There are specific hardware requirements for RapidSense to function successfully. Special attention should be given to the following to achieve optimal performance.

PC/controller and OS requirements

Minimum required PC specifications

• Intel i7 or equivalent or better

• 128GB RAM minimum

• I/O ports for sensors (either USB or Ethernet)

• SSD hard drive preferred

• nVidia or AMD GPU required (mandatory)

OS Requirements

• Ubuntu Linux 22.04+

Native sensor support

RapidSense provides native support for a set of sensors.

Supported Sensors

• Standard applications: Intel RealSense D455

• High-resolution applications (requiring tight tolerances ~2-10mm): Photoneo Motioncam 3D depth + IR cameras (any size, both color and non-color)

Calibration requirements

Aruco tags are required to calibrate the sensors. RapidSense supports the April 36H11 tag format. See section "Calibration Tag Information" for further information.

Mounting calibration tags

Open

Example of an April 36H11 Calibration tag

In order to ensure a robust and repeatable calibration process, it is required that the calibration tag be fixed to the robot.

Users must design and install a mount for the calibration tag. It is recommended that the mount be attached to the end-effector.

The following are examples of robot-mounted tags.

 

Open

 

Sensor Mounting and Occlusions

Occlusions are the parts of the environment that a camera cannot see due an obstruction. It is recommended that sensor placement is made so as to minimize occlusions (e.g., it is not optimal to place a sensor directly behind a robot).

Sensor placement also depends on the robot application. For example, for a pick and place application, ask yourself, where will the robot move? Where will the occlusions be, and for which sensors? 

This process tends to be somewhat trial-and-error. 

To achieve optimal performance out of RapidSense, keep the following guidelines in mind when mounting your sensors:

1. Make sure to position the sensors in your workspace to maximize coverage of the environment. Most sensor providers provide CAD models of the sensors and include the field-of-view (FOV) in the model, which this can be used to visualize and verify the sensor ability to see the Volume of interest.

2. The Intel Realsense specifications are for +/- 2% depth accuracy when they’re within 2 meters of an obstacle, so we recommend placing cameras across the workspace so that obstacles are within this distance. 

3. Space the sensors apart from one another and at different angles to provide different perspectives of the scene. 

4. Mount the sensors securely and rigidly. If a sensor vibrates relative to robot motion, it will return noisy data resulting in false positives for obstacles. It can also cause sensors to shift and invalidate the calibration data.

5. Connect the sensor cables with strain relievers. 

Lighting Conditions

It is recommended to control lighting conditions as best as possible for applications through adding covers to deflect direct overhead lighting, sunlight, or other direct ambient light sources. Ambient lighting directly cast onto the sensors and the environment or that is reflected off of highly reflective surfaces can introduce noise and wash out images needed by the sensors and should be avoided and/or controlled (i.e., minimized). Changes in lighting conditions during operation, such as the effects of sunlight, should be prevented. Depending on sensor selection, the effects of lighting to the system may vary.