Welcome to Part 2 of our Self Balancing Bot using 96Boards blog series. In this blog, we are going to see the finished Self Balancing Bot in action… yay :D . You may notice a huge delay in getting the Part 2 out, reason for that is explained below.
In case you missed, here is the quick recap of what happened in previous parts:
Introductory blog - This is the introductory blog for the Self Balancing Bot. Here we introduced the project, BoM and roadmap.
Self Balancing Bot
As I promised, the Self Balancing Bot has been created using 96Boards CE. The detailed instructions including Schematic diagram, step by step instructions on building the code is provided in the 96Boards-projects repository.
But, the bot is not functioning up to the mark :( There are a couple of reasons for that which are listed below:
1. Use of software Sensor Fusion algorithm
As you can see from the previous blog, I have used Complimentary Filter for fusing the Accelerometer and Gyroscope readings together. The issue here is, Accelerometer data can get affected by noise and Gyroscope data tends to drift over time. Since the filter is a software model, accuracy is not that great. Because of these reasons, Bot fails to stabilize on its axis.
2. Lack of Optical Encoder to determine position of motors
Because I have purchased a ready-made chassis and my poor knowledge towards Robotics (this is what a newcomer say :P), the motors I’ve purchased with Encoder doesn’t fit with the Chasis. So, I decided to use general gear motors for this project which prohibits me from precisely reading the position of the bot. This is also one of the reasons why the bot is not able to stabilize properly.
Why huge delay for Part -2 ?
Well, before starting this project I assumed that only PID tuning part will prove to be tough. But, now I learned that for a good Robotics project it is important to learn how to choose components properly. Because I made the wrong selection of components right from the motor to battery charger, my entire project gets delayed.
In between, the sensor MPU6050 also went bad so I ordered one more from Amazon but the received one was faulty (doesn’t have few vias present on board). Then I ordered couple more from the trusted vendor and finally able to make it.
Then, as I expected initially tuning PID controller was also tough. It requires more trial and error approach.
Because of the above-mentioned reasons, it took a while to get the Part-2 out.
As I said, PID controller tuning took more time. So I have prepared the below instructions to help anyone tune PID better as the constants worked for me may not work for you. In the project, PID constants are mentioned here.
- KP - Proportional Constant
- KI - Integral Constant
- KD - Derivative Constant
- Start with zero for all PID constants
- Increase KP until the motors start oscillating. Here KP will provide the necessary torque for the motor to move.
- For the motor controlling project, KI is not needed generally. But a little negative value may prove handy for stabilization.
- Finally, adjust the KD value (should be small) so that the oscillation damp out and the bot balances itself.
Note: PID constants depends on the motor, chassis, sensor and lot other factors. So, the PID value provided in the project will not work straightaway for you. That’s why PID tuning is necessary.
The video demonstration of the Self Balancing bot
Because of the above-discussed reasons, Bot fails to stabilize for a long time. This can be overcome heavily by using on-chip DMP in MPU6050. DMP takes care of fusing accel and gyro reading together hence avoiding drift and noise.
So, I’ll call the bot which uses DMP as Rev 2 and the current implementation as Rev 1. We are also planned to demonstrate the Rev 2 of this bot in Linaro Connect.
We are at the end of Part-2 blog of our self Balancing Bot series. I hope this blog provided some info on my experience in creating the self-balancing bot using 96Boards CE. If you have any idea of how to make this project better or any suggestion over the current implementation, please provide those in comments. We’d love to hear back from the community!
Want more? Continue on to Part - 3 of the series.