Robotics II

Poznan University of Technology, Institute of Robotics and Machine Intelligence

Laboratory 3: Cones classification using LIDAR sensor intensity data

Back to the course table of contents

1. Intensity data

Intensity is a value connected with the laser distance measurement. It describes the strength of returning laser pulse for every measured point. Value of intensity varies for different types of objects, surfaces and colors. Moreover the measurement can also be affected by object roughness, moisture content, range or scan angle. Intensity data can be used in lidar point classification.

Formula Student Driverless track marking

Formula Student track for Driverless competitions purpose is marked with 4 types of cones:

Positions and function of every used cone class is as follow: - big orange - start / stop line - small orange - brake area for Acceleration and Skid Pad competitions - small yellow - right side of the track - small blue - left side of the track

Assumptions are illustrated by figure below:

Cones in LIDAR view

Figures below show accordingly yellow cone, blue cone and the Trackdrive in view of PointCloud2 message from laser scanner.

2. Dataset

Regarding to the lack of measurement intensity data in Formula Student Simulator (in general in AirSim-based simulators) rosbag files shared by Edinburgh University Formula Student AI Team will be used. During laboratories fsai.bag and single_lap_processed.bag (available here) will be utilized for evaluation purposes. However training, validation and test datasets were created with the use of other rosbag files.

3. Tasks

Part I

1. Make a copy of Google Colab interactive tutorial to your Google Drive. Follow the instructions in the notebook.

2. To the eKursy platform add exported classification model in ONNX format.

Part II

1. Run rosbag file using rosbag play command, start rviz tool (rosrun rviz rviz) and visualize lidar PointCloud2 message.

2. Create depth image. #ToCheck

3. Create intensity image. #ToCheck