Intro to Inertial Measurement Units (IMU)

hey guys thanks for stopping by microengineering my name is michael rona and i thought i'd make a video talking all about inertial measurement units or imuse because right now i'm working on developing a quadcopter flight controller completely from scratch and i'm officially naming it the hummingbird flight control unit so i'm doing everything completely on my own from the sensor selection to the pcb design to flight software development i'm doing everything on my own and since the inertial measurement unit or imu is arguably the most important sensor on board my drone i thought i'd make a video maybe a couple of videos talking all about it i apologize that this video is a bit technical in nature but i'm going to try my best to explain things at an easy to understand level and hopefully end up learning a thing or two about inertial measurement units there's a lot to do so let's jump into things okay all right so an inertial measurement unit or imu is a combination of sensors that measure orientation and motion with respect to an inertial reference frame and before we get into talking about those sensors i think it'd be a good idea to understand what i mean by an inertial reference frame so this inertial reference frame is located at the center of earth and earth is actually rotating about this inertial reference frame so hopefully this analogy helps a little understand it a little bit so right now i am standing still in my bedroom but due to earth's rotation my position with respect to the inertial reference frame is actually changing and so therefore i currently have some constant velocity with respect to the inertial reference frame and for all of you nerds out there who are in the know about this this inertial reference frame i'm referring to is called the earth centered inertial frame or eci frame so now that you know what the inertial term means in inertial measurement unit let's get into talking about what sensors are actually in an imu so when you go online to like adafruit or amazon or robot shop and look up imu you're going to come across two different varieties of imews you're going to come across a either a six degree of freedom imu or a nine degree or feed a minute a six stop imu consists of a three axis gyroscope and a three axis axis accelerometer and then a 9. imu adds on a 3-axis magnetometer to those other two sensors let's start off by talking about the gyroscope because the gyroscope is arguably the most important sensor that's going to be on board our drone and we use the gyroscope to measure orientation angles and so you know if we want to measure our orientation really accurately really precisely we're going to need a really really good gyroscope and so gyroscopes are basically angular rotation rate sensors they measure how fast our drone is rotating in the x y and z directions in radians per second or degrees per second you guys should be using radians not degrees radians are better and what we do with these angular rotation rates is we integrate them to get angles now unfortunately this direct integration is not practical due to what's called gyro bias and gyro drift and i think the best way to explain these two things are going to be from looking at some graphs so let's pull some up quick so if we take our gyro sensor and set it on our desk and keep it completely still and take measurements for a little while we'll get a plot that looks like this one the gyro rate isn't changing much you know the radiance per second number but as you can see the gyro is not exactly zero the gyro is reading a non-zero angular velocity even though it's completely stationary on the desk and we call this offset gyro bias or gyro offset and so therefore gyro biases are constant offsets in our gyro rate measurements and this is a quality found in all gyroscope sensors whether it be on board a tiny little drone like this or the gyroscope found in a submarine they're all going to have gyro biases and if we were to integrate our raw angular rate measurements to compute angles we'll get a plot that looks like this one and as you can see although the sensor is completely stationary on our desk the computed angle is still increasing and this drift in integrated gyro measurements is called gyro drift and since our raw gyro rate measurements had a non-zero bias our gyro angle measurements began to drift away from true and so therefore gyro bias causes gyro drift very very important quality of gyros fortunately there are ways to correct for this gyro bias and gyro drift and the way i'm going to be doing it on board my flight controller is by using what's called an extended kalman filter so i'll be using my column filter to estimate gyro biases and correct for them on the fly in real time and next i'd like to talk about the second most important sensor on board our drone and that is going to be the accelerometer and accelerometers simply measure accelerations in the x y and z directions unfortunately they're a little bit more difficult to use than gyros mainly because they are inertial sensors so let's pull up the mathematical model that describes accelerometer measurements so this icky equation you're looking at is the mathematical model that describes accelerometer measurements i know it's a little bit complex but i label the terms for you a little bit easier so not only do accelerometers measure linear accelerations they also measure gravitational acceleration centripetal acceleration and coriolis acceleration the first term in this equation is a combination of linear accelerations and the acceleration due to gravity the second term is the coriolis acceleration term which as you can see depends on earth's rotation and our velocity and the third term is the centripetal acceleration term which depends on again earth's rotation and our position and for precision applications you definitely need to take the centripetal and coriolis accelerations into account but for hobbyists like you and i you know just making a simple little drone the sensors we have access to are not sensitive enough to measure these terms so we can make an engineering approximation fancy word and ignore those terms simplifying our accelerometer measurement model to one that looks like this much nicer right and also since we know that gravity points downwards whoa big shocker there uh we can use accelerometers to measure our tilt angles with respect to gravity so you know how level our drone is and we call these angle these tilt angles roll and pitch angles denoted by phi and theta respectively and if you recall from your high school physics classes i know that might be a little while ago if we integrate accelerations we can get velocities and if we integrate velocities we can get positions but unfortunately just like gyroscopes this direct integration of our accelerometer measurements is not practical they're going to drift away just like gyroscope measurements can and so yeah accelerometers also have biases that we need to somehow account for i'm going to be using an extended kalman filter to estimate these accelerometer biases on the fly and help correct accelerometer measurements back to true and then also we can supplement our accelerometer velocity and position estimates with gps readings to kind of like help correct those uh velocity and position estimates back to like normal through sensor fusion and again calm and filtering but that's really complex you know that's for another video and also i should note accelerometer sensors are very subject to vibrations and noise accelerometers are very noisy sensors so it's pretty common to apply a low-pass filter to accelerometer measurements to help reduce some of the noise a little bit but do not filter your gyroscope measurements that's not good to do accelerometers you can you can get away with filtering them a little bit to help reduce noise but do not filter your gyroscope measurements so hopefully that explanation wasn't too complex and it gave you a little bit of an idea how we're going to use accelerometers on board our drone and so now i like to move on to talking about how imus are classified in general so let me show you this uh this table right here now you can feel free to pause the video and examine this table a little bit closer for yourself but imus are typically classified by their sensor biases and how much their navigation solutions drift over time and so you know different applications are going to require different grades of imus the lowest grade of imus like the ones you and i are gonna be using in our projects we'll have accelerometer biases around point zero one g's or a little bit more if they're not as good and general bias and gyrobioses near 100 degrees per hour so that means our gyro measurements are going to drift like 100 degrees per hour and then so as you can see the highest end imus found in things like submarines and spacecraft are going to have accelerometer biases zero zero zero zero one g's and gyro biases of less than a tenth of a degree per day that is such a huge that's such a huge range that's insane that's actually insane that's really cool all right so that was probably a lot of info thrown at you i don't blame you but hopefully that gave you a bit a better understanding of how the imu is going to be used on board our flight controller and you might be thinking to yourself well didn't you say you're going to use a gyroscope to measure orientation angles and then are we also going to be measuring orientation some orientation angles with the accelerometer well yeah we are and you might also ask yourself well for measuring the same thing with two different sensors is there some way to combine them to make them a little bit better and to that i say yeah definitely that's what the whole that's what the whole field of sensor fusion studies is how to fuse sensor measurements together to output a better estimate of some sort of uh measurement parameter so yeah definitely i'm going to be using an extended common filter to fuse gps gyroscope and accelerometer data together to form like a better estimate of like positions and velocities um orientation you know all of those different what are called state variables it's pretty complicating and i definitely want to make a video talking about the sensor fusion algorithms and approach i'm going to be using to estimate my drone state variables so stay tuned for those videos those are definitely going to be a bit on the technical side because common filters aren't exactly one plus one equals two math um hopefully i learned a thing or two about accelerometers and gyroscopes and imus and uh yeah i'm really looking forward to sharing my progress uh on this drone flight controller development project i guess until the next video see you later [Music]

Channel: MicWro Engr

Views: 34,020

Keywords: imu, inertial measurement unit, imu fusion, inertial navigation system, ins, ins fusion

Id: LjeFZetmfYc

Length: 12min 16sec (736 seconds)

Published: Sat Jan 23 2021