watonomous.github.io
[ Electrical Division : Radar Driver Spring 2020 OKRs ]
Created by [ Frank Yan], last modified on May 08, 2020
[S20.O1] Multi-Radar Development
Finish multi-radar development by deleting duplicate points and performing more filtering algorithms.
[S20.O1.KR1] Delete duplicate points
One major weakness that the multi-radar set-up has is that when an object is detected by both radars, the object will show up in both point-clouds in RViz. We need to drop one of the point-clouds to prevent false positives and to get cleaner data.
[S20.O1.KR2] Implement better detection algorithms
As of now timestamps are the determining factor on whether or not something should be dropped. Implementing an algorithm that takes into account multiple factors such as RCS, distance, velocity, and timestamps will better determine what data needs to be dropped.
[S20.O1.KR3] Filter bad data
Before sending the data off to software, removing duplicate points, unnecessary noise, and inaccurate data will help software use the radars’ data better.
[S20.O2] Implement tracker
Implement tracker for dynamic object detection.
[S20.O2.KR1] Research object tracking
Talking to processing to get more insight on how object tracking is currently handled will help us get a better idea of what to do for the radars.
[S20.O2.KR2] Start development
Determine if the object tracker is going to be another node in the software stack or as another file in the multi-radar node.
[S20.O2.KR3] Dynamic velocity thresholding
Subtract the speed of the car to find absolute velocity of moving/stationary object.
[S20.O2.KR4] Determine necessary accuracy
Determine what level of object detection is needed before handing off to perception/processing.
[S20.O2.KR5] Testing
Dynamic object tracking will need to be tested vigorously with both rosbags and live radar once they get mounted to the car; will also need to be tested with objects both bigger and smaller than people.
[S20.O3] Radar Integration
Plan out how radars are going to be used for year 4 competition.
[S20.O3.KR1] Set up a meeting with power systems
Plan out how the radars should be connected to the car’s power.
[S20.O3.KR2] Determine how software is going to use the radar’s data
Discuss with the software teams on how they want the radars’ data to be used. (Dynamic objects and how to integrate the radars with the processing/path-planning code)
[S20.O3.KR3] Discuss sensor fusion with sensor-interfacing
Ask sensor interfacing if the radars should be a part of the sensor suite or as a separate network/configuration.
[S20.O4] Concept Design Report
Finish design report before August 2020 so Steven can read and edit the document.
Current document for the radar report: https://docs.google.com/document/d/13S7rl7l5iFEcxTU1Ptikjyx5iEBaASHPiL1C06j3zNg/edit
\
Comments:
+———————————————————————–+
| [] |
| |
| We're creating a object tracker library in prediction - unless there |
| are specific things you can do with radar, it might be easier for you |
| to add to what we're working on. |
| |
| {.smallfont align=”left” style=”color: #666666; width: 98%; margi |
| n-bottom: 10px;”} |
| 
|
| Posted by a293pate at May 09, 2020 17:16 |
| |
+———————————————————————–+
| [] |
| |
| I'll get back to you on that because Continental might be giving us |
| their tracker. |
| |
| {.smallfont align=”left” style=”color: #666666; width: 98%; margi |
| n-bottom: 10px;”} |
| |
| Posted by qf2yan at May 09, 2020 18:32 |
| |
+———————————————————————–+
Document generated by Confluence on Nov 28, 2021 22:40