Sensor for detecting cyclists / URM06 ultrasonic sensor

Hello forum!

Recently, my team partnered with our local traffic authority. We want to achieve more safety for cyclists in road traffic. As we found out, a big problem are cyclists which are not recognized by turning cars. So, our goal is to set up two posts with one Arduino for each one of them. One post has a warning lamp, the other one a sensor. As the sensor detects a passing cyclist, the warning lamp lights and warns turning cars.

Ideally, we want to use Arduino Nanos, communicating via NRF-modules. Now for my question:

We are not really sure which sensor we should use. We thought about various types of sensors and I want to explain what we need and we found out:

Our requirements

  • ability to measure distance
  • range between 20cm - 4m
  • accuracy of ±5cm
  • ability to detect regardless of color/material (everything a cyclist could be made of)
  • ability to withstand different weather conditions (it’s meant to be outside the whole year)
  • rapid refresh rate (cyclists can go fast)

1. radar sensors
Not really an option as it is hard to measure distance

2. infrared sensors
Cheap or selfmade sensors are way to inaccurate. May be affected by high sun exposure/lit materials. We already contacted Terabee which manufactures high-accuracy infrared sensors, their sensors could match our requirements but they are not very well compatible with Arduinos and expensive.

3. ultrasonic sensors
Slower refresh rate than infrared sensors. The cheap common Arduino ultrasonic sensor may be triggered by raindrops. This sensor also needed a fairly flat surface for reflection which we cannot guarantee in daily use.
Now I found out about the URM06 sensor sold by Arduino. The specs seem to match our requirements pretty well but I’m not entirely sure about the following points:

  • is it weatherproof?
  • is it triggered by raindrops?
  • will a passing cyclist trigger the sensor in every case?
  • will it withstand continuous operation, like 27/7 over several weeks/months?

That’s for my questions.
Looking forward to hear what you think about it. Would be happy if I got some response from people with experience working with sensors in general or the URM06 in particular.
You don’t have to answer every question asked, I’m happy for every help I can get!

Thank you very much!

ammoniumperchlorate:
accuracy of ±5cm

What is the reason for needing this level of accuracy?

You want to create a sensor that can detect a cyclist under all conditions for a few dollars, while even humans and systems worth hundreds of dollars struggle with detection of random objects.

Additionally, I think adding a sensor would endanger more cyclists than it would protect. After a short initial period, car drivers would get used to the sensor and rely on it. I think it would be better to create a signal that reminds drivers to look for cyclists regardless.

ammoniumperchlorate:
We are not really sure which sensor we should use.

The human eye has been known to be reliable. All that is needed is to ensure better connection to the human brain. This sounds very much like somebody wanting to avoid doing their job properly by trying to re-invent the wheel instead - and landing up with a square one.

Klaus_K:
You want to create a sensor that can detect a cyclist under all conditions for a few dollars, while even humans and systems worth hundreds of dollars struggle with detection of random objects.

Additionally, I think adding a sensor would endanger more cyclists than it would protect. After a short initial period, car drivers would get used to the sensor and rely on it. I think it would be better to create a signal that reminds drivers to look for cyclists regardless.

I may have put that the wrong way: We don't want to use a cheap sensor. We are willing to invest money of the sensor matches our requirements. Maybe I misunderstand your point there but we don't want our sensor to tell what the passing thing was, as we can ensure that it was a cyclist by adjusting the range of the sensor.

For your second point: We also thought about a reminder first, but as we found out there was already a sign reminding drivers to look for cyclists. This sign didn't help at all as the city still recorded several accidents there including one being deadly for the cyclist. Furthermore, this is just a pilot project to find out about things you mentioned, e.g. drivers getting too used to the device. We are not sure at all if this idea will work but that's why we want to try it out.

pert:
What is the reason for needing this level of accuracy?

As we only want to detect cyclists on the bicycle path in front of the sensor but not the cars passing behind them, we need a high accuracy to ensure that the trigger range matches the bicycle path.

My wife, frequently, bikes to work. She has had the issue of nearby or approaching cars not seeing her.

I, originally, tried ultra-sonic to find that the entire ultrasonic process of send a signal, receive a signal, and process to be to slow.

I then tried with a TFMini, which works great in darkness and bright daylight, though less sensitive on the brightest of days.

The box, mounted on the back of her bike, sends out a pulse, reads the reflected distance, and lights up some LEDs, based on distance to car. Runs off batteries.

Nick_Pyner:
The human eye has been known to be reliable. All that is needed is a proper is to ensure better connection to the human brain. This sounds very much like somebody wanting to avoid doing their job properly by trying to re-invent the wheel instead - and landing up with a square one.

As much as I like your metaphor, it doesn't quite fit our case. It might be not clear enough in my original post but the cyclists are not recognized by the cars because they are hidden behind cars coming from the other direction. So it is not an isse of drivers not paying attention but a rather dangerous construction of the crossing.

Idahowalker:
My wife, frequently, bikes to work. She has had the issue of nearby or approaching cars not seeing her.

I, originally, tried ultra-sonic to find that the entire ultrasonic process of send a signal, receive a signal, and process to be to slow.

I then tried with a TFMini, which works great in darkness and bright daylight, though less sensitive on the brightest of days.

The box, mounted on the back of her bike, sends out a pulse, reads the reflected distance, and lights up some LEDs, based on distance to car. Runs off batteries.

Cool, sounds like a very interesting project!

Did you find any downsides of using the TFMini, like inproper detection or measurements during "extreme" weather conditions or exposure to non-reflective material? Is it weatherproof by itself or is it completely housed inside the box?

Of course the car detection degrades with rain and fog but she does not bike ride during those conditions.

Unlike a SR04, the TFMini is very accurate. I previously used the TFMini as an imaging device in my home, I now use a TFMiniPlus. The TFMiniPlus operates better than the Mini. I weatherproofed the Mini, the TFMiniPlus comes in a weatherproof case.

The reliance safety issues , I see as a problem .
Arduino fails- light doesn’t come on- car runs over cyclist . Who is at fault ?

Idahowalker:
Of course the car detection degrades with rain and fog but she does not bike ride during those conditions.

Unlike a SR04, the TFMini is very accurate. I previously used the TFMini as an imaging device in my home, I now use a TFMiniPlus. The TFMiniPlus operates better than the Mini. I weatherproofed the Mini, the TFMiniPlus comes in a weatherproof case.

Good to know, thank you very much!

ammoniumperchlorate:
We don’t want to use a cheap sensor. We are willing to invest money of the sensor matches our requirements. … we don’t want our sensor to tell what the passing thing was, as we can ensure that it was a cyclist by adjusting the range of the sensor.

With sensor I mean the whole system including the Arduino. Its a cheap system compared to AI detection system build by engineers to detect traffic. And these are still struggling.
Cyclists are random objects. They do not stick out from the environment like cars which are much more uniform (big junk of metal which can be detected by ground loops).

I think a solution without a sensor would be better. I agree with hammy.

I am in favor of you using a low-tech solution. A pneumatic tube sensor that the bicycle/unicycle must ride over. Replace the tube on a scheduled basis. A pressure sensor can detect the pressure change.

Paul