PathMaster

The Road to PathMaster.

During the Russian special military operation, Ukrainian artillery destroyed a column of Russian military equipment. It was actively observed that the movement of military transport required the renewal of road information as regularly as possible. This led us to the idea that we can use satellite data to determine the quality of roads, more precisely - whether the roads are not bombed and passable.

To recognize the importance of road infrastructure is to promote a safer and more optimal organization and movement of people and military units. We are developing a pathfinder and route selection assistant for the exercise of large mobile units based on satellite imaging and a combination of weather and environmental databases to analyze road terrain conditions.

With satellite imaging, we are mainly focusing on roads and the terrain around them - official or those detected/found by satellite. Thus, we must create a skeleton network of roads where every segment (max 10km) has an updating profile with a value to determine how it could carry out a commute, and another algorithm to offer the best option.

The map of the routes and partly the quality of the road we determined by direct images from the Copernicus Sentinel-1 and Sentinel-2. Asphalt has a unique albedo. This unique value can be used to determine changes made to a part of a road by observing the difference in albedo on the same road under similar weather conditions. While it may not be possible to determine the exact quality of a specific part of the road, we can still use this information to identify the amount of road that has changed by comparing it to older images and thus also collecting a lot of historical data. For instance, assuming we have a pixel that covers a square of about 10 by 10 meters, we can observe the difference in albedo on this specific pixel and use it as an indicator of changes in a segment. The precision of this method is influenced by weather and sun angle. For instance, if it has rained recently, the albedo will be significantly higher, and the precision of the observations will improve, as asphalt is highly reflective. Additionally, if there are steep differences in albedo on a specific part of a road, it can be considered suspicious and investigated further via different means of quality determination such as IP cameras, airplanes, drones, or infantry recon units. Through this technique, we actually get around one huge drawback of these old EU satellite systems - the resolution. By simplifying the data processing to a binary level - passable road or bombed road, we get a model that is able to assess the quality of the road with high accuracy, based on the stored historical data.

All the mapped routes would be portioned into segments based on some condition that make it different from the next one - a crossing or a road type. However, if a segment is <1km or >10km, then it would accordingly be added to another or split into two. The whole point of this is accuracy and precision. The more separate profiles with their own individual values, the better path finding calculations. 

The routing and the other half of road quality determination is based on algorithmic and database work. An algorithm that suggests an optimal option. It ingests problem-relevant data from the user - for example, the type of military unit used for what commute - and using the satellite-assisted mapping model outputs a path for it. When would it decide to go on a different road? When would that be optimal? This is pretty much the traveling salesman problem. For every road segment we create a profile which contains multiple values to help calculate the best path. Calculating best possible path model takes in consideration following parameters:

1. Segment Length

2. Road Quality

   1. Road type (asphalt, dirt, etc.,)

   2. Surrounding environment (forest, urban, plain, etc.)

3. Weather data

4. Type and amount of transport units

Satellite imaging lets us get new data in 1 week compared to other services that base they’re imaging on more involved and attendance-based practical techniques. For example, to update google maps over Ukraine, a plane must fly over restricted air space, but the satellite infrastructure is already there and regularly can refresh its imaging. We must make use of what we can.

Next steps to make our solution more capable:

  1. Analyze adjacent territories to main roads and evaluate their passability. In case if the road segment is damaged then our solution will try to find a new route even through fields and swamps taking into consideration climate factors.

   2. Improve evaluation of military transport unit impact on road quality. For example, calculate how much a military convoy will damage the road. 

   3. In time we could further rely on automation and algorithmically associate the image with other higher-quality maps for improved precision. Thusly, we could continue to integrate with other data sources (drones, sensors, etc.)

   4. Additional features that can predict most possible enemy location and heightened risk areas.

   5. A ML algorithm that projects road damage based on the planned usage.

   6. Use other waveleghts to be able to determine more data about the impact of damage. For example, how deep are the holes.

Bussines potential

Our market auditory is military but we see the potential of this product also in the private logistic sector.

Proposed business model - participating in business accelerators and grant programs to get base financing for product development. Price of our solution is calculated by the length of routes that are calculated every day. Price in a year would be approximately 20 000 euro for a 1000 kilometer route calculation every day.

About the team

Our team consists of highly specialized and motivated individuals.

Core facts:

• 4 IT specialists from different fields and walks of life.

• Cybersecurity and data analysis with over 5 years of experience.

• Experience with software development and research field as well.

The members:

Kaspars -  an innovation developer at Accenture Baltics and University of Latvia MedTech makerspace DF LAB. Co-founded a startup Tandeems.

Daniels - a final year master's student at the University of Latvia's Computer Science Faculty with over 5 years of experience in data processing and almost 2 years of experience in the cyber security industry.

Peter - Third-year student at the University of Latvia's Computer Science Faculty, with a year of experience in cybersecurity and 3 years of experience in the Latvian National Guard. Particular interest in unique solutions to unique problems.

Katrīna - a freshman at the University of Latvia's Computer Science Faculty with an interest in networks and cybersecurity and additionally helps out as an assistant at the LU DF MedTech.

Link to Presentation and Visual Materials:

https://docs.google.com/presentation/d/1vCvhZUAAdhqCragUjt1lTufshicAZYfft1fAqOaww2E/edit#slide=id.g22660e3a3b3_0_200