Weeble is a cute 2 wheeled self-balancing robot made with FTC parts and a fully 3D printed body. Weeble uses a Linear Quadratic Regulator (LQR) controller to balance and drive around!

It took a lot of work to fit the many features of Weeble into such a compact fully 3d printed cute robot let alone get him to balance and drive so stable with the slow 20ms loop times of the Control Hub and being 7.5 pounds! Hope you like him!

• Fully 3d Printed Body
• Linear Quadratic Regulator (LQR) controller to balance and drive around the 7.5 pound Weeble!
• A 1080P Camera embedded in the right eye, used for computer vision (openCV)
• Object detection and tracking with embedded camera in right eye
• A head that can move up, down, left and right super quickly
• 2 massive adorable eyes, that both move side to side
• Moveable eyebrows to express complex emotions
• 2 cute arms powered by servos
• 2 epic 100mm Wheels powered by 312 rpm motors, high torque because Weeble is heavy!
• 3 2M TOF distance sensors for autonomous driving and obstacle detection
• Efficient LQR controller and 12V battery means long 30+ min battery life.

• Robot Reveal Video
• Onshape CAD Assembly
• How Does Weeble Balance?
  • Linear Quadratic Regulator (LQR) Controller
  • About LQR Controller
  • State-Space Model
  • Optimal Control Law
• What is Weeble Made Of?
• How Does Weeble Work?
• Why Was Weeble Made?

Watch the video below to see Weeble in action! The video shows the progressions of Weeble from the first version to the current version. Also, the video is better with sound!

Weeble - Robot Reveal (YouTube)

Follow the link below to view the full CAD assembly of Weeble. The assembly includes every part of Weeble, from the big 3d printed parts to the smallest screws!

Weeble balances using a Linear Quadratic Regulator (LQR) Controller. The previous versions of Weeble used a PID controller, but the LQR controller is much more efficient and stable. Also, with the slow 20ms loop speed of the REV Control Hub, the PID was unable to be effective enough. Below are the key java classes that handle keeping Weeble balanced (github links):

• **LQRController.java ** - The LQR controller class. It calculates the optimal gains and calculates the optimal control input for the motors.

• **GyroDrive.java ** - The class that handles the balancing of Weeble. It uses the LQR controller class and the states of Weeble.

The LQR controller uses the state-space model of the robot to calculate the optimal gains to keep Weeble balanced. The state-space model is a mathematical model that represents the behavior of a physical system as the set of input, output, and state variables. The gains are used with the state of the robot ( angle, angular velocity, position, and velocity) to keep the robot balanced and driving. Future improvements will be made to make the LQR controller PID-assisted to improve velocity control.

Below are the two most important components of the LQR controller, there are many more equations and components that make up the LQR controller but these are the most important and easiest to understand.

The dynamics of the system are modeled as a set of linear differential equations:

ẋ = Ax + Bu

• x is the state vector (angle, angular velocity, position, velocity)
• u is the control input vector (motor power)
• A is the state matrix (how state evolves over time)
• B is the input matrix (how control input affects state)

The control input u is computed using the feedback law:

u = -Kx

• K is the feedback gain matrix (computed using LQR)
• x is the state vector (angle, angular velocity, position, velocity)
• u is the control input vector (motor power to keep the robot balanced)

• Fully 3D printed body
• 2x 100mm rubber wheels with 3D printed hubs
• 2x 312 RPM GoBilda DC motors with encoders (Wheels)
• 1x REV Control Hub (Brain)
• 2x REV Servos (Arms)
• 1x Axon Mini+ Servo (Neck)
• 4x SG90S Servos (Head, Eyes, Eyebrows)
• 3x REV 2M Distance Sensors (Front, Left, Right)
• 1x Innomaker Arducam (Embedded in Right Eye, OpenCV)
• 1x 12V Battery (Power)

All 3 versions of Weeble were fully designed in OnShape and 3D printed parts were made with a Prusa Mini+ 3D printer.

Onshape CAD Assembly

Weeble is programmed in Java using the FTC SDK as he uses a leftover REV Control Hub from the 2024 FTC season. He also uses OpenCV to track objects and faces. The robot is controlled wireless with a Logitech Gamepad over a WiFi connection to the REV Control Hub.

I made Weeble as a challenge to myself to see if I could make a balancing robot. I also wanted to make a robot that was fun with a personality, hence the moving head, eyes, and eyebrows. Now that Weeble is built, I plan to use him as a platform to work on computer vision projects, point cloud mapping, and autonomous driving.