Data

This project uses the UTIAS Multi-Robot Cooperative Localization and Mapping Dataset (Mr. CLAM) from Autonomous Space Robotics Lab [3]. This 2D indoor dataset collection contains odometry and measurement data from 5 robots, as well as accurate ground-truth data for all robot poses and the positions for 15 landmarks.

The robots are two-wheel differential drive iRobot Create robots. Each robot has a camera and a barcode identifier. Odometry information in the dataset is given as a list of (forward velocity, angular velocity) sampled at approximately 67 Hz.

The landmarks are fixed poles with barcode identifiers.

Each robot, using its camera, finds the range and bearing of all barcodes (of landmarks or other robots) in its view. The dataset preprocesses this data to return measurements as a list of (timestamp, subject number, range, bearing) information. The frequency of the measurements is much less than the frequency of the odometry information and is not sampled at the same time.

The ground truth information is provided as pose information for each robot at 100 Hz with 10 external cameras.

We resample all the data with linear interpolation at 10 Hz so that all the odometry, measurement, and ground truth data are on the same timebase. 10Hz was chosen as a trade-off between computation time and accuracy of the odometry-based prediction step. The movement of the robots is relatively slow, over an approximately 25-minute dataset.

This dataset provides real-world test data for the implementation of SLAM with different scenarios of cooperative localization. This dataset was chosen in particular because it does not require the use of computer vision to implement SLAM.