Technical Specs
Warning: The stereo INS component self-calibrates the IMU during its initialization. It is therefore required
that the rc_visard is not moving and sufficient texture is visible during startup of the stereo INS component.
6.3.2 Available state estimates
The rc_visard provides seven different kinds of timestamped state-estimate data streams via the rc_dynamics
interface (see The rc_dynamics interface, Section 8.3):
Name Frequency Source Description
pose 25 Hz best effort Pose of camera frame, slightly delayed but most accurate
pose_ins 25 Hz Stereo INS Pose of camera frame, slightly delayed but most accurate
pose_rt 200 Hz best effort Pose of camera frame
pose_rt_ins 200 Hz Stereo INS Pose of camera frame
dynamics 200 Hz best effort Pose, velocity and acceleration in IMU frame
dynamics_ins 200 Hz Stereo INS Pose, velocity and acceleration in IMU frame
imu 200 Hz Stereo INS Raw IMU data
Best effort here means that if SLAM is running, then it contains the loop-closure corrected estimates and is equiv-
alent to the stream from Stereo INS when SLAM is not running.
Camera-pose streams (pose and pose
_
ins)
The camera-pose streams called pose and pose
_
ins are provided at 25 Hz with timestamps that correspond to
image timestamps. The former stream is the best-effort estimate, combining rc
_
slam and rc
_
stereo
_
ins if the
SLAM component is running. If SLAM is not running, then both data streams are equivalent. Pose values are
given in world coordinates, and also refer to the rc_visard’s camera frame origin (see Coordinate frames for state
estimation, Section 6.3.1). They are the most accurate estimates, taking all available rc_visard information into
consideration. They can be used in modeling applications, where camera images, depth images, or point clouds
have to be aligned highly accurately with each other. To ensure the greatest possible accuracy, these pose values
are delayed until a corresponding visual odometry measurement is available.
Real-time camera-pose streams (pose
_
rt and pose
_
rt
_
ins)
Two real-time pose streams called pose
_
rt and pose
_
rt
_
ins are provided at the IMU rate of 200 Hz. The
former stream is the best-effort estimate, combining rc
_
slam and rc
_
stereo
_
ins when the SLAM component
is running. If SLAM is not running, then both data streams are equivalent. They consist of the pose estimates of the
rc_visard’s camera frame origin (see Coordinate frames for state estimation, Section 6.3.1) in world coordinates.
The values given in these streams correspond to the values in the real-time dynamics streams, but give the pose of
the sensor/camera coordinate frame instead of that of the IMU coordinate frame.
Real-time dynamics streams (dynamics and dynamics
_
ins)
Two real-time dynamics streams called dynamics and dynamics
_
ins are provided at the IMU rate of 200 Hz. The
former stream is the best-effort estimate, combining rc
_
slam and rc
_
stereo
_
ins when the SLAM component
is running. If SLAM is not running, then both data streams are equivalent. The estimates can be used for real-
time control of a robot. Since the values are provided in real time and visual odometry computation requires some
processing time, the latest visual odometry estimate may not be included. Therefore, these estimates are in general
slightly less accurate than those in the non-real-time camera-pose streams (see above), but are the best estimates
available at this instant. The provided dynamics streams contain the rc_visard’s
• translation p = (𝑥, 𝑦, 𝑧)
𝑇
in 𝑚,
• rotation q = (𝑞
𝑥
, 𝑞
𝑦
, 𝑞
𝑧
, 𝑞
𝑤
)
𝑇
as unit quaternion,
6.3. Sensor dynamics 41










