Summarizing my AGV work this winter

This was my first winter with AGV. Work on the bot from my side was complete. At least that was what I thought. I thought I had implemented the control systems of the robot by tuning the two PID loops and making it run smoothly. And then my senior Ayush approached me for performing an in depth analysis for the control systems. And it was then that I realized that I couldn’t have been more wrong.

The work commenced with me learning about state space forms and the interconversions. That was something rudimentary and was dealt in a day or so. We derived upon the basic equations of transfer functions of gearless DC motors and their applications. The study indeed was quite helpful in building concepts related to control systems. I partnered with Anurag for the same and it was good for both of us for learning the core concepts of control systems, including brushing up the methods of linear algebra.

Then the main job of optimizing the controls kicked in. It was then when I went on to taking real world data and optimizing the overall methods for the same.  I started off with the already awesome tuning of the Roboteq closed loop motor driver system for the steering column of Eklavya 4.0. I took both open loop and closed loop data and tried to get a suitable relation for the plant transfer function for the motor and plot the corresponding PID plot. Though the experiments were not that successful, subsequent usage of MATLAB tools made a good estimate for a closed loop transfer function that fit our bill.

Next we moved on to the velocity control of the BLDC motor up front. The agenda was clear, make custom ROS nodes to provide custom input voltages to the BLDC motor driver. And then plotting responses after filtering out the noise using the inbuilt filters, comparing their outputs and hence estimating their transfer function as a whole.

Overall the second task was fairly successful, and barring a few constraint evaluations, the TF we got was fairly accurate as evident from the similarity of the expected and real graphs for the tuned PID constants.

Winters 2015 were an amazing learning experience in this regard. I’m looking forward to work on more research orientation in this field, i.e. Control Systems. Cheers to the oncoming 2016!

As a footnote, Here is the link to the Documentation I had created as a part of the work in a PDF format.

Leave a Reply

Your email address will not be published. Required fields are marked *