The Hidden Variables: Challenges in Measurements and Instrumentation

Recently I came across a youtube video by Scott Manley discussing the failure of the Japanese Lunar Mission — Hakuto-R lunar lander. The primary reason for the failure after the investigation came out to be a problem with Sensor Fusion. While on the fly-by, the lander came across a crater 3km deep. This caused the Altimeter radar reading to drop drastically. The sensor fusion system thought this was a bug and the sensor has started malfunctioning thus started ignoring the sensor reading in its calculation and using the fallback method instead. This caused the lander to assume it has reached the landing altitude before it actually reached there commanding the engines to throttle down eventually causing a crash landing.

Similar problems were also experienced in various earlier space programs like Beresheet Lander by SpaceIL (Israel) and Vikram Lander by ISRO (India). While this could have been avoided by doing simulations before launch, it raises a few bigger questions.

How do you use and trust sensors in an unknown environment which you know very less about?

Sensors whether electronic or otherwise basically convert a form of physical quantity to something which can be measured and recorded. Later on, post-processing is done on this data to extrapolate the actual measurement in terms of physical units.

Let’s take a simple example of temperature sensor. A thermocouple uses two different materials stuck together to measure temperature. When a temperature change occurs, these material expand differently causing mechanical stress to develop in the materials resulting in electronic voltage which can be processed and measured. Based on the pre-calibration where each voltage is measured for a known temperature, we can determine or extrapolate the temperature later based on the measured voltage.

Now, even though this is a very simple example, consider the below limitations.

1. What if there are other unknown variables affecting the voltage developed apart from temperature? The real world is not perfect, what if light or radio frequency or sound or gravity affects the voltage developed which is not known? In a controlled environment, we might be able to get rid of some of these variables and establish causality, but what happens if this sensor finds a new out-of-bound variable especially when it goes into space and feels weightlessness or darkness, etc which it has never experienced in our labs?
2. What if we just can’t extrapolate the data for extreme conditions? Maybe the sensor just doesn’t behave linearly or in a known fashion when it experiences extreme temperatures.
3. What if the sensor produces the correct data but the post-processing system ignores it thinking it is malfunctioning similar to what happened in the above-discussed missions?

I know many of the above are solved problems and great minds have been thinking and working to isolate and study these phenomena, but many times, these are taken for granted when designing a system if not studied in detail resulting in costlier further learnings for future missions.

How do you conduct research relating to causality when you can’t control or tune down variables?

Often Engineer’s job consists of finding and controlling variables affecting other variables. Establishing causality is an important part of this job. If you’re a Software Engineer, you’ll comment out a code or run it step by step in order to debug the source of the issue, if you’re lucky, you’ll find and squash the bug, if not, you’ll spend hours/days/months/years in suspense. If you’re an Electronics Engineer, you will disconnect a part of the circuit and take readings using your instruments which keep in mind are just sensors and can malfunction if the limitations are unknown. Similarly, many engineering fields will have some form of way in which you can cut down on the variables and find the root cause and establish causality for the stuff you’re trying to study.

But what about natural sciences? How do you turn off the gravity of a body to see how it’d affect another moving body? How do you turn off a star so that its light’s wavelength doesn’t affect the object you’re trying to observe? How do you study the atmosphere of a planet if the lander which you’re using to land there itself pollutes it by burning elements which don’t belong there? Not just astronomy, how do you study the effect of exposure to social media over 30 years if you can’t cut off a group of individuals from their daily routine? or how the planet would have evolved if there wasn’t a sun.

What does measurement even mean?

When we say the boiling point of water is 100 degree Celsius, it is an incomplete statement. The complete statement would be boiling point of water is 100 degree Celsius at 1 atmospheric pressure if the water is free from any contaminants. The radius of the sun is 695700 kilometers, under what wavelength of light? The height of the atmosphere is 10,000 kilometers. How? Does the air just stop existing after that limit? The height of Mt. Everest is 8,848 meters above sea level, but the sea level changes every time of day, during tides, due to rainfall and evaporations and glaciers melting, what does that measurement mean?

How do you even define a measurement system? Earlier unit of length “foot” used to be derived from the actual size of the foot of the king, what happens once he dies? pound, a unit of weight, equal to 16 ounces, 7,000 grains but the grains are not uniform.

Slowly we transitioned to physical constants used to define units and we have still not completed that transaction. Now, a meter is defined as the length of the path traveled by light in a vacuum during a time interval of 1/299,792,458 of a second. Whereas a second is defined by taking the fixed numerical value of the cesium frequency ∆νCs, the unperturbed ground-state hyperfine transition frequency of the cesium-133 atom, to be 9,192,631,770 when expressed in the unit Hz, which is equal to s−1. and so on.

Whenever we are defining any measurements, there is n number of preconditions and assumptions we conveniently skip. While this is okay for day-to-day conversations, when going in-depth and extremes, these are important factors to consider.

Conclusion

Questions like the ones above are what many minds ask every day and have been asking for eons for humanity to stand where we are. Employing various techniques and learnings from the past, standing on the top of the giants, we fail to appreciate the knowledge we have learned through centuries. While most of us might not need to employ these tools in our day-to-day execution, it is just interesting to wonder about the ingenuity of these. Countless variables and dependencies go into even building something very simple. There is a wonderful video by A.J. Jacobs where he starts his journey to thank people responsible for his coffee, let’s apply the same to science and appreciate known and unknown faces who bought us to the level of understanding we have today and contribute whatever we can so the next generation has something more to thank for.

Credits and References

And countless videos and articles I lost track of.