ASR research log-log 2

After getting familiar with what I have to use, I start to think about what I do for.Like what I said in the previous log, to build the reliable rocket system isnot only the problem of the reliable hardware construction, it mostly depends on the reliability of the software system, which is the soul of the entire system.At first, we low estimated the importance of such thing, thought that this is not a very serious problem. Like always, things didn’t appear like what wethought. The more function that we added to the system, the more complicated the system became, which lead to the appearance of the unpredictable errors. Irealize that to cut the functional part of the system is not a preferable choice, so I need to reorganize the structure and the process of the systemworking. I worked for one whole night to get a self-testing scheme. In thisscheme, I make two control systems that are placed in the different positions,the ground station system and the rocket control system, and I let them to do a series of verifying of each other. Since both the system are made by us, so we should consider both of them as the unreliable systems before the testing,which invisibly increase the difficulties of the testing. Suffering from the thousands times of failing and debugging, I finally understand one truth: neverbelieve your system is stable before the system is tested by you, no matter how good and reliable the chip seems to be. This truth is also related to the dilemmaI talked about: the chip is more functional means that includes the more complexity----maybe the chip is stable in some settings but the chip will be ascrazy as the mad man after changing a little parameter by our carelessly action.So, killing all the undesirable bugs in the program is not an option while what we need to do is to give the system a way to handle the every possible situationand mistakes. For instance, we need to locate our rocket in order to find it after it has landed. We design several ways to achieve such aim: the GPSlocating, the radio direction finding and smoke sign marking. Each one has its own merits, like the GPS can offer the precise and real-time location datawhile the radio direction finding is much more reliable. Anyone of these methods can lead us to get the exact location that the rocket is, but also anyone ofthese methods can be dysfunctional. So what I did is to combine all of them and make sure that there is at least one way to find the rocket. However, in theother parts of the system, the combination of the different methods will not give the result that we want but it will cause the components to contradictwith each other and lower the succession rate. For instance, the control of theopening of the parachute also has various ways to decide when to give the signal of opening. For instance, I used to think use the attitude calculationto get the real-time acceleration and velocity in the ground coordinate system,which shows us when the rocket has reached its highest point---when the velocity is zero-----and ready for the opening of the parachute. But the precision and the reliability of this method are severely depending on the precisionof the sensor we used and the performance of the offset algorithm we add in the attitude calculation. More detailed speaking, the not good enough sensors willgive the data varied in a large range even they are placed in the rest place. Being constricted by the limit budget, we could only afford the relatively lowerperformance sensors. The offset algorithm I wrote is not capable enough offixing such error. All of the above will make the calculation become much more unpredictable.In the worst case, the wrongly calculated time will cause our rocket to open its parachute when its engine is still working, which will cause the parachute immediately to be tangled and torn by the rocket main body. (An important note: such resultwill only be occurred when we put the parachute in the head of the rocket. I have designed the different plan---like open the parachute from the bottom ofthe rocket----which I will write more details in the following logs.) In the other hand, we can use the timing method to open the parachute, the triggeringtime of time which is fixed in the program. This method is much more reliable than the previous one if the triggering time we calculated is precise enough. Thismethod is so “honest” that it will not change a single bit by itself indifferent situation. We are worried about whether we could give an acceptable time only with the thrust data of the rocket engine working test and the idealaerodynamics simulation. I once have thought about combine two of them in a way of calculating the function of the reliability of each method in differentperiods. For example, the acceleration sensor becomes much more unreliable as the rocket is going up since the vibration which mostly influences the reliabilityof the acceleration sensor becomes much more intense when the engine is keep working and accelerating. So I give the acceleration sensor a decreasing reliabilityfunction as time is going up until the engine is stop working. And since the triggering time that worked out by our self is fixed and will not be affectedby the time, it has one constant reliability value. Therefore, we could use the product or other mathematics result of these reliability functions and theirreal-time data as one parameter in our triggering procedure. This sounds to be a good plan, but the reliability functions are not as ideal as I thought, or eventhere is no kind of such function that we use to evaluate the result of the different unreliable data producers. (This drives me to read and learn about theDPPX---the distributed processing---which also face the problem of how to get aunited result from the system that contains different processors that has a possibility to be broken. The mathematicians have proved that there is no way to give aperfect way to solve this problem while in order to give an approximating result, we need to bear some extents of mistakes to be constantly existed inthe systems. That’s very enlightening and maybe in the future we can use such kind of ideas to solve the complicated sensor systems.) As a result, we give upall the complex plans and choose the fixed time triggering as the final plan. We got the information that the other armature rocket making groups abroad mostlyused the fixed time triggering. This may have taught us a lesson: maybe sometimes the silly-looked-plans may have the higher value.