Robot kinematics
Robot kinematics – definition
Robot kinematics is the mathematical description of motion without modeling the forces that cause it. In mobile robotics, it defines how commanded wheel velocities, steering states, or body motions translate into robot movement in space. For unmanned ground vehicles (UGVs), kinematics is used to relate actuator commands to the robot base frame, estimate motion from wheel encoders, and constrain planning and control algorithms.
In ROS and ROS 2, robot kinematics is expressed through standard coordinate frames, motion models, and message interfaces. The most relevant conventions are the ROS coordinate frame conventions in REP 103 and the mobile platform frame semantics in REP 105. For a ground robot, the kinematic model usually maps wheel angular velocity to planar body velocity expressed as linear velocity along x and angular velocity around z. This is the basis for odometry, local control, trajectory tracking, and navigation stack integration.
For Fictionlab platforms, kinematics is primarily discussed in the context of differential-drive UGVs. Leo Rover uses a four-wheel skid-steer layout. In practice, this is often approximated as a differential-drive platform depending on the controller and the level of approximation. Raph Rover, as a larger UGV platform with higher payload capacity, also depends on a calibrated kinematic model for predictable low-level motion, odometry fusion, and autonomous navigation.
Why kinematics matters on UGV platforms
On a mobile robot, the kinematic model is not a theoretical detail. It directly affects how velocity commands are interpreted and how accurately the robot reports its pose over time. If the model is wrong, the robot may drift, rotate under straight-line commands, or fail to follow trajectories generated by a planner.
For ROS 2-based UGV systems, robot kinematics is used in several subsystems at once:
- the motor controller converts desired body motion into wheel commands
- the odometry node estimates motion from encoder feedback
- the localization stack fuses odometry with IMU, LiDAR, camera, or GNSS data
- the navigation stack assumes specific nonholonomic motion constraints
This is especially important on small and medium research rovers. Leo Rover can be used with ROS 2, typically Humble or newer, on onboard compute such as a Raspberry Pi. On such hardware, a simple and well-calibrated kinematic model is often more useful than a computationally heavy but poorly tuned estimator.
Kinematic models used in mobile robotics
Several kinematic models exist for UGVs. The correct choice depends on wheel arrangement, terrain interaction, and controller assumptions. For Leo Rover and similar four-wheel skid-steer robots, the common approximation is planar differential drive or skid-steer. On loose terrain, skid effects become significant, so the model becomes only approximate.
| Model | Typical platform | State variables | Main limitation |
|---|---|---|---|
| Differential drive | 2-wheel or differential base approximations | x, y, theta |
Assumes limited lateral slip |
| Skid-steer | 4-wheel or tracked UGV | x, y, theta |
Slip depends strongly on terrain |
| Ackermann | Steered UGV | x, y, theta |
Not applicable to differential platforms |
For a differential-drive model, the standard equations are:
v = r * (wr + wl) / 2
omega = r * (wr - wl) / b
Where:
v– linear velocity of the robot base in m/somega– angular velocity around the vertical axis in rad/sr– effective wheel radius in mwr,wl– right and left wheel angular velocities in rad/sb– wheel separation or effective track width in m
These equations are used both for forward kinematics, from wheel motion to body motion, and inverse kinematics, from body command to wheel targets.
How robot kinematics appears in ROS 2
In ROS 2, kinematics is implemented through message flow, transforms, and controllers rather than a single package. A mobile base usually accepts velocity commands on geometry_msgs/msg/Twist, publishes odometry as nav_msgs/msg/Odometry, and maintains frame relations in tf2.
The most common interfaces are the following:
/cmd_vel– commanded body velocity, usually interpreted in thebase_linkframe by convention/odom– wheel- or fused-odometry estimateodom -> base_link– transform produced by odometry or localizationbase_link -> sensors– fixed transforms defined in URDF
REP 103 defines standard units as SI units and right-handed coordinate conventions. REP 105 defines the semantics of frames such as map, odom, and base_link. These REPs matter because kinematic consistency depends on frame correctness. A sign error in wheel velocity or an inverted frame can make odometry unusable.
A typical ROS 2 inspection sequence may look like this:
ros2 topic echo /cmd_vel
ros2 topic echo /odom
ros2 run tf2_ros tf2_echo odom base_link
Key parameters and metrics
The quality of a kinematic model depends on a small set of measurable parameters. These values should come from platform documentation, CAD, or calibration experiments. For research-grade navigation, they should be validated on the real vehicle, not only copied from nominal dimensions.
| Parameter | Meaning | Typical unit | Impact |
|---|---|---|---|
| Wheel radius | Effective rolling radius | m | Linear velocity scale |
| Wheel separation | Distance between left and right drive sides | m | Angular velocity scale |
| Encoder resolution | Counts per revolution | CPR or ticks/rev | Odometry quantization |
| Odometry update rate | Published estimator frequency | Hz | Controller stability and fusion quality |
| Slip ratio | Deviation between wheel rotation and real motion | dimensionless | Accuracy on rough terrain |
In practical ROS 2 systems, odometry is often published at 20-100 Hz depending on the controller and compute budget. IMU data is commonly available at higher rates, often 100-400 Hz depending on the sensor. The exact rates depend on the hardware vendor and driver configuration and should be verified in the device specification and the running ROS graph.
Leo Rover and Raph Rover context
On Leo Rover, robot kinematics is central to any autonomous stack because the platform is not an autonomous vehicle out of the box. It provides a skid-steer mobile base, ROS integration, and the ability to integrate navigation software. The kinematic layer must therefore be configured correctly before SLAM, Nav2, or sensor fusion can perform reliably.
Typical Leo Rover tasks include:
- mapping indoor and outdoor test areas with LiDAR or depth cameras
- teaching odometry, frame trees, and controller tuning in education
- testing wheel slip and localization drift on mixed terrain
For Raph Rover, the same principles apply, but higher payload and larger inertia change tuning requirements. A larger platform often carries heavier sensors such as 3D LiDAR, RTK GNSS, or industrial compute modules. In that case, kinematics is still the motion foundation, but it is usually paired with stronger state estimation and terrain-aware control.
URDF, controllers, and calibration
In ROS 2, the robot description should encode wheel geometry and frame positions in URDF. The control layer then uses these parameters to compute wheel motion and odometry. If URDF dimensions and controller parameters disagree, the robot may behave correctly in simulation but drift in hardware.
A simplified YAML-style controller configuration may include values such as:
diff_drive_controller:
ros__parameters:
left_wheel_names: ["left_front_wheel_joint", "left_rear_wheel_joint"]
right_wheel_names: ["right_front_wheel_joint", "right_rear_wheel_joint"]
wheel_separation: 0.36
wheel_radius: 0.0625
publish_rate: 50.0
base_frame_id: base_link
odom_frame_id: odom
These values are examples only. Real parameters must be taken from the actual platform geometry and verified experimentally. Calibration usually consists of straight-line tests, in-place rotation tests, and comparison against external reference data such as motion capture, visual odometry, or RTK GNSS.
Limitations and trade-offs
Kinematics does not model torque, traction limits, or suspension dynamics. For UGVs on grass, gravel, ramps, or deformable ground, this matters. Wheel encoders may indicate motion that did not happen, or fail to capture lateral slip that did happen. This is why kinematic odometry alone is rarely sufficient for long autonomous runs outdoors.
In practice, a UGV stack combines kinematics with additional sensors:
- IMU for short-term rotational stability
- LiDAR or visual SLAM for drift correction
- GNSS or RTK for global position outdoors
This separation is important. Kinematics gives the motion prior. Localization corrects it.
Normative references and standards
The most relevant references for this topic in ROS-based mobile robotics are REP 103 for standard units and coordinate conventions and REP 105 for mobile platform coordinate frames. For message and transform implementation, the relevant interfaces are geometry_msgs/msg/Twist, nav_msgs/msg/Odometry, sensor_msgs/msg/JointState, and tf2. For controller behavior, ros2_control documentation and the differential drive controller documentation are the primary implementation references. For terminology, IEEE robotics literature distinguishes kinematics from dynamics in the standard way: kinematics describes motion, while dynamics includes forces and torques.