Apps that help to fix the problems of your customers

The calculation does not have a certain correct solution, but is potentially infinite. This means that the problem had to be broken down into individual possible solutions and then defined what the ideal solution could look like. Successively, several different strategies of the loading process were implemented, whereby the most effective one was ultimately chosen.

Škoda Auto provided various 3D trunk models for all supported vehicle models. These had to be converted into a uniform form. The first challenge was to standardize the documents in a format that could be used for calculations:

• The trunk of the car was divided into virtual small cubes with an edge length of 5 cm, which proved to be ideal in tests. Most of the luggage is loaded into this net and the calculation is faster than if something is manually stowed in the trunk. Computer scientists and mathematicians visualize this in the form of the so-called backpack problem.

Since the calculation using an app should not take longer than the calculation in the head, the algorithms had to be optimized and a balance between accuracy and speed had to be found.

• The load in the trunk is calculated in four ways. The most effective method of loading is then displayed directly in the mobile device using vector masks.

The input of the dimensions of the individual items should not degenerate into boredom. Therefore, ways were considered how to simplify and facilitate the associated routine work.

• Users can enter the baggage dimensions and any information on the fragility or softness of the load into the app themselves. However, for some packages - such as those for furniture kits - the dimensions are already given. In this case it is sufficient to pull out the camera and leave the input of the dimensions to the intelligent text recognition.

For the representation of the objects in the cargo area of the car, it had to be assumed that the respective cargo area with an open trunk was merely rendered in 2D. Therefore, a special way of calculating the perspective and projecting the objects into the 2D space was developed - including masks for displaying the hidden parts.

The models of the interior of the cargo area are provided in the standard triangular format. The task was to devise a discrete format suitable for the calculations, including a transformation of the coordinate system suitable for representation and calculation.

A mere manual numerical input of the size of the objects would be insufficient and not user-friendly. For this reason, the input of an object is based on visualization and animation, and it is even possible to handle the object including its size adjustment by means of multi-touchgest. With the help of OCR, the app can recognize an object and rewrite its size. The user has the possibility to measure the object with the help of AR, if he does not have a measuring stick at hand or does not know the exact size.

Technical facts

•  iOS already online as part of the My Škoda App, Android on the road
• The version for Android is based on MVVM architecture. The coding was done in Kotlin. The app for iOS is kept in Swift in MVVM-C architecture. Since Load-in is currently part of the app MyŠkoda, everything has been packaged into a library for use in other apps using common package management systems.
• The algorithm running on the backend was developed in the Python language, which has a very good set of tools for working with scientific computations and which allows fast iteration. For the final solution, the code is also translated into C using the Cython tool, which makes the calculation even faster. All in all, this runs on Azure Cloud - with intelligent scaling, any number of users can be served.
• To handle the trunk packing problem, several strategies were successively used - Residual Space and Minimal Space, both in two variants: lean and brute-force.

Where is the data stored?

The data is on-premise at the end customer