Querying streaming sensor data on the Arduino board

Question

I am new to Arduino and this forum, so please forgive me if I am asking something really trivial and stupid.

I have an Arduino board which collects sensor readings from an MPU6050 sensor (acceleration+gyroscope). The next step for me would be to process these data directly on the board. I want to somehow store N last sensor values and then query this set (collection, vector, array, whatever you call it) to fetch certain values. This set will be constantly updated with new values, whereas older values will be pushed out, so that only the most up-to-date values are stored in the set. For example, I want to know if the acceleration exceeded 10 m/s in the last minute.

In the literature this kind of set with most recent values is sometimes called window - this approach allows limiting an unbounded sequence of constantly generated tuples to a bounded bag and then apply traditional SQL-like query operators. For example, Oracle has come with CQL - an engine and a query language for handling streaming data.

Before I start implementing this, I am wondering if there is a similar (and, of course, much simpler) library for Arduino? If not, which data structure should I use to implement such functionality?

Also, my intention is to query not just individual values, but certain patterns and sequences - e.g., I want to know if the acceleration first decreased and then increased again (something similar to the WHERE condition of an SQL query). Does it look realistic to implement with Arduino?

Answer 1

It sounds like you are after my Average library:

• https://github.com/MajenkoLibraries/Average

It allows you to store a rolling set of data where you get to "push" values into it and the oldest one drops off the end. (It's implemented internally using a circular buffer).

You can then do various forms of analysis on the data - get the mean average, the mode average, calculate the standard deviation, find the maximum and minimum values of the set, and even access individual entries within the set.

It's a fully templated library, so you can use it with any numerical data type (float, double, int, long, etc).