with a nitrogen and carbon dioxide detector on the exhaust pipe. The gas will start coming out the pipe the moment the engine starts. Don't count on heat as it might take a couple of second for the whole area to heat up especially on a cold day, same goes for the hat of the gas. If you have a system that finds the fuel opening you wont have any difficulties finding the exhaust pipe.