sisl / AutomotiveDrivingModels.jl

Driving simulation architecture for Julia

Other

63 stars 29 forks source link

[WIP] New feature extraction mechanism #72

Closed MaximeBouton closed 4 years ago

MaximeBouton commented 4 years ago

This PR implements a new feature extraction mechanism and interface.

70

Users can extract features from a list of scenes using the extract_features function.
```
dfs = extract_features((posgx, posfs, vel), scenes, roadway, ids) 
```
[x] possibility to have parameter in the feature type (e.g. DistanceTo(:bob), CollisionWith(:alice), ...). This is conflicting with the use of symbols and needs to be discussed.

Most of the features from the previous version are implemented

MaximeBouton commented 4 years ago

I have a first draft of the feature extraction working all the way. You can see the example in test/new_test_features.jl.

My implementation is using a lot of broadcasting. Some stuff is not efficient, like constructing a dictionary all the times. If we only want to keep extract_features as user facing, there might be some optimization possible in the lower level functions.

I decided to use dictionary instead of NamedTuples because it is a bit more flexible in terms of key types. If the ids are integer we cannot do df.1 .

cc @wbrannon @mattuntergassmair

coveralls commented 4 years ago

Coverage increased (+4.4%) to 79.154% when pulling 5be72108f012ccba3a2433bbc369c7e27719e923 on feature-extraction into 94a1dd8cddbd54fc318f3d8b0770f4a58e8ca955 on master.