This repo contains results collected during my period as HiWi at the ASL Lab under Michael Pantic work on Radar sensors.
Related repositories are;
- radar_rig_sensor_fusion: Contains information on the developed radar rig
- web-app: Nifty web-app installed on the radar rig to record data easily
- zadarlabs ROS driver: Wrapper for provided ROS driver
- zadarlabs_implementation branch: Branch of rio repository containing the integration of the zadarlabs sensor. (Here, simply look at the changes between the two last commits to find changes made.)
RadarLiteratureReview.pdf: File in this repo, containing literature review done on radar odometry and highlights some potential areas.
Data can be downloaded from GoogleDrive.
Different 'tests'/evaluations were made;
- RIO_RH_LH_Analysis compares RIO estimates using the TI sensor and determining, what effect a flipped z-axis has on the estimate. This evaluation is done because the TI sensor outputs data in a left-handed (LH) coordinate frame, and it was not apparent, if and where RIO considers this.
- ZadarlabsPointCloud aggregates the datapoints from the ZadarLabs radar sensor to determine, whether structures become visible.
- RIO_Comparison compares the RIO estimates using TI or ZadarLabs sensor.
Data can be found on GoogleDrive.
Data should be downloaded and placed in the data folder.
Here is an overview of the different data collected and some comments:
| Directory | Explanation | Comment |
|---|---|---|
data/AllSensors_noRIO/ |
Data was collected specifically for ZadarLabs point-cloud aggregation. Used in the ZadarlabsPointCloud evaluation. |
Data was initially collected with incorrect tf for ZadarLabs sensor and was post-processed to have correct transformations. |
data/RH_LH_raw/ |
Raw data collected to make the Right-Hand vs Left-Hand coordinate frame comparison for TI radar sensor, found in RIO_RH_LH_Analysis. |
Does not contain ZadarLabs data. |
data/RH_LH/ |
Contains the output of the RIO estimate made on the RH_LH_raw datasets. defaultRIO refers to default implementation where the modRIO is with the z-axis flipped (in the RIO ROS package). |
The RIO estimate was made in post after recording the bag files. |
data/RIO_Comp_raw/ |
Data where TI & Zadar data was collected in order to compare RIO estimates side-by-side. For TI and Zadar there is each one dataset, where only one of the radar sensors was recorded. This could maybe be used to see whether cross-interference is relevant. |
|
data/RIO_Comp |
In raw/ we find the unprocessed recordings made with the master.launch file from the radar rig. Using the scripts rio_ti.launch and rio_zadar.launch the estimates are made and the datasets with respective names were generated. |
The first few (~5) seconds are cut from the respective bag files (hence '...cutfromfront.bag' naming). |
Software Guide and driver can be found on GoogleDrive.
CAD files can be found on GoogleDrive and more details on the radar-rig at the repository.