RoboTech Vision - Sensor Box - Autonomous Navigation Sensor
The RTV Sensor Box is a sophisticated combined sensor suite devised for autonomous navigation. It empowers robots to autonomously navigate to destinations, even in complex intersections, by providing critical functionalities such as route planning, obstacle detection, and real-time environment mapping. This device integrates a 360° camera for path segmentation, a front camera with object recognition capabilities, and a 3D LiDAR to ascertain obstacle size and distance. The components are meticulously calibrated to each other, ensuring high-accuracy outputs. It operates as a universal device compatible with existing platforms but is ideally used with dedicated robotic platforms. The RTV Sensor Box scans its surroundings with a 3D LiDAR providing a full 360° view, which is crucial for obstacle dimensioning and distance assessment. It also employs neural networks for path segmentation and dynamic obstacle classification. It runs on an 8-core ARMv8.2 CPU, enhanced by a 512-core Volta GPU, supported by Linux Ubuntu with ROS Melodic, and meets the IP65 protection standard, making it robust for diverse environments.
The RTV sensor Box is a combined sensor that enables the robot to plan a route, autonomously pass to the destination even through complex intersections, detect and avoid obstacles, or preprocess data for other algorithms. The task of the sensor is to segment the path, detect static and dynamic obstacles and recognize them. Using data from the sensor, it is possible to create a map of the environment which can be used for navigation of autonomous devices equipped with an RTV sensor Box.
The sensor combines a 360° camera that provides path segmentation data, a front camera with object recognition, and a 3D lidar to give an idea of the size and distance of obstacles. The individual parts of the sensor are calibrated to each other, which ensures high output accuracy. It is a universal device that can also be deployed on existing platforms. The optimal solution is to use one of our robotic platforms.
Scans the environment
The RTV sensor Box scans the environment using 3D-Lidar. It has an idea of the obstacle´s dimensions or distance in the range of 360°.
Segments the path
With 360° camera and algorithms using neural networks, the sensor is able provide information about the suitable path to the destination.
Recognizes objects
Static and dynamic obstacles detected by Lidar are recognized and classified into the groups by a front camera and neural network.
Dimensions
- Width: 170 mm
- Height: 330 mm
- Depth: 170 mm
- Weight: 2 kg
Components
360° camera, front camera, 3D scanner
Power supply
18 – 36 VDC
CPU
8-core ARM v8.2 64-bit CPU, 8MB L2 + 4MB L3
GPU
512-core Volta GPU
Operating system
Linux Ubuntu with ROS Melodic
Software
Path segmentation, detection of static and dynamic objects, face recognition, environment scanning
Outputs
- raw data from sensors (image from 360° camera, image from the front camera, point cloud from 3D scanner)
- detected objects marked with a bounding box (coordinates of corners, object designation, and success rate in percent)
- live view with marked detected objects (bounding boxes)
- path segmentation using a mask
- live view with path segmentation using a mask
- projection of data from a laser scanner into a 2D image mask
- path segmentation mask with subtracted 2D mask projection (path segmentation with removed obstacles that were not recognized)
Environment
Protection class: IP 65