fennhelloworld / car

Radar + Camera Perception Pipeline for an Autonomous Car

mercedes-clk-perception

1. Install ROS melodic:

Instructions are available at: http://wiki.ros.org/melodic/Installation/Ubuntu

2. install dependencies:

pip3 install -r requirements.txt

Most of the packages are already provided with ROS. Virutal environment is preferable.

3. Download rosbags:

https://drive.google.com/a/sjsu.edu/file/d/1-OQuWuRwvfPK0akXgAcgkNtSuKIMSzwA/view?usp=sharing

4. Run nodes:

python3 fusion_node/main.py
python3 gps_node/main.py
python3 radar_node/main.py
python3 2d_detection_node/main.py #(additional instructions in 2d_detection_node/README.md)

5. Play rosbag:

roscore
rosbag play 720loop1.bag

6. Check if nodes are working:

If everything worked out successfully

$ rostopic list

Should contain the following elements:

#standard ros topics
/rosout
/rosout_agg

#raw data:
/fix
/ti_mmwave/radar_scan_pcl
/zed_node/left/image_rect_color

/2d_node/output_image #2D detection pipeline
/2d_node/results

/gps_node/heading #gps pipeline
/gps_node/speed

/radar_pipeline/clusters #radar pipeline

Your final setup can look simmilar to this one:

Results

There are some examples of the inference below. All examples of 2D inference generated with the system are avalable in https://github.com/tomek-l/mercedes-clk-perception/tree/master/images

For more information, go see the paper https://github.com/tomek-l/mercedes-clk-perception/blob/master/docs/CMPE297_03_PilotA.pdf