Use cases

At the moment, Sonae has a self-scanning app for customers in the loyalty program, Continente Siga, that allows users to scan products and do self-check-out in the store, among other features.

The goal, within the PEER project is the development of a personalized in-store picking assistant in the hyper-market stores using Siga APP as the interface.

This could be done, for example, with the development of an algorithm that estimates the location of products in-store and calculates optimized and personalized routes with turn-by-turn navigation.

Besides this, the assistant could also suggest items throughout the customer journey, improving their overall shopping experience.

At the moment, Proditec inspection machines are based on “classical” image processing to detect defective tablets or capsules. For each product, a “recipe” has to be created by adjusting a whole set of parameters and thresholds.

This task can be perceived as too long and as a source of mistake, as the users may not be skilled enough to play with algorithm parameters.

When trying to use deep-learning models instead of “classical” algorithms, this is the building of the model that can be a challenge for the users. The goal within the PEER project is the development of an intuitive AI assistant in order to help the user achieve the best recipe - or deep-learning model - that will get the best out of the machine sorting capabilities.

One of the most important steps in the tire production process is extrusion where rubber is pressed into a chamber that heats the rubber and forces it through a so-called die to produce the desired shape or profile. Extrusion machines are quite big - the material traverses several hundred meters from one end to the other end of the extrusion line.

During this traversal the material continuously changes its consistency and shape. At the end of the line the material is checked whether it matches certain quality criteria on several properties. If not, a material is classified as scrap and needs to be discarded or re-worked. To tackle this problem, we plan to create an AI algorithm that can help the operator by predicting a scrap event occurring at the end of the pipeline already at the very beginning of the pipeline and suggesting ways to resolve the quality issue.