Reference Projects

SmartBee

How did the project “Smart Bee” start?

A colleague at EFS, learnt from his Father-In-Law a beekeeper by profession, about the Varroa Mite being the main reason of bee colony collapse. These mites were introduced to Europe from their indigenous environments and do not have any natural enemies here. The European bees are incapable of fighting these mites and that is why these parasites multiply here exponentially. The bees however, are capable of surviving in extreme temperatures, whereas a Varrao Mite dies above a temperature of 42 degrees Celsius. EFS plans to help the beekeepers solve this problem, using its engineering perspective.

How does this project fit EFS Profile?

EFS occupies itself typically with safety-critical software used to secure critical situations and with the integration of autonomous controlled systems. This knowledge can also be used for the Mite problem and enables us to find an elegant technical solution. EFS has amassed expertise over several important components, in order to bring newly developed technologies like a safety-critical software onto a car. This project requires just a simple yet efficient technology to be implemented for a beehive.

What is the solution?

We plan to use the increase of temperature in the form of a Bee-Sauna. In the circle of experts, the process is called “Hyperthermia”. For the implementation of this Bee-Sauna, we need knowledge from all the different fields like Software Development for Control Systems, Data Analysis, Data Management and also, Security components.

Image Detection / Deep Learning for Image Description

Problem Statement:

Our Data management division has been working together with our Technology center division to develop an algorithm for classification of different automobile parts.

As a part of the documentation procedure in the field of corrosion research, a lot of pictures are taken of automobile parts under different lightings and in different perspectives, such that approximately 70000 pictures have been taken up until now. In order to study the corrosion effects, it was required to classify all these different automobile parts, which would manually require approximately 300 hours.

Our Approach:

We built a neural network to achieve the classification. The neural network was trained to identify different parts of a car and to perform classification, based on the location of the identified part on the car. For example, the engine space, car rear, occupant compartment. Subsequently, the pictures were further processed and optimally prepared for the neural network. The neural network could identify and classify these pictures with 80% accuracy.

One Step Ahead:

We implemented and run this neural network model on a development ECU. This step was taken as a solution for the future, where the pictures can be uploaded directly on a Web-App, after being taken and can be immediately classified and labelled. For this purpose, the recording device source has also been kept flexible (Computer, Mobile, Tablet or camera). The user can take a picture with an integrated camera which will then be classified within few seconds. The user receives a result in the form of four most probable categories and their respective probabilities. In the end, one also has the option to check the classification and in case of an error, to correct it. The neural network then identifies its mistake and improves itself continuously on the run.

 

 



Group E-Bike

Initial Stage

In the year 2014, EFS started a cooperation project with Audi and Volkswagen, to develop a modular platform model set for innovative E-Bikes with never before seen features, which would be common to everyone under the VW-Group. Besides implementing a number of unique features, the focus lied in the transfer of our well established competence in modern car development to the branch of E-Bikes development and to create a completely new driving and safety experience. Another important aspect of this project was to generate interest in the public and introducing them to this new and alternative concept of mobility with the help of comprehensively designed simulators.

Details

In this project, EFS was responsible for the concept, architecture and development of a comprehensive electrical and electronic network for a bicycle system, including their respective software building blocks. Together with a team of 26 partners, we determined the construction and network parameters of individual components, sensors/actuators and ECU’s and also built a prototype, with all these components on an electrical wire harness.

In order to get a realistic picture of the realization and perspective aspects of such a bicycle building system, the development of the prototype was performed as close as possible to the series development of such bikes. Thanks to EFS’s years of experience in the field of automotive hardware and software development ranging from pre-development to series production, and in accordance with the standards of the automotive industry, we were able to design and develop an ECU for the bike, within a very short period of time.

The usage of multicore processors, production OS and AUTOSAR conform SW development led to a solid fundamental for this product, which also helped in the realization of never before seen safety related key features, besides the connectivity services like display and smartphone connectivity. In addition to this, with the help of our cooperation partners in the already established branch of automatic gear mechanism in the automobile Industry, where gear changing strategies can be tuned with the help of certain parameters, a customized dynamic shock absorber control and world’s first bicycle-ABS for target domains was developed. Both these systems were integrated in the system by EFS. The safety and stability of the system could be increased multifold, by detecting the blocking of front and rear tires which helped in the detection of a threatening increase in the Rollover and its control. Another unique safety feature of the system which was developed and patented is the strategy of calculation of wheel speed for bicycles in extreme situations. The use of magnetorheological shock absorbers leads to an increase in the comfort factor. The required damping factor can be calculated based on the driving state, ground conditions and the selected driving program applied in a matter of milliseconds.

EFS’s expertise in the development of control systems and strategies could be applied in all the above mentioned fields.

Result

Three different ready to drive prototypes were successfully developed by the end of this project with different branding, on which one can experience all the above mentioned as well as many other innovative features.

Pessimizer: To the sun and back

How long would it take to travel to the sun and back in a car? And that too 88 times?

The distance between the sun and the earth is approximately 150 million Kilometers (one way), so with a return trip it is about 300 million Kilometers. This total when multiplied by 88, results in 26,400 Million Kilometers. This distance when transformed to the number of travelling hours in a car, gives us the amount of time an autonomous function has to be tested, in order to statistically call it a secure function.

To clock so much time, was a tad bit too much for our colleagues at EFS, and so they looked for solutions, to secure a driving function in a more faster and uncomplicated way. Current projects are already leading in an increase in acute requirements to safeguard Car Dynamic controllers in the automotive field. At EFS a numerical approach was implemented in a product which would then be used for the security of trajectory following controllers and secure development of highly automated driving functions. In the future, such controllers can be secured in a given parameter space with the help of Pessimizer..

The project started last year in the summer. This whole project is taking place in cooperation with University of Lübeck and especially with one of our past colleagues Prof. Dr. Georg Schildbach.

EFS provides its customers a basis for safeguarding of driving functions with the help of Pessimizer. Consequently, improving customer satisfaction and thereby, setting itself apart from its contemporaries and winning new customers.

All about Pessimizer

Due to the newness of the field of autonomous driving, it’s safety is a controversial and intensively discussed topic. A newly developed driving function should cover a lot of different details to ensure its safety. With reference to statistical safety, a function must first be tested for a long range of kilometers. This can be reduced with the help of Pessimizer.

Pessimizer deals with the concept of securing a trajectory following controller. Due to the non linearity of such a controller and effects of different external influences, the ensuring of such a complex function can be a challenge. Classical methods such as those based on requirements, can therefore, not be used as the only proof of security of such a function. This is where Pessimizer comes into play.

Pessimizer creates a number of different critical trajectories, which are then tested under simulation together with the trajectory controller. The Pessimizer tries to generate large deviations from the reference path during these simulation runs. These deviations can be achieved for example, by producing multiple curves in the reference path, thereby, trying to destabilise the trajectory controller. The goal is to test the controller intensively under such difficult conditions, to optimize and to evaluate its security under the applied parameter space.

With the help of Pessimizer we can provide a customer with the assurance, that the developed controller is capable of staying stable for a range of trajectories and inside the given limits. Due to the targeted optimization of the controller for corner cases, with the help of Pessimizer, the functional capability of the controller can be ensured for the given parameter space. A hundred percent guaranteed stability of the controller is not possible, however, the stability for a number of critical maneuvers in the given space can be guaranteed.

The idea of Pessimizer or its sale as a product for our customers is not planned yet, but could be take place in the future. Right now, the Pessimizer is being implemented in our internal projects, to gain more experience with it.

For those, who want to know more details:

The idea of Pessimizer comprises of the combination of simulation of a car and the trajectory following controller with an optimization step. In order to achieve this, systematic sequence of trajectories and external influences for example, cross winds, weight distribution, road conditions like bank angle and gradients are generated. Instead of maximizing the performance of the trajectory following controller, the Pessimizer tries to identify a scenario with the worst performance from the controller. As criteria for the optimization, measurements such as lateral deviation of the car from the given trajectory is used. The optimization function is additionally fed with certain constraints. Constraints such as maximum curvature, car velocity and lateral acceleration and other parameters are selected, so that the constructed scenarios are realistic for the solution space and also, realizable for the cars.