Udana Jayasekara

All mechatronics work

Self-balancing robot

Context
Control Systems project, RMIT
Team
Team project
Discipline
Mechatronics, Embedded
The built prototype on the bench
The built prototype on the bench

A two-wheeled inverted pendulum, modelled and simulated in MATLAB/Simulink, then built and tuned until it balanced on the bench.

A two-wheeled robot that balances itself like an inverted pendulum. It uses geared DC motors on an L298N driver, an MPU-6050 gyro and accelerometer, an Arduino Uno and two 3.7 V lithium cells on a stacked plate frame.

What I did

  • Derived and linearised the system model, obtained the transfer function and compared PD and PID controllers against the specified transient response in MATLAB/Simulink.
  • Built the prototype and tuned the controller on hardware, then analysed why measured behaviour diverged from simulation.
  • Added a detachable 3D-printed limiter (the red part) that catches the robot when it falls, so it is not damaged during tuning.

What went wrong, and the fix

  1. Gains calculated from theory (Kp 705, Ki 70, Kd 19) did not work on the real robot, because of friction and errors in the model parameters.

    Retuned on the bench to Kp 23, Ki 90, Kd 1.8, after which the robot balanced reliably.

From the bench

  • SolidWorks model of the stacked-plate frame

  • Battery pack, IMU and the red fall limiter

  • Geared DC motor and wheel

  • Arduino Uno and L298N wiring

  • Detachable 3D-printed limiter

SolidWorks model of the stacked-plate frame
SolidWorks model of the stacked-plate frame1 of 5