Tuesday, January 13, 2015

Child Safety Seat Using the KEA-64 to Prevent Heat Stroke

This past summer has been somewhat of a sad one. With this past summer not being particularly hot there was still an ongoing media presents of infant fatalities related to heat stroke due to these unfortunate children being left in cars. This past year alone in the USA there were approximately 30 fatalities due to heat stroke[1]. For some people and even the car companies this sounds like a small number and a number that does not warrant the need to invest any engineering resources, but for each of those children and parents or for someone who needs to deal with the aftermath, there may be a different view. For this reason I started thinking what would be an easy and simple way to help these children.

This graphic SIMULATED video gives an awfully real feeling of what a child goes through as they experience this horrible ordeal.

Issue & Resolution
It is known that the greenhouse effect is the major cause for a car heating to well above the outside temperatures. This happens when sunlight passes through the a window and strikes an object inside the car converting the sunlight to a longer wavelength that cannot escape thus increasing the temperature in the car.

Figure 1. Greenhouse Effect Causes a Car to Heat Up Quickly

A seemingly simple solution would then seem to be to break thermal seal allowing for the heat to escape and reducing the internal temperature. If the windows of the car could be opened then the temperature would not increase as fast and may even decrease. The objective behind this solution was to find an immediate low cost low complexity solution. Allowing strangers to help also seemed a reasonable option. This assumption was made due to the realization that in most cases where children were saved from extreme harm and even death, it was often a stranger that helped, so why not start with that as a part of the solution.

There have been other attempts to resolve this issue but they are usually complex solutions requiring a decent amount of infrastructure. As car companies are dealing in the millions of units, anything that adds even a 1¢ quickly adds to $10,000 over the manufacture of the product. While this may seem trivial, to a CEO and shareholders this is money no longer in their pockets and therefore lost revenue [2]. For this reason alone it is worth keeping the system to minimal cost input and try use systems already available in a car.

System Hardware
Freescale recently released their KEA family of ARM processors. These microcontrollers are made for automotive applications and therefore have specs that allow them to operate in the temperature range of -40 ℃ to 125 ℃ (-40 ℉ to 257 ℉). Using the KEA64 as a starting point a very simple and basic prototype was developed to see how well it performs, if the basics could be proven to work and show its reliability and safety than it would be worth attempting to incorporate such a system in an actual car.

Figure 2. TRK-KEA64 Development Board

Keeping to simplicity and minimalism the demo system is composed of only 4 components. These components are the KEA64, an ADT7420 and for the demo a stepper motor controlled by a ULN2003. Missing from the demo are pressure sensors used for child detection. The system operation follows a simple linear flow removing potential logic confusion or code errors.

Some extra hardware was used in this demo, but the basics of the system is already found in most modern cars today. While some extra coding may be needed, the overall system could be implemented with no additional hardware added by the car company. The addition of pressure sensors as well as communication from the car seat to the car could be added to the cost of the car seat removing this from automotive production cost. This no extra cost method may allow for more serious consideration and implementation by a manufacturer.

System Operational Model
The program in the KEA64 is interrupt driven allowing for simplicity as well as maximum power consumption. The flow of the program is outlined in figure 3 below.

Figure 3. System Flow Diagram

The system is initialized when the car is powered up (change of battery) or when the back door is opened. This is done to ensure whenever a child is in the back seat the system is in a known state. During initialization the ADT7420 is set to sample once per second and interrupt on a high temperature. Once initialization is completed the system goes into a low power state. The system will remain in this state until either the back door is opened at which point the system is reset or until a high temperature interrupt is fired.

If a a temperature interrupt is fired the system wakes to start doing a situation assessment. The first step taken is to check the car's internal temperature. This is done to prevent the system activating based on a false interrupt. Should the interrupt in fact be false the system is put back in power saving mode to once again await a temperature interrupt.

However, should it the system verify the cars internal temperature is high the system will then attempt to check if a child is present in the car. This is done by reading three pressure sensors fitted in the child's car seat. The sensors would be fitted to the bottom, the back and shoulder section of the seat. If a child is in fact in the seat and not something else 2 of the 3 sensors should indicate so. Using this method of two out of three the system would then assume a child is present  and move to the next step. This would involve opening the windows and monitoring the car for a reduction in temperature, if the car does not begin to cool the horn and flashers could be activated in an abnormal fashion to attract the attention of other people in the vicinity.

Figure 4. Locations of Pressure Sensors in Child Seat

If instead it is found that although the car is at a high temperature but there is no child present the system would disable the temperature interrupt and go back into a low power mode. The disabling of the interrupt is to prevent further false alarms until a child is possibly placed in the back seat. This would be determined by one of the rear doors opening. Once opened the system would once again reset itself starting the monitoring cycle anew.

Code for the above system can be found here

Initial Concerns and Solutions
When starting of the project one major issue that was a recurring concern among people consulted for this project was theft. There appeared to be a concern that if someone could somehow heat the car to the critical point, the windows would open and the car could be easily broken into. For this reason as well as others the need for sensors in the child seat to detect whether it is in fact a child present is crucial. In the initial attempt it has been assumed that a three sensor system would be sufficient. By placing a sensor in the seat, back and shoulder parts of the child seat it would be possible to determine if a child or something else is in the seat.

Another concerns was power consumption. While it is often assumed a car has a large battery that could not possibly drain easily this thinking needed to be reconsidered. The first issue pointed out was the other systems using power, each device added could easily tip the consumption scale causing  unsustainable power consumption . Another scenario is a car left for a long duration without being driven. In this case the battery has no chance to recharge and therefore if a car was left for several months the battery would be dead. For this reason the system is put into a low power state to reduce power consumption. The KEA64 when in run mode uses 6.7 mA (~2.25 Ah over 2 weeks) and 5.5 mA in WAIT mode (~1.85 Ah over 2 weeks). If the KEA64 is however in STOP mode the KEA64 only draws 145 µA (~48.7 mAh over 2 weeks) even with the ADT7420 sampling at 1 sps which has a current draw of only 65 µA (~21.5 mAh over 2 weeks) the total over a two week vacation would be 70.2mAh.

It is for this reason if an interrupt is fired and a child is deemed to be absent the interrupts are disabled. This prevents the system needing to wake each time the temperature gets above a set value when it is already known a child is not present.

System Testing
The system was tested at various stages during the project as well as upon completion. During the project a hair dryer was used to simulate high temperatures. While the whole system was not subjected to these elevated temperatures this method was sufficient to test the functionality of the system. Included in this testing was successive cycling to ensure such cases as a car being cooled by AC and than left to heat up and then cooled again would not negatively affect the system.

Upon completing this first stage the system was tested in thermal chamber. The chamber was set to 65 ℃ although the critical temperature is set to 50 ℃. This was done to speed up the testing since while the chamber may register 65 ℃ the air inside takes a while longer to reach the set temperature. Figure 5 below shows the thermal chamber as well as the system inside.

Figure 5. Thermal Chamber with System inside

Figure 6. System in Thermal Chamber Resting on Cardboard to Insulate the Two Systems

The full system test can be seen in the video below. As expected when the system reaches the critical temperature the “window” rolls down. The video shows this test being performed once but the actual test was completed several times to ensure repeatability.

While not shown in the video, this test was performed a few times in close succession again to ensure the system could handle a modest cycling of the surrounding temperature.

Figure 7. ULN2003 Arlington Array Used to Control the Stepper Motor

System Issues (KEA64)
During testing, with both the hair dryer as well as in the thermal chamber, it was noticed that the KEA64 had an unusual and unexpected behaviour at some points. This was evidenced by the LEDs turning on when nothing had happened. With the hair dryer this was witnessed when the hair dryer was initially turned on but not near the unit. In the oven, as can be seen in the video below, it happened when the oven was turned on.

It was not possible to reliably reproduce the issue with the hair dryer, however, with the thermal chamber it was reliably reproduced three times as seen in the above video. Each time the issue was reproduced, it was only when the chamber was turned on, having an initial large power draw, but never when the chamber was being turned off. The few times it did occur with the hair dryer it was also only when turning on the hair dryer.

While further testing needs to be conducted to better understand the issue, there are a few similarities between the two situations which leads to some initial thoughts. In both cases the power outlet is shared by the KEA64 system and either the hair dryer or thermal chamber. With the hair dryer the KEA64 was powered from a laptop that was sharing the same outlet and with the Thermal chamber the USB power supply was plugged in the same outlet as can be seen in figure 8. In the case of the hair dryer and the laptop it may be though that the power supply as well as the USB subsystem would shield the KEA64 from any negative power effects seen at the outlet. In the second case, with the thermal chamber and the USB charger, again it would be assumed that the charger would provide some form or isolation.  In both cases without proper testing the isolation can not be be guaranteed, perhaps in later testing a battery pack could be used to ensure complete isolation from the power mains.

Figure 8. KEA (Top Right) Plugged into Same Socket as the Thermal Chamber (Bottom Left)

These symptoms have led to the assumption that it is a power issue of some sort and related to the KEA64 and not the ADT7420 generating false interrupts. This is assumed because the motor does not run as it should if it was an interrupt generated falsely by the ADT7420. It is therefore assumed the KEA64 is susceptible to something either EMI or some power disturbance that is causing some bits to be flipped.

Figure 9. Power Supply Used to Power the System During Thermal Chamber Testing

Figure 10. LEDs on KEA64 Lit Showing Some Issue & the Breadboard Isolated from the Chamber with Cardboard

Moving Forward
The next steps for this project would be to attempt to control the windows in a car via the CAN bus. As these commands are not published this will require some reverse engineering. The CAN bus will be accessed via the OBD port as this is the the easiest and most convenient access point but this may provide other issues. Some ODB ports do not allow all messages to be transmitted to the CAN bus but rather only those needed for diagnostics. Another issue is controlling the windows when the car is off. In this case it would be needed to turn on the electronic subsystems of the car in order to power the windows. Each one of the issues pose their own challenges and will hopefully be dealt with to accomplish the end goal of this project.

Original post on Element14 can be found here