We are currently working on...
At Digiteq Automotive, we contribute to the development of the software stack E3 2.0 (aka One SW stack) which will include a unified operating system for vehicles from the Volkswagen Group brands. One of the key features will be Autonomous Driving Level 4 readiness, meaning that customers can fully hand over the steering to the car. As a 100% member of the Volkswagen Group, we participate in the most interesting projects of the Group’s various brands in the following fields:
We are developing in all areas of IT, from series SW running on high performance machines with Linux, to programming cloud applications, frontend, and DevOps. SW Engineering as a discipline in the Automotive industry grew out of embedded SW development, running on small ECUs. Gradually, the solutions became more complex.
Nonetheless, Series grade SW, which runs in cars, is our main focus, and to work on it, the following know-how areas are crucial:
- SW Architecture
- Requirement engineering
- SW Development
- SW Testing and Validation
- Validation Frameworks
- SW Quality Assurance
Each area of SW development also requires a certain level of domain know-how, ranging from knowledge of communication protocols, security concepts, vehicle diagnostics, to control systems, computer vision, and AI.
We have one specific goal, and that is to develop a new robust SW platform for all vehicles from VW Group brands. This platform or software stack is called VW.OS, and will include a unified operating system and a set of services. Another key feature will be level 4 readiness, meaning that customers can fully hand over the steering to the car.
Overall, there are many big and interesting challenges - from choosing the right SW and System architecture, making the stack as HW independent as possible, putting together a multitude of necessary OS services, supporting features with big computation demands, as well as seamlessly integrating into the cloud.
Challenges in SW Engineering lie not only in solving complex technical projects, but larger projects also require efficient and lean infrastructure that help the engineers in development.
SW project infrastructure
Each project has a similar goal, which is to automate and optimize all aspects of the project. Having Continuous Build, Integration Test, and Delivery is a vital paradigm that helps teams achieve efficiency and focus on developing features.
Autonomous driving platform (ADP)
Tasks that focus on machine learning lead to a large amount of data that must be used for the training and evaluation of neural nets. Infrastructure that handles this high load is being developed, utilizing the MS Azure and MS ML Frameworks.
Beyond automated build and delivery, SW projects also require a multitude of custom tools that are being developed to help achieve the project goals. This includes, for example, Data labelling, Requirement Traceability, ARXML parsing, and other tasks.
Throughout the SW landscape of the different projects, it is essential that we in Digiteq can rely on the state-of-the-art DevOps infrastructure that we provide all of the SW teams across our project and product portfolio.
Sensor Data and Control
An important part of autonomous driving is to understand the scenery and environment around the car itself. But there is another step to consider; there must be appropriate action that reacts to environment changes in a very short time.
New cars, with the potential to be autonomous, use numerous sensors, based on different physical phenomena - ultrasonic, camera, laser, radar, etc. Every sensor type brings its own pros and cons. Our role is to fuse the data together, using Kalman (or other) Filters, to lower uncertainty to a bare minimum.
Camera systems have always been very important for our company. We focus mainly on tasks related to video perception, such as multi-camera system calibration, stereovision and hazard detection, structure from motion, free-space detection, and deformation models.
When a car understands its surroundings, it needs to take actions. At first glance, it looks like an easy task; in reality, the opposite is true. The design of precise and robust controllers is the key, along with the optimization of the car’s trajectory. For both topics, we use state-of-the-art approaches to reach as smooth/pleasant as possible an experience for the passengers.
Leadership and Teamwork
We grow, we learn, we have fun. Our true values are openness, teamwork, and innovation. Moreover, our solutions are driven by Scaled Agile Framework (SAFe). To ensure our mindset stays the same, we strongly support true leadership (not only as a buzzword 😊). We enable people to grow in their strong competencies, and support knowledge sharing, using chapters. Perfect Product owners and Scrum Masters are the key to having great teams, as well as focus on valuable delivery.
- Product Owner
- Scrum Master
- Agile Coach
- Chapter Lead
- Business Owner
Car perception is an area in which we apply artificial intelligence approaches the most. Since automotive requires safe and secure algorithms, our biggest challenge is certainly the validation of machine learning algorithms. Precise object detection and classification are essential, as well as a perfect understanding of the scene. Deep Learning helps us enormously and brings promising results. Especially the use of state-of-the-art methods, such as CNNs, encoders, and multi-task learning.
- Object detection
- Semantic segmentation
- Active learning
- NN Architectures and Optimization
- Data operations
Validation and Testing
When you think of SW Testing, you might imagine clicking on web page menus and evaluating button colours. SW Testing in the automotive industry is much more challenging.
The creation of automation scripts and the development of testing frameworks are the most important parts of our daily work. Our professionals also focus on test architectures. It sounds simple, but imagine a situation where you have to create a concept to test or validate algorithms for computer vision algorithms. Algorithms you developed without a car, in virtual reality.
Autonomous driving brings many challenges. One that we try to tackle is how to validate algorithms that change, based on data. A good example could be the validation and interpretability of neural networks. There are many papers and research projects available, nevertheless, research is just one part, and the application of algorithms to the car itself is another.
In situation, we already have a sample of HW and sensor system testing, which is starting to play a prime role. Preparing and designing all of the electronic gadgets to simulate real rain, as well as small steps forward. Hardware-in-the-loop design and then testing is the final phase, before we test everything in a prototype car.
All of those activities have to be tracked and discussed with the researchers, developers, and system architects, in order to deliver as safe and reliable a product as possible. This means that we focus heavily on test management. Our test managers are not suppliers to development; they are partners.
Our Dev Stack
Our Brains are busy with
connected cars ?
of next release? ?
of data per day ?
of CNN Models ?