3-RRR parallel manipulator: a planar robot steered by tilting an IMU
A planar 3-RRR parallel robot whose three servo legs move a platform in x, y and rotation, steered by a hand-held IMU or by ROS 2.
- When
- since 2022
- Origin
- Open-source robotics project and course
- Built with
- ROS 2 Humble
- Gazebo
- ros2_control
- MuJoCo
- Python
- Arduino C++
- MPU6050 IMU
- URDF and xacro
The 3-RRR is a planar parallel robot. Three servo-driven legs hold a triangular platform, and together they move it in x, y and rotation. Tilt a hand-held IMU and the platform follows, or send targets from ROS 2. Without a PC, the Arduino firmware runs the IMU teleoperation on its own.
The repository covers the whole chain. It has a tested maths library with no ROS dependency (kinematics, Jacobians, singularities, workspace, dynamics, trajectories and IMU fusion), a pipeline that turns the CAD model into a ROS robot description, ROS 2 packages for control, Gazebo and launch files, MuJoCo physics of the same model, new Arduino firmware, and twelve lessons that teach the project step by step.
Scenarios that show a real problem
| Scenario | What it shows |
|---|---|
| Gesture teleoperation | Tilt a hand-held IMU to steer the stage, with no joystick or screen |
| Singularity guard | A request to rotate by 40° would cross a parallel singularity, and the guard checks for this before the robot moves |
How it works
The robot has three degrees of freedom: x, y and the rotation of the platform. Its inverse kinematics has a closed-form solution, so the controller turns a target pose into three leg angles without iteration. The forward kinematics is harder, because one set of leg angles can match up to six platform poses, so the control package includes a monitor that keeps track of the real one.
A parallel robot can pass through poses called singularities, where it loses control of the platform. The controller includes a singularity guard that looks at the planned motion first.
The operator’s hand-held controller carries an MPU6050 IMU, and the leg angles can be driven either by the Arduino firmware alone or by ROS 2 on a PC through a serial bridge. The same description file drives both Gazebo and MuJoCo, so a scenario can be tried in physics before any hardware moves.
Accuracy depends on the actuator
In the physics simulation with a hobby-servo model, the position error changes a lot with the actuator, and the maths is the same in all three cases:
| Actuator | Position error |
|---|---|
| Hobby servo | 0.66 mm |
| Hobby servo, finer command resolution | 0.56 mm |
| Smart servo | 0.06 mm |
Tested
| Check | Result |
|---|---|
| Core maths tests (kinematics, Jacobians, energy balance of the dynamics, trajectories, IMU, servo) | 23 passed |
| Firmware build for the ATmega328P | 19.2 kB flash (58 %), 752 B RAM |
| Firmware inverse kinematics against the Python version | Checked by host unit tests |
Limits
The accuracy of a built robot is set by its servos, its mechanical build and its calibration, as the table above shows. The repository includes an industrialization guide with parts lists, wiring, a firmware protocol, calibration, ISO 9283 testing and safety, for anyone who wants to go beyond a hobby build.