Dyros Robot Controller¶

Dyros Robot Controller is a ROS 2 C++ library for mobile robots, manipulators, and mobile manipulators. It provides a unified controller API backed by Pinocchio for kinematics/dynamics and OSQP for quadratic-programming solvers, along with Python bindings and MuJoCo simulation examples.

:material-robot-arm: Manipulator control

Joint-space PD, CLIK, OSF, QP-based inverse kinematics and inverse dynamics, and hierarchical QP — all with CBF safety constraints.

→ Joint-space control · → Task-space control · → QP / HQP
:material-car: Mobile robot control

Kinematic support for differential-drive, mecanum-wheel, and caster-wheel mobile bases.
:material-robot: Mobile manipulator control

Whole-body QP / HQP interfaces that split optimized velocity or torque commands into a mobile-base vector and a manipulator vector.
:material-code-braces: Python and C++ examples

Minimal MuJoCo simulation loops for the Franka FR3, Summit XLS, and FR3-on-XLS robots.

Package Layout¶

Path	Purpose
`include/dyros_robot_controller/`	Public C++ headers
`src/`	C++ controller, solver, and binding implementations
`drc/`	Python wrapper package
`examples/C++`	C++ MuJoCo examples
`examples/python`	Python MuJoCo examples
`examples/robots`	Example robot URDF/SRDF models and assets

Controller Classes¶

C++ class	Python module	Purpose
`drc::Manipulator::RobotData`	`drc.manipulator.RobotData`	State storage and kinematics/dynamics (manipulator)
`drc::Manipulator::RobotController`	`drc.manipulator.RobotController`	Control computations (manipulator)
`drc::Mobile::RobotData`	`drc.mobile.RobotData`	State storage (mobile base)
`drc::Mobile::RobotController`	`drc.mobile.RobotController`	Control computations (mobile base)
`drc::MobileManipulator::RobotData`	`drc.mobile_manipulator.RobotData`	State storage (whole-body)
`drc::MobileManipulator::RobotController`	`drc.mobile_manipulator.RobotController`	Control computations (whole-body)

Quick Start¶

from drc.manipulator import RobotData, RobotController
from drc.type_define import TaskSpaceData
import numpy as np

robot_data = RobotData(dt=0.001, urdf_path="path/to/robot.urdf")
robot_ctrl = RobotController(robot_data)

# Set QP-IK weights
dof = robot_data.get_dof()
robot_ctrl.set_QPIK_gain(
    w_tracking   = np.array([100]*6, dtype=float),
    w_vel_damping = np.ones(dof) * 0.01,
    w_acc_damping = np.ones(dof) * 0.1,
)

# Control loop
while True:
    robot_data.update_state(q, qdot)                          # update from simulator

    task = TaskSpaceData()
    task.x_desired         = T_desired                        # target SE(3) pose
    task.xdot_desired      = np.zeros(6)
    task.x_init            = T_init
    task.xdot_init         = np.zeros(6)
    task.control_start_time = t_start
    task.current_time       = t_now

    ok, qdot_cmd = robot_ctrl.QPIK_cubic({"end_effector": task}, duration=3.0)

See Installation for build instructions and Examples for complete simulation loops.