Please enable JavaScript.
Coggle requires JavaScript to display documents.
Radar REM - Coggle Diagram
Radar REM
Scenario Breakdown
Dynamic Context
Drivable Path
Road Boundaries
Road Semantics
Applications
HD-Map Creation
Localization
Road Segment Level
Decimeter accurate
Absolute
Relative
Parking
Parking lot occupancy
Routing
Traffic Data Collection
PoC
Collect location redundant road scenarios
city
sub-urban
Separate Foreground and Background
Ego-Motion compensation
Remove detected dynamic objects
cars
motorbikes
other
Develop signatures, that can encode the background
Prerequisite for crowd sourcing due to bandwidth limitations
Train End-to-End how to encode Sensor Signals to be invariant to scene dynamics, while providing high localization sensitivity.
Utilizing an occlusion model
Train models to localize the ego vehicle
Road-Segment Level
Decimeter accurate localization within Road-Segment
How to get highly accurate ground truth ?
Train a model to estimate the relative location differences in confined spatio-temporal neighbourhoods by utilizing host data.
Quality assurance
Visualize
Uncertainties
Foreground / Background Separation
Road segment scene deviations
Highlight potential scene changes
Interesting Sources
Bosch/TomTom Paper (Radar Road Signature)
Pros
Unaffected by low light or bad weather conditions
Lower bandwidth than video
Pinpoint lane location to within a few centimeters
Can be used alongside other sensors
Cons
Video+Lidar would be more reliable
HDR could be used for low light conditions