This is a proposal for the development of a Winter Road Maintenance project. The main problem we faced in https://devpost.com/software/road-friction-forecasting was insufficient data. The data were collected in such a way that every day of observations contains only a small part of all unique coordinates visited over the entire period of observations. Moreover, daily observations are a series of records with constantly changing time and coordinates, which gives only one data point for specific location and timeper day. For the entire observation period, the number of data points per uniquecoordinates is between 1 and 32 with average 2.93. So, for each time we have only only one measurement and for each location - between 1 and 32. In such conditions, it is impossible to build a model taking into account the time and location of the collected data. But if we fix this issue, then I suggest using Graph Neural Networks as the most natural way to process road data, when roads are basically a graph.
- OSMNX - for building roads graph and collection geospatial data
- wwo_hist - for collecting weather data
- pyowm - for collecting UV-index data
- dgl - for GNN training
label = accidents rate of roadway condition * accidents rate of friction interval
| Friction | Rate |
|---|---|
| < 0.15 | 0.8 |
| 0.15 - 0.24 | 0.55 |
| 0.25 - 0.44 | 0.25 |
| 0.35 – 0.44 | 0.2 |
| Friction | State |
|---|---|
| Dry | 0.12 |
| Moist | 0.16 |
| Wet | 0.16 |
| Icy | 0.53 |
| Snowy | 0.3 |
| Slushy | 0.18 |
| Friction/State | Dry | Moist | Wet | Icy | Snowy | Slushy |
|---|---|---|---|---|---|---|
| < 0.15 | 0.096 | 0.128 | 0.128 | 0.424 | 0.24 | 0.144 |
| 0.15 – 0.24 | 0.066 | 0.088 | 0.088 | 0.2915 | 0.165 | 0.099 |
| 0.25 – 0.34 | 0.03 | 0.04 | 0.04 | 0.1325 | 0.075 | 0.045 |
| 0.35 – 0.44 | 0.024 | 0.032 | 0.032 | 0.106 | 0.06 | 0.036 |
pip install -r /path/to/requirements.txt
Preprocess road conditions data and collect weather data using utilities from /util and add OSM information using process_osmnx.py.
Continue modifying the gnn.py code:
- in function build_graph():
- replace src array with indices of points where roads start
- replace dst array with indices of points where roads end
- keep new graph bi-directional, OSMNX will handle directions
- for graph G set:
- G.ndata["feature"] - np.array with information about weather in points
- G.edata["feature"] - np.array with information about road conditions in edges
- G.edata["label"] - safety metric
- G.edata["train_mask"] - train-test split
A. Krutsylo and Md. Abrar Jahin (2021) DIT Source code.