MyController — mc_rtc Humanoid Robot Controller

---

Overview

This controller demonstrates the core building blocks of whole-body humanoid control using mc_rtc:

• Posture regulation — keeps the robot near its default half-sitting configuration
• Center of Mass (CoM) control — maintains balance by regulating the CoM position
• End-effector control — moves the right wrist along a sinusoidal vertical trajectory
• Contact management — declares and enforces foot contacts with the ground

The target robot is the JVRC1 humanoid, using mc_rtc's built-in JVRC1 robot module.

---

Controller Architecture

MyController
├── Constraints
│   ├── ContactConstraint       — foot contacts stay valid
│   └── KinematicsConstraint    — joint limits respected
│
└── Tasks (solved jointly by QP solver)
    ├── PostureTask             — joint-space regulation (weight: 10, stiffness: 1000)
    ├── CoMTask                 — Cartesian CoM regulation (weight: 10, stiffness: 1000)
    └── EndEffectorTask (R_WRIST_Y_S)
                                — sinusoidal z-axis hand motion (weight: 5, stiffness: 1000)

At each control tick (run()), the QP solver computes joint torques/velocities that satisfy all constraints and minimise the weighted sum of task errors. The right-hand target is updated every tick to produce a smooth sinusoidal oscillation (±10 cm at 1 rad/s).

---

Dependencies

Package	Version tested
Ubuntu (WSL2)	20.04 LTS
mc_rtc (libmc-rtc-dev)	2.12.0
mc-rtc-data	1.0.7
mc-rtc-utils	2.12.0
CMake	≥ 3.1
GCC	9.4

Install mc_rtc on Ubuntu 20.04

curl -1sLf 'https://dl.cloudsmith.io/public/mc-rtc/stable/setup.deb.sh' | sudo -E bash
sudo apt install libmc-rtc libmc-rtc-dev mc-rtc-data mc-rtc-utils mc-rtc-gui

---

Building the Controller

cd MyController
mkdir build && cd build
cmake ..
make -j4
sudo make install

After installation the controller plugin is available at:

/usr/lib/x86_64-linux-gnu/mc_controller/MyController.so

---

Running the Controller

1. Configure mc_rtc

Create or edit ~/.config/mc_rtc/mc_rtc.yaml:

MainRobot: JVRC1

Enabled:
  - MyController

RobotModulePaths:
  - /usr/lib/x86_64-linux-gnu/mc_robots

ControllerModulePaths:
  - /usr/lib/x86_64-linux-gnu/mc_controller

2. Launch the controller

mc_rtc_ticker

Expected output:

[info] Loading controller MyController
[success] MyController init done

3. Visualise (optional)

In a second terminal:

mc_rtc_gui

This opens the mc_rtc stand-alone GUI showing live task targets, robot state, and solver output.

---

Design Decisions

Why PostureTask as a regularisation term? Without a posture task, the QP has an under-constrained null space — joints drift freely. The PostureTask at low weight acts as a soft regulariser that keeps joint configurations human-like without interfering with higher-priority Cartesian tasks.

Why CoMTask for balance? JVRC1 is a floating-base robot. Controlling the CoM position relative to the support polygon is the simplest approach to passive balance in a quasi-static regime. A full stabiliser (e.g. LIPMStabilizer) would be needed for dynamic motion but is out of scope here.

Why a sinusoidal end-effector trajectory in run()? Updating the target every tick (rather than setting it once in the constructor) demonstrates how a controller reacts dynamically to changing references — the pattern used when tracking sensor input, GUI interaction, or FSM transitions in production controllers.

Task weight choices:

• PostureTask: weight 10 — low, acts as regulariser only
• CoMTask: weight 10 — moderate, balance is important but should not block hand motion
• EndEffectorTask: weight 5 — lower than CoM so balance is preserved if the two conflict

---

Key mc_rtc Concepts Used

Concept	API
Controller base class	mc_control::MCController
QP solver access	solver()
Joint regularisation	mc_tasks::PostureTask
CoM balance	mc_tasks::CoMTask
Cartesian end-effector	mc_tasks::EndEffectorTask
Contact constraints	contactConstraint, solver().setContacts()
Kinematics constraints	kinematicsConstraint
Logging	mc_rtc::log::success()
Plugin registration	CONTROLLER_CONSTRUCTOR macro

---

Possible Extensions

• FSM controller: use mc_control::fsm::Controller to sequence states (e.g. HalfSit → ReachForObject → ReturnToStand)
• LIPM stabiliser: replace CoMTask with mc_tasks::lipm_stabilizer::StabilizerTask for robust dynamic balance
• Bilateral symmetry: add a second EndEffectorTask for the left hand, creating a synchronised reaching motion
• GUI integration: expose the hand target via mc_rtc::gui::Point3D so it can be dragged interactively in mc_rtc_gui

---

Notes on WSL2

mc_rtc_ticker may segfault on WSL2 Ubuntu 20.04 due to missing real-time kernel features that mc_rtc's internal timing loop relies on. The controller compiles, installs, and loads correctly; the runtime issue is environment-specific and does not affect the controller logic. Running on native Ubuntu or in Choreonoid resolves it.

---

Author

Giacomo Demetrio Masone
Technical assessment — CNRS-AIST Joint Robotics Laboratory
April 2026