Project Overview
To balance on a ball you need to be able to move it in any direction, to remain on top of it. This project uses robot omni wheels, which can be driven in one direction whilst rolling freely at right-angles. Three wheels angled at 120° to each other can therefore provide movement in any direction. Gyroscopes and accelerometers are used to sense the slightest imbalance, which triggers the control system to drive each motor to respond to the error. The 2.4 GHz Wi-Fi link is used with a Wii Nunchuk to control the robot, and to convey dynamic data which can be displayed as graphs on a PC screen for detailed analysis, whilst the robot is moving. Click the image on the right to watch a video of my BallBot in action-->

All of the robot chassis components were produced on my 3-D printer and modelled using the free 3-D design package from RS Components, DesignSpark Mechanical. For this design I also constructed an assembly of the complete robot, to ensure that all of the parts aligned. This included simple models of all bought components like the motors and H-bridge driver pcb’s. Given the limitations of 3-D printing, the chassis consists of a number of parts screwed together using nylon countersunk screws. The top plate has compartments for the two batteries and apertures for socket strips to take the micro and Wi-Fi transceiver.

The lower body consists of two plates to provide a strong platform on which to mount the motorized legs, which are adjustable to suit a range of ball sizes. You will need small drills and metric taps to be able to put threads in all of the holes in this design.

The robot is powered from two 18650 Lithium batteries, delivering 7.4v to the system. An On/OFF switch is mounted in the centre, adjacent to a push buton switch which gives you the ability to select modes of operation. On power-up the robot should be on a level surface, as it immediately enters into a calibration routine, which trims out any offsets in the accelerometers and removes gyro drift.

Red and green LEDs are also mounted in the top face, which provide moving patterns of light to indicate what the robot is doing and to indicate when the robot is being moved towards vertical as its self-balancing trigger point. The graphing app, I mentioned earlier, is also used to tune the robots PID controller using sliders presented on a PC screen. These adjustments are sent to the robot over the Wi-Fi link and their effect can be seen instantly. So if you choose a ball which is larger than the one I have used, you should still be able to find sets that achieve a stable self-balancing condition.

The circuit diagram for the robot is shown here on the left, with the Arduino connected to the three H-bridge drivers, providing high current PWM  for the motors, the MPU 6050A 3-axis gyroscope and 2.4GHz transceiver, and the six LEDs, which are connected back to back and multiplexed in code to save on wiring. As the motors are only rated up to 6v I have used diodes to drop the battery voltage fed to the drivers. A simple resistor network is used to subdivide the supply voltage so that the Arduino can measure the voltage remaining in the batteries, which should not be allowed to fall below 3 volts per battery.

The wireless transceiver command pin is connected to the NANO so that the code can set up the device using AT+ command codes. By default it is set to 9600 baud, but needs to run at 115200 for PID tuning and graphing data.

I developed the combined PID controller and graphing tool app using Processing. This took quite a lot of effort, but the end result makes it very easy to change PID controller setting on the fly, as the robot is moving around, simply by moving sliders on the PC screen. You can temporarily save a set of settings you think are close to what you need, make further changes, then recall them if you want them back

You can switch back and forth between controller adjustment and the graphing tool, so you can also see the effect on data sent from the robot. The tool allows you to access different sets of data from the NANO, just by clicking on a number field. The graph can be dynamically adjusted automatically, showing min/max values or set to specific ranges, and if the graph is paused you can inspect the data points simply by moving the mouse pointer over the graph, to draw on the underlying data.