Hi @Pushkar_S ,
Welcome to this Community!
There are quite a lot of things that you should consider when it comes to selecting an IMU for your robotic project. [Disclaimer: These are from my opinion / point of view, some people may or may not agree with all the points.]
The first ting you need to consider is, if you can attach the IMU on your robot in the center of your robot. This makes things really easy without having to worry about x-y offsets. Attaching the IMU at an offset in your robot model can bring additive errors in your IMU readings. So you should calculate offsets as accurate as possible.
Second, make sure that you have enough space between your IMU location on your robot and your motor wheels. This is quite important because, if the motors quite close to your IMU, you might have electro-magnetic interference that will add more noise to your digital IMU (analog IMUs are quite pricey and bulky, therefor you want/need only digital IMUs). If you don’t have enough space between your IMU location and the motors, you should consider using a Faraday shield/cage around your IMU’s PCB. Faraday shield is just a metal wire mesh around something that is susceptible to EM interference.
Third, you need to select an IMU that has less noise factor. Every IMU manufacturer will provide datasheets of their chips / chipsets. You need to specifically look at the noise standard deviation measure on the specifications.
If possible, try to buy IMU breakout boards from any online store like SparkFun or Adafruit or any Robot kit sellers. Unless you compare different models by hand yourself, you cannot know which one is best for your use case.
For your purpose, a 6-DoF IMU should be sufficient. But when your are looking for IMUs and find a 9-DoF IMUs priced only a bit higher than 6-DoF IMUs, then go for the 9-DoF ones. Choose MEMS based IMUs.
Lastly, from my experience working with IMUs, I would recommend the models that use BNO055 or LSM6DSOX + LIS3MDL based IMU’s. I have also worked with TDK’s MPU 6050/6150/9250 and they were very noisy and requires a lot of software filtering. I have not tried any other models.
I hope you find this information useful!
Regards,
Girish