Introduction

Hello and Welcome to the final part of our Self Balancing Bot using 96Boards blog series. After hitting several roadblocks, we are finally able to stablize the bot with the help of few tweaks. Before getting into the details, here is the quick recap of what happened in the previous parts:

In Self Balancing Bot using 96Boards series

1. Introductory blog - This is the introductory blog for the Self Balancing Bot. Here we introduced the project, BoM and roadmap.
2. Part - 1 - In this blog, we discussed about interfacing IMU with 96Boards CE. This involves 3D rendering the IMU data using python OpenGL.
3. Part - 2 - Part 2 blog showcased the first revision of the Self Balancing Bot with the help of 96Boards CE doing the processing of IMU data, PID and Arduino doing the motor control.
4. Part 3 - This blog showcased the second revision

Self Balancing Bot - Rev 2

In the first revision of the Balancing bot, I have mentioned a couple of pain points which hindered the stability of the bot. This revision addresses all of those and makes the balancing bot completely balancing ;-)

For making it stabilize, below changes are done:

1. Using the DMP in MPU6050

This is the most important change required to handle the gyroscope drift and accelerometer noise from IMU. In the first part, we used the simple Complimentary filter for fusing the accel and gyro values. But, that means there was no heavy filtering of data done which will eventually have no control over the IMU drift and noise.

In order to overcome this issue, I planned to make use of the DMP (Digital Motion Processor) present in MPU6050. But, the documentation for using DMP is very vague and that will lead to reverse engineer the DMP. But thankfully there was a library exists which makes use of DMP, i2cdevlib written by Jeff Rowberg and it is a collection of C++ based i2c libraries for AVR/Arduino and other MCU’s.

So, I decided to integrate this library with our source code. That helped me to overcome the gyroscope drift and accelerometer noise to provide smooth and steady IMU data. Generally, DMP can be used in two methods: interrupt and polling modes. I used the later which avoids the use of an extra pin for sensing interrupts.

2. Offloading PID control to Arduino

Even though the IMU data is steady, I experienced some issues with my current PID implementation. After trying hard, I reached a point where hosting the PID algorithm in 96Boards CE didn’t provide any benefit. So, I switched to use the PID library in Arduino.

That eliminated the need for decoding the commands from CE board and provided a simple but efficient way to use PID controller.

3. Using MotorController library

In the first revision, I have used some simple functions to control motors. But, since there exists a library for controlling motors through Arduino, so I decided to make use of it.

With the help of above-mentioned software components, I was able to stabilize the bot finally :-)

Building and Programming

All the source code has been uploaded to our 96Boards Project Org. Since this project involves two revisions, I have partitioned the sources into rev_1 and rev_2 which can be found under src/ directory.

Instructions on building and programming the Self Balancing Bot Rev 2 can be found in project’s README.

Conclusion

So, we are at the end of the Self Balancing Bot using 96Boards blog series. This project gave me the first exposure to robotics, which coupled with success and pain, but overall I learnt a good lesson and it turned out to be a good noobie project ;-) But the biggest takeaway for the community is to know how easy is it to develop any applications using 96Boards ecosystem. The combination of the suitable base and Mezzanine board can help anyone to build anything using 96Boards!