Skip to main content
Version: 3.5.0

System Requirements

SENSR is a powerful perception engine that requires adequate computing resources for optimal performance. You can run SENSR on a pre-configured device from Seoul Robotics or on your own custom hardware. This guide details the requirements for each option.

For a streamlined, out-of-the-box experience, we recommend using the LiDAR Processing Unit (LPU), a dedicated edge computing device optimized for SENSR.

The LPU-O is our pre-configured hardware solution, designed and qualified to run SENSR efficiently. It provides a stable and powerful foundation for your deployment.

  • Processing Power: Can simultaneously process data from up to 8 LiDARs (32-channel models).
  • Coverage Area: Capable of covering an area of approximately 15,000 m².
  • Scalability: For larger sites, multiple LPU-O devices can be networked together to form a distributed system.

For inquiries about purchasing an LPU, please contact our sales team at sales@seoulrobotics.org.

Option B: Using Your Own Computing Device​

If you prefer to use your own hardware, SENSR can run on most modern machines that meet the following specifications.

GPU Requirements​

SENSR leverages deep learning for highly accurate object detection, which relies heavily on a capable GPU. Using a host machine with a dedicated NVIDIA GPU is strongly recommended for optimal performance.

  • Primary Factor: GPU load is primarily determined by the detection range. A larger coverage area requires more GPU power. The number of points and objects are secondary factors.
  • Reference Guideline: An NVIDIA RTX 3060 (with 3,584 CUDA cores) can effectively process an area of roughly 300x300 meters (90,000 m²).
  • Estimating Core Count: As a rule of thumb, you can estimate the required CUDA cores for your site with the following formula: Required CUDA Cores ≈ Area to Cover (m²) * 0.04

CPU Requirements​

CPU usage is influenced by several factors:

  • The specific LiDAR manufacturer's driver, as some are more CPU-intensive.
  • The total number of points being processed.
  • The number of concurrent objects detected in the scene.
  • Use of SENSR's built-in features like driver recovery and tilt auto-correction.

Reference Guidelines (using 32-channel LiDARs):

  • For a single LiDAR: Intel Celeron J1900 (Quad-core, 2.0GHz) or equivalent.
  • For up to 8 LiDARs: Intel 8th Gen i5 or better (i7 preferred) with a clock speed >2.0GHz.

You can fine-tune performance by allocating CPU resources within SENSR under Algo Settings > Performance. To monitor CPU usage, a tool like htop can be used to check the processes corresponding to each LiDAR driver.

RAM Requirements​

RAM consumption depends on the number of points and objects being processed at any given time.

  • Recommended: A minimum of 32 GB of RAM is recommended for smooth operation.
  • Minimum: 16 GB of RAM may be sufficient for smaller sites with 1-2 LiDARs and a detection area under 2,500 m².

Storage Requirements​

A minimum of 50 GB of available disk space is required to install SENSR software images, store logs, and ensure smooth operation.

Supported Operating Systems​

SENSR will only run on the following officially supported operating systems:

  • Ubuntu 20.04 LTS
  • Ubuntu 22.04 LTS
  • Ubuntu 24.04 LTS
  • Red Hat Enterprise Linux (RHEL) 8.x
  • NVIDIA JetPack 5.0
  • NVIDIA JetPack 6.0

If you require SENSR to run on a different OS, please contact our support team at support@seoulrobotics.org to discuss your needs.

Environment & Other Requirements​

  • Operating Mode: SENSR can run with a graphical user interface (GUI) or in a headless (non-GUI) mode. Both modes place significant load on the system, but headless mode uses slightly fewer resources and is recommended for production environments.
  • Graphics Library: To use the GUI, the host machine must have a graphics card that supports OpenGL.
  • File Permissions: The installer places project files and settings in the /opt/seoulrobotics directory by default. Please ensure the /opt/ directory has read/write permissions for the user running the installation.

If you have any doubts about selecting the right hardware for your deployment, please feel free to contact the Seoul Robotics support team.

Logo
If you need access, please contact