The automotive industry has mobilized the global economy for decades. German automobile manufacturers (OEMs) alone employ more than 1 million people worldwide and generate sales of more than USD 500 billion. Since a Google and Stanford team won the Darpa Self-Driving Vehicles Challenge 2006 with the help of machine learning, the industry has been undergoing rapid change. Machine learning opens up brand-new business models, from autonomous driving to smart production to personal assistance in the car. However, the use of machine learning requires a different infrastructure than that found in traditional OEMs. Technology-first companies like Waymo or Tesla threaten to overtake established OEMs with billion-dollar market capitalization. Autonomous vehicles produce terabytes of data every day. This data can be immensely valuable in developing machine learning-driven functions. However, substantial challenges remain in the way of using this data. Visit this talk to hear about these challenges to help turn the automotive industry from a mechanical engineering to a software industry.

#UseThePlatform
#StencilJS
#Svelte JS

Our frontends have become increasingly more complicated. More and more business logic is moved from back- to frontend. Frameworks like Angular are great for this paradigm. They offer great support for routing, dependency injection, testing, etc. But – as with all things in life – these benefits don't come for free. By relying on these frameworks we actually depend on it. Our code cannot run without them, and framework upgrades may sometimes break it. Also, these frameworks are often heavy in size, slowing down our applications.WebComponents on the other hand offer a framework agnostic and lightweight alternative. Modern compilers like StencilJs or SvelteJs help us to keep most of the benefits mentioned above. In this talk, we will migrate a typical angular application into WebComponents and discuss the advantages and disadvantages.

Now that we've launched our shared mobility platform, with our first customer, a fully electric moto sharing service in Barcelona, we want share our experience! Learn more about our journey towards creating "the best shared mobility platform ever" - we'll share our lessons learned about the culture, technologies and agile methods that have brought us this far.

Do you believe that a team can make all technical, functional and business decisions on its own and create great, high-quality products without being controlled by people with specific roles such as a software architect or team leader? In this talk, we will take a tour of Software Development in a balanced team where people can be open, listen to each other and trust each other. The result of this way of working is high-quality software, created by extremely motivated people who take pride in their work and the resulting products.

Today’s vehicles have transformed from a means of transportation with a few in-vehicle electronic control units into internet-connected datacenters on wheels that capture and process volumes of data about the vehicle and its surroundings. The autonomous car needs to be kept as secure and up to date as any other connected system.

SUSE and Elektrobit have joined forces to provide a future software platform for automobiles that fulfills all the key requirements around openness and transparency, security and safety, and seamless system updates over the air. All backed up by a broad open source community that provides constant innovation and a large talent pool.

We’ve combined SUSE’s almost three decades of experience with hardening and maintaining open-source software for the most mission-critical use cases with Elektrobit’s expertise in building software solutions that meet and exceed the requirements of the automotive industry.

Daniel Siegl from LieberLieber Software, together with Hermann Gollwitzer from Volkswagen consider and show how MBSE is introduced at Volkswagen to develop software on the basis of reference-architecture. Fundamental interrelationships for cooperation in the multi-layered development environment for the benefit of all participants and for the development of sustainable, constantly expandable solutions are presented. Additionally argued is how Automotive SPICE requirements are perforable in a related setup as an example and how configuration management can be realized for models.

After the rise of self-driving cars and electric mobility: What about the infrastructure? Will the future of mobility also demand that the industry builds its own infrastructure?

Based on real-world projects at Porsche we will show you how we make use of Spring WebFlux and Kotlin in an environment of multiple legacy systems to slip through their natural boundaries. You will get an insight into our process of choosing a technology stack which resulted in Spring WebFlux. Expectations on our side have been to reduce resource consumption, increase efficiency, maintaining a high level of abstraction in our code and having a better way of dealing with the inherited long response times of systems we depend on. Furthermore, bundling Spring WebFlux with Kotlin allows us to write code in a very expressive and functional style, making our code more readable and maintainable. We will explain our learning curve with some examples and address some of the common pitfalls we faced. Join us as we share our ups and downs when expectations meet reality.

Continuous delivery is not a new concept anymore. It has been already in many technical talks for quite a while. Usually, these talks focus on standard software development: where the deliverable is a piece of software. This deliverable can be presented as an installable or deployed to be consumed as a service. When we speak about Data Pipelines data is considered the deliverable, which implies some changes in the continuous delivery process. In this talk, we will present the technical decisions and design that we followed in a recent project to be able to apply continuous delivery into a Data Processing pipeline.

In this presentation, we demonstrate the challenges of validating a computer vision (CV) pipeline that runs inside the car and discuss how the task of validating such a pipeline can be solved by using Artificial Intelligence outside the car. It is too expensive to manually label enough reference (ground truth) data for validating the CV pipeline inside the car. One solution is to train a more powerful CV Deep Learning system that runs outside the car on more powerful GPUs and not necessarily in real-time and can provide reference object detections. The latter are compared to the results the algorithm inside the vehicle would produce. The reference deep learning system needs training data to be trained upon. The process of acquiring this initial training data can be optimized by semi-automatic annotation.

Presenting different cases and main challenges at indoor localization and what are the possible solutions. Technical overview of the main technologies, comparison(pros and cons) plus real examples from our experience working with them on various customers – WiFi analytics, Beacons, Image Processing and Sensors.

We at Transmetrics solve a variety of planning problems for the Logistics Industry. We tackle issues that are typically performed by human experts, and they are responsible for them. After we model these problems and propose a solution, the planners often have some hard-to-quantify knowledge, way of thinking, or biases the models don’t know. Thus a proposed solution could be looked at with skepticism, even if it formally satisfies the requirements and minimizes cost.

We work on some methods of letting the user result from an optimization problem and explore neighboring solutions. This way, the model can consider new information and preferences so that they are more inclined to use it.

In this talk, we discuss integrating UI and optimization techniques to help human planners have more trust in our optimization models.

Developing a robust machine learning code is hard, but debugging it is even harder! When developing machine learning (ML) solutions for complex problems such as autonomous driving we tend to focus on the hard research problems and develop theoretical models whose implementation we take for granted. In practice, however, bugs always creep in. How do you detect bugs when working with multidimensional arrays containing millions of parameters? How do you identify sources of error when building dynamic computational graphs? In this talk, I will provide an overview of the existing methods for debugging ML code and showcase the first of its kind visual 3D debugger which makes debugging deep learning models a breeze.

We tend to think of advancing mobility in terms of highly optimized software and enterprising development – which it is. However, many don’t realize that the topic of mobility hinges on intelligent applications of the mathematical bases for route optimization. At the root level, this began with the Traveling Salesman Problem, better known as TSP. While working with TSP provides answers on pathfinding between two points, mapping these distances onto conventional streets, involving traffic and other factors, presents extra challenges. Here, we’ll discuss these essential concepts at work behind Matrix Routing and show you how you can use TomTom’s resources to build an application for food delivery.