A Raspberry Pi robot arm with inverse-kinematic teleoperation and autonomous colour-based pick-and-place, developed for the QUT Robotics Club's 2018 workshop series.
The project shipped as two things: a working arm, and a progressive curriculum that took club members from servo control through kinematics to a computer-vision pipeline. It was modelled on QUT's Introduction to Robotics unit, and funded through the Student Clubs and Projects (SCAP) Fund.
Status: archived. Built in 2018 and not maintained. The Python 2 scripts and the OpenCV version it targets are both long superseded; treat this as a teaching reference rather than a current codebase.
Kinematics half. A position-based inverse-kinematic controller that drives the tool point along a trajectory from keyboard input, with USB gamepad teleoperation as an alternative. Link lengths in the DH model mirror the laser-cut structure.
Vision half. An autonomous picking routine: the arm locates a coloured ball with a single Pi Camera, plans a trajectory from a set of reference points defined in its own frame, picks the ball up and places it in a box.
The camera is uncalibrated and the vision pipeline is colour-threshold based, so this is a demonstration of the full sense-plan-act loop rather than an accurate manipulation system.
- Raspberry Pi 3 B+ with a NOOBS SD card
- Raspberry Pi Camera
- SparkFun Pi Servo HAT (PCA9685, over I²C)
- Hobby King servos
- Laser-cut acrylic structure — CorelDraw files and a full parts list are in this repository
Enable I²C and the camera in raspi-config before starting.
# keyboard capture and I2C
pip install getch
sudo apt-get install python-smbus
# vision half
sudo apt-get install python-opencv python-matplotlib
# gamepad teleoperation
pip install evdevTeleoperation (Python 3), in Teleoperation_Demo_Python3:
python3 RobotArmRaspberryPi_Complete_Code.pyAutonomous vision picking (Python 2), in VisionArm_Demo_Python2:
python VisionArm.pyGamepad teleoperation (Python 2):
ls /dev/input # find which event the gamepad is; event3 by default
python RobotArm_Gamepad_teleop.pyServo calibration, in Robot Arm control Workshops:
python3 calibration.pyVision thresholding, in Computer_vision_Workshops:
python test_video_blobdetection.pyThe teaching material is the part of this repository most likely to be useful to someone else, and it is all here:
- Workshop slides with the underlying theory
- Python scripts for each workshop task, separate from the final demo scripts, so a session can build up incrementally
- A MATLAB inverse-kinematics simulator (
RobotArmSimulator.m) for working through the kinematics away from the hardware - Laser-cut files and parts list for building the arm itself
The progression runs servo control → forward and inverse kinematics → trajectory following → computer vision → autonomous pick-and-place, with each stage runnable on its own.
MIT — see LICENSE.
Built for the QUT Robotics Club 2018 workshop series, during my term as club president. The mechanical design was a collaboration; I led the final integrated design, and wrote the control software, kinematics, vision pipeline and workshop curriculum, and sourced the funding.
- Krishan Rana — CAD and laser-cut design from the end-effector through to the first joint
- Marty — debugging assistance on the computer-vision task
- Student Clubs and Projects (SCAP) Fund, 2018 — project funding