Release Notes for Intel® RealSense™ SLAM Library

Version v2.0.4

Intel® RealSense™ SLAM Library provides Simultaneous Localization and Mapping capabilities. This is the SLAM Library Beta release supporting Ubuntu 16.04 LTS (Desktop version).

New Features and Changes

Please refer to the readme file for the features included in the SLAM Library. The following changes are introduced in this release:

  • Fixed Bug: Temporary freeze immediately after starting SLAM, causing tracking to fail
  • Fixed Bug: Occupancy residue left on 2DMap upon objects moving in scene
  • Fixed Bug: Calling get_occupancy_map_update without configuring SLAM did not return an error code
  • Fixed Bug: Calling get_camera_pose without configuring SLAM did not return an error code
  • Added get_occupancy_map_as_rgba(…) API as a convenient way to get a visual representation of the occupancy map
  • Removed enable_debug_image(…) API
  • Removed set_camera_pose(…) API

Compatible Platforms

The library is developed and tested on the following platform:

Software Requirements

  • Intel RealSense Cross Platform API - v1.12.1
  • Intel RealSense SDK - v0.7
  • OpenCV 3.1

Hardware Requirements

  • Intel RealSense ZR300 (DS4.1) Rev11 G sensor camera (DS4.1 Firmware 4.1, Adapter Board Firmware (CX3) 1.29, Motion Module Firmware (MM# 951805)
  • Intel Joule 570x Developer Kit with expansion board (MM# 951308)
  • Note: This middleware module was validated with one ZR300 camera connected to Joule.

Known Limitations

This release includes the following known limitations:

  • Stream/Camera Initialization:
    • For accurate results, tracking should be started when the camera is still. The system will start producing accurate tracking only after the camera has moved a few decimeters.
    • When starting Depth and Fish Eye streaming, the first few frames (between 2-10) may be dark. This is due to a camera limitation. It may be necessary to skip these frames before starting SLAM.
  • Lighting:
    • The system will not work in the dark or in low-visibility environments, such as inside clouds, heavy smoke, or at night. It also will not work when facing a very bright light or in conditions of direct sunlight.
    • Mixed lighting conditions (moving from bright to dark environment or moving from dark to bright environment) and extreme lighting conditions (too bright and too dark environments) negatively impact the localization and the 2D occupancy map.
  • Camera Movements:
    • Fast motions or rotations and strong accelerations may produce inaccurate results. Specifically, localization and mapping may be less accurate.
    • When mounted to a robot, the camera should be mounted in such a way as to minimize vibration. Vibration from the robot motion can cause tracking to fail. Using larger wheels and mechanically isolating the camera can both improve tracking quality.
    • Use in elevators is not supported: Radical drift in trajectory while and after using the elevator.
  • Scene’s Visibility:
    • Re-localization only occurs when the camera sees substantially the same scene it has seen before. The amount of difference is scene dependent, but in an open office environment, the camera must come back within a few tens of centimeters of its prior position, and within a few degrees of prior orientation, to recognize the location again.
    • If a large percentage of the visible scene is moving (non-static), accuracy will be reduced or tracking may fail.
    • Having no features visible to the camera, for example being directly against a smooth white wall, will result in tracking failure or inaccurate results.
  • CPU Utilization:
    • An overloaded CPU that is causing capture frame drops will result in inaccurate results.
    • Applications that are developed on the SLAM module should avoid utilizing more than 90% of the CPU (at peak).
  • The following are details of the best known configuration:
    • Depth resolution and format: 320x240x30, z16
    • Fisheye resolution and format: 640x480x30, raw8
    • Depth area of interest for 2DMap building on Z: [0.7m, 3m]


Copyright 2017 Intel Corporation Licensed under the Intel RealSense Pre-Release License Agreement, provided in this package as '/usr/share/doc/librealsense-slam2/copyright' **Other names and brands may be claimed as the property of others*