Bad road conditions are a plague for smart cities, with increasing electrified vehicles and heavier loads. We aim to leverage road maintenance and safety of autonomous vehicles as well as pedestrians by utilizing the remote sensing data taken by satellites and space shuttles, as well as the road roughness data. By integrating multiple datasets we aim to provide a comprehensive solution for future infrastructure.
What it does
Crater gets satellite data (cloud density) and converts it to weather data to then inform the traffic department and drivers about the road conditions. Based on this weather data it predicts the road degradation levels. On the other hand, it takes the road roughness data into a classification model, combined with the remote sensing image analysis, a "road score" is given to provide reference for road maintenance staff, autonomous vehicles as well as pedestrians. Road maintenance staff can be notified if the road score is low enough, indicating potential bad road conditions. Autonomous vehicles can utilize that information to bypass the roads with bad conditions, increasing passengers' safety, and pedestrians can be notified by an augmented-reality mobile app which visualizes the roads sections in trouble in real time.
How we built it
We built the prototype given the dataset granted by Space Dataset (by the Swedish Space DataLab) and NiraDynamics. We estimated cloud density based on the satellite imagery from the space dataset. The dataset of NiraDynmics gave us information about the level of road roughness; with that data; we get a classification model to predict if the road needs repair.
Challenges we ran into
Meeting the deadline was one of our biggest challenges as our group was in different time-zones. It took us some time to coordinate to establish the problem and we did what was possible during the remaining time. Another challenge comes from the time limitation for training a better model that is more representative even in other areas. Further, we spent a long time understanding various datasets as well as using their custom APIs to gain access to the data.
Accomplishments that we're proud of
One of our key achievements is by getting reliable real-time information for users impacting the mobility system with improved efficiency and sustainability, specially under severe weather conditions (snow storms). We are able to combine multiple information from various perspectives, and extract valuable information out of it.
What we learned
The datasets provided by SpaceDataset and NiraDynamics are precious sources of information that will progress in the mobility network management. Drivers and Traffic Department will be able to avoid congested areas and count on more reliable roads.
What's next for Crater
In the next stage, we aim to incorporate more neural network models derived from more comprehensive datasets, in order to provide a more representative model that could actually be applied to autonomous vehicles. A mixed-reality 3D interface is also desired at the next stage