Nuclear energy requires operations to be performed by remote handling, due to radiation levels. Thus, accurate vehicle mapping and localisation in this complex environments is essential to move forward on today’s clean energy pursue. SLAM(Simultaneous Localisation And Mapping) with an EKF approach is one method of achieving this when the reliance on external sensors to the robot is not a plausible scenario. On this project, the aim is to implement and test this algorithm using artificial landmarks to help a International Thermonuclear Experimental Reactor(ITER) remote handling transport cask prototype navigate through space.
The project was developed for the Autonomous Systems course in Instituto Superior Técnico, Universidade de Lisboa.
For further understanding, you may want to read the report on this.
We provide simulated as well as real datasets. The real datasets include odometry from the ITER prototype at a 100ms rate, images of visual markers taken at 1s rate, respective .txt files containing the [timestamp, id, bearing, range] of each marker seen by the robot and functions to concatenate this data.
To run the algorithm use the file ekf_slam.m and define sim as 0 or 1, to use real or simulated data, respectively. If it's real data, load the respective data file from the data/ directory or create your own with the information above. Get some popcorn and watch the almost reasonable results.
The camera used is an uEye LE USB 3.1 Gen 1.
The ITER prototype used was handed pre-built with NXT Lego hardware. There is a user guide.
Both ueye and opencv version 3-3-0 were used for working with the image processing under Ubuntu 14.04.
ITER data was collected under Windows OS.
arUco library was used for the visual markers. 12 different 4x4 arucos with id from 0 to 11 and 15cm of length were used.
The camera was calibrated using arUco boards also provided in this repo as well as with 65 images of a chessboard, producing the calib_arucoboard.xml and calib_chess.xml.
For once, we thank Microsoft Windows for being the only OS to connect to ITER.
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent