Robert Bosch (Taiwan)

Deep Learning has enabled remarkable progress over the last years on a variety of tasks, such as image recognition, speech recognition, and machine translation. One crucial aspect for this progress are novel neural architectures. Currently employed architectures have mostly been developed manually by human experts, which is a time-consuming and error-prone process. Because of this, there is growing interest in automated neural architecture search methods. We provide an overview of existing work in this field of research and categorize them according to three dimensions: search space, search strategy, and performance estimation strategy.

This review presents the synthesis, characterization techniques, processing and potential applications of silicon-based ceramic materials derived from organosilicon polymers. The Si-ceramics are prepared by thermolysis of molecular precursors. The influence of the initial molecular structure of the precursor on the properties of the final ceramic material and its applications is discussed. The thermolytic decomposition of suitable Si-based polymers provides materials which are denoted as polymer-derived ceramics (PDCs). In particular, this procedure is a promising method for the preparation of ternary and multinary silicon-based ceramics in the system SiCNO. There is no other synthetic approach known to produce e.g. SiCO or SiCN based ceramics. In the case of PDCs route, common preceramic polymers are poly(organosilazanes), poly(organosilylcarbodiimides) and poly(organosiloxanes). One basic advantage of the PDC route is that the materials can be easily shaped in form of fibers, layers or bulk composite materials by applying processing techniques established in the plastic industry. The PDCs in general exhibit enhanced thermomechanical properties, i.e., temperature stabilities up to approximately 1500°C. Recent investigations have shown that in some cases the high temperature stability in terms of decomposition and/or crystallization can be increased even up to 2000°C if the preceramic polymer contains some amount of boron. The composition and microstructure of the PDC are a result of the molecular structure of the preceramic polymer. Therefore, the observed differences in the macroscopic properties are also closely related to the variation of composition and solid state structure of these materials.

There is much current interest in identifying the anatomical and functional circuits that are the basis of the brain's computations, with hope that functional neuroimaging techniques will allow the in vivo study of these neural processes through the statistical analysis of the time-series they produce. Ideally, the use of techniques such as multivariate autoregressive (MAR) modelling should allow the identification of effective connectivity by combining graphical modelling methods with the concept of Granger causality. Unfortunately, current time-series methods perform well only for the case that the length of the time-series Nt is much larger than p, the number of brain sites studied, which is exactly the reverse of the situation in neuroimaging for which relatively short time-series are measured over thousands of voxels. Methods are introduced for dealing with this situation by using sparse MAR models. These can be estimated in a two-stage process involving (i) penalized regression and (ii) pruning of unlikely connections by means of the local false discovery rate developed by Efron. Extensive simulations were performed with idealized cortical networks having small world topologies and stable dynamics. These show that the detection efficiency of connections of the proposed procedure is quite high. Application of the method to real data was illustrated by the identification of neural circuitry related to emotional processing as measured by BOLD.

<div class="htmlview paragraph">VDC is a new active safety system for road vehicles which controls the dynamic vehicle motion in emergency situations. From the steering angle, the accelerator pedal position and the brake pressure the desired motion is derived while the actual vehicle motion is derived from the yaw rate and the lateral acceleration. The system regulates the engine torque and the wheel brake pressures using traction control components to minimize the difference between the actual and the desired motion. Included is also a safety concept which supervises the proper operation of the components and the software.</div>

We introduce the SE(3)-Transformer, a variant of the self-attention module for 3D point clouds and graphs, which is equivariant under continuous 3D roto-translations. Equivariance is important to ensure stable and predictable performance in the presence of nuisance transformations of the data input. A positive corollary of equivariance is increased weight-tying within the model. The SE(3)-Transformer leverages the benefits of self-attention to operate on large point clouds and graphs with varying number of points, while guaranteeing SE(3)-equivariance for robustness. We evaluate our model on a toy N-body particle simulation dataset, showcasing the robustness of the predictions under rotations of the input. We further achieve competitive performance on two real-world datasets, ScanObjectNN and QM9. In all cases, our model outperforms a strong, non-equivariant attention baseline and an equivariant model without attention.

<div class="htmlview paragraph">The selective catalytic reduction (SCR) based on urea-water-solution is an effective technique to reduce nitrogen oxides (NO<sub>x</sub>) emitted from diesel engines. A 3D numerical computer model of the injection of urea-water-solution and their interaction with the exhaust gas flow and exhaust tubing is developed to evaluate different configurations during the development process of such a DeNOx-system. The model accounts for all relevant processes appearing from the injection point to the entrance of the SCR-catalyst:</div> <div class="htmlview paragraph"> <ul class="list disc"> <li class="list-item"><div class="htmlview paragraph">momentum interaction between gas phase and droplets</div></li> <li class="list-item"><div class="htmlview paragraph">evaporation and thermolysis of droplets</div></li> <li class="list-item"><div class="htmlview paragraph">hydrolysis of isocyanic acid in gas phase</div></li> <li class="list-item"><div class="htmlview paragraph">heat transfer between wall and droplets</div></li> <li class="list-item"><div class="htmlview paragraph">spray/wall-interaction</div></li> <li class="list-item"><div class="htmlview paragraph">two-component wall film including interaction with gas phase and exhaust tube</div></li> </ul> </div> <div class="htmlview paragraph">The single modeling steps are verified with visualizations, patternator measurements, phase-doppler-anemometer results and temperature measurements. CFD simulations of a SCR DeNOx-system are compared to experimental data to determine the decomposition parameters for urea-water-solution droplets. Numerical results for an injection including all processes addressed above are discussed.</div>

<div class="htmlview paragraph">Passenger car DI Diesel engines need a flexible fuel injection system. Bosch develops a common rail system for this purpose. Besides variation of fuel quantity and start of injection, it permits to choosing freely injection pressure inthe rangeof 150 to 1400 barand injecting fuel in several portions. These new means will contribute to further improvements of DI engines concerning noise, exhaust emissions and engine torque.</div>

Our study draws on institutional and signalling theories to postulate relationships between board characteristics and corporate reputation. Based on a sample of 324 firms featured in Fortune's list of most admired corporations in the USA, our findings indicate that board characteristics significantly influence the assessment of firm reputation by the business community. Specifically, we found that firms with a greater proportion of outside directors and those with larger boards exhibited better reputation than those with smaller boards and a higher proportion of insiders. In addition, we observed an inverted‐U relationship between the average tenure of outside directors and corporate reputation. However, contrary to expectations, our findings indicate a negative association between independent leadership structure (i.e. absence of duality) and corporate reputation.

Contemporary office buildings commonly experience changes in occupancy patterns and needs due to changes in business practice and personal churns. Hence, it is important to understand and accurately capture the information of such trends for applications in building design and subsequent building operations. Detection of occupant presence has been used extensively in built environments for applications such as demand-controlled ventilation and security, and occupancy profiles are widely used in building simulations. However, the ability to discern the actual number of people in a space is often beyond the scope of current sensing techniques. This paper presents a study to develop algorithms for occupancy number detection based on the analysis of environmental data captured from existing sensors and ambient sensing networks. Both wireless and wired sensor networks are deployed in the Robert L. Preger Intelligent Workplace (IW) at Carnegie Mellon University, comprising six different types of sensors. An average of 80 % accuracy on the occupancy number detection was achieved by Hidden Markov Models during testing periods. The findings also offer encouraging possibilities for incorporating the algorithms into building management systems for optimizing energy use while maintaining occupant comfort.

<div class="htmlview paragraph">Since its introduction in March 1995, the market demand for Vehicle Dynamic Control systems (VDC) has increased rapidly. Some car manufacturers have already announced their plans to introduce VDC on all their models. Particularly for compact and subcompact cars the system price needs to be reduced without sacrificing safety and performance. Originally designed for optimal performance with economically feasible components (sensors, hydraulics and microcontrollers) and using a unified control approach for all vehicle operating situations the system has been extended to include various drive concepts and has continuously been improved regarding performance, safety and cost. This paper describes the progress made in the development of the Bosch VDC system with regard to the design of the hydraulic system, the sensors, the electronic control unit, the control algorithm and safety.</div>

Car-to-car and car-to-infrastructure communications are considered as a key technology for safe and intelligent mobility in the future. In the German project sim <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">TD</sup> , <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">1</sup> car-to-x communication is shifting from a pure research topic toward a first deployment of such a system: the goal of sim <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">TD</sup> is to test car-to-x applications in a real metropolitan field trial. The sim <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">TD</sup> scenario includes vehicles, roadside units, and infrastructural facilities for traffic and test management. Besides these domains, several third parties are involved to provide access to additional services. The main communication partners are furthermore distributed over a wide area including highway, suburban, and urban scenarios. As a result, such a system requires a commonly accepted architecture of the individual components and a seamless communication network for reliable and efficient information interchange. In this article we propose the sim <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">TD</sup> system architecture and the components necessary for successful deployment in the large-scale field trial.

&lt;span class="label"&gt;1&lt;/span&gt; &lt;div class="htmlview paragraph"&gt;High precision high pressure diesel common rail fuel injection systems play a key role in emission control, fuel consumption and driving performance.&lt;/div&gt; &lt;div class="htmlview paragraph"&gt;Deposits have been observed on internal injector components, for example in the armature assembly, in the slots of the piston and on the nozzle needle. The brownish to colourless deposits can adversely impact driveability and result in non-compliance with the Euro 4 or Euro 5 emission limits.&lt;/div&gt; &lt;div class="htmlview paragraph"&gt;The deposits have been extensively studied to understand their composition and their formation mechanism. Due to the location of these deposits, the influence of combustion gas can be completely ruled out. In fact, their formation can be explained by interactions of certain diesel fuel additives, including di- and mono-fatty acids. This paper describes the methodology used and the data generated that support the proposed mechanisms. Moreover, approaches to avoid such interactions are discussed.&lt;/div&gt;

&lt;div class="htmlview paragraph"&gt;As new smart systems for passenger cars are assisting me driver to handle manoeuvres in critical and normal situations, brake systems are required to fulfill the compatibility and interface demands. These advanced brake systems will be operated in a remote mode during normal braking and for autonomous brake interventions. BOSCH is developping a brake-by-wire system on a hydraulic basis, called ‘Electrohydraulic Brake EHB’. Brake pressure buildup is supplied by a high pressure accumulator. Generation of the high pressure is done by an electric motor driven pump, similar to current ABS-systems.&lt;/div&gt; &lt;div class="htmlview paragraph"&gt;Pressure at the wheel brakes is individually controlled by closed-loop pressure control, consisting out of inlet, and outlet valves, pressure sensor and corresponding algorithm. It is specified, that this control must be completely noiseless, proportional, fast, and highly accurate. To raise the acceptance of such a system, it will be introduced with a conventional hydraulic backup. The backup actuates the front wheel brakes. In the normal operating mode the master cylinder is switched to a hydraulic pedal travel simulator to give the right feeling and sensitivity at the brake pedal.&lt;/div&gt; &lt;div class="htmlview paragraph"&gt;The system comes together with ABS, ASR, and VDC functions, optimized by using the wheel brake cylinder pressure information and proportional brake pressure control. It incorporates electronic brake force distribution between front and rear and even left and right, thus improving stopping distances and stability, making better use of the rear brakes than conventional systems.&lt;/div&gt; &lt;div class="htmlview paragraph"&gt;It can be shown that ABS and other regulations can be done fully hidden for the driver. No noise from the pressure control or pedal reactions are noticed.&lt;/div&gt; &lt;div class="htmlview paragraph"&gt;Autonomous vehicle guiding systems, such as advanced cruise control, collision avoidance (assist) systems, necessary for Intelligent Vehicle Highway System IVHS, and functional upgrading like hill-bolder systems, and parking aids will have an ideal brake basis to act on.&lt;/div&gt; &lt;div class="htmlview paragraph"&gt;Further concepts of integrating various other drivetrain and comfort systems will have a brake system that fulfills their needs. Functional enhancement can be added to the brake system with minimal hydraulic modifications.&lt;/div&gt;

&lt;div class="htmlview paragraph"&gt;Common Rail provides additional flexibility for the design and application of a diesel injection system.&lt;/div&gt; &lt;div class="htmlview paragraph"&gt;Contrary to conventional injection systems pressure generation and injection are decoupled in the common rail system.&lt;/div&gt; &lt;div class="htmlview paragraph"&gt;The injection pressure can be selected independent of engine speed and injected fuel quantity within certain limits.&lt;/div&gt; &lt;div class="htmlview paragraph"&gt;The fuel combustion and the corresponding noise can be improved by increasing the fuel pressure up to 1400 bar and introducing pilot injection or multiple injection.&lt;/div&gt; &lt;div class="htmlview paragraph"&gt;Furthermore the common rail system can replace conventional injection systems without requiring major engine modifications.&lt;/div&gt; &lt;div class="htmlview paragraph"&gt;BOSCH will provide this new injection system for the whole range of applications from light duty (30 kW per cylinder) to heavy duty vehicles (50 kW per cylinder).&lt;/div&gt;

&lt;div class="htmlview paragraph"&gt;The development of future engine generations for Gasoline Direct Injection requires sophisticated combustion systems to reach reduced fuel consumption and future emission standards. The design process of these combustion systems has to be based on a fundamental knowledge of the interacting mixture preparation mechanisms. Beside the air motion inside the cylinder mixture preparation is mainly feeded by the fuel spray quality, injector performance respectively. The article therefore presents a fundamental analysis of the GDI mixture preparation and affords an insight into the injector development. Comprehensive experimental studies were performed in high pressure/temperature vessels using Phase Doppler Anemometry, Laser Induced Fluorescence and video techniques to define the significant fuel spray features for GDI. CFD-calculations were additionally applied to study the temporal behavior of the mixture preparation under injection parameter variation. The adjustment of the spray and injection parameters is described under the view of engine application requirements. Using the results from these fundamental studies an appropriate injector design could be developed very efficiently.&lt;/div&gt;

Deep learning methods have resulted in significant performance improvements in several application domains and as such several software frameworks have been developed to facilitate their implementation. This paper presents a comparative study of five deep learning frameworks, namely Caffe, Neon, TensorFlow, Theano, and Torch, on three aspects: extensibility, hardware utilization, and speed. The study is performed on several types of deep learning architectures and we evaluate the performance of the above frameworks when employed on a single machine for both (multi-threaded) CPU and GPU (Nvidia Titan X) settings. The speed performance metrics used here include the gradient computation time, which is important during the training phase of deep networks, and the forward time, which is important from the deployment perspective of trained networks. For convolutional networks, we also report how each of these frameworks support various convolutional algorithms and their corresponding performance. From our experiments, we observe that Theano and Torch are the most easily extensible frameworks. We observe that Torch is best suited for any deep architecture on CPU, followed by Theano. It also achieves the best performance on the GPU for large convolutional and fully connected networks, followed closely by Neon. Theano achieves the best performance on GPU for training and deployment of LSTM networks. Caffe is the easiest for evaluating the performance of standard deep architectures. Finally, TensorFlow is a very flexible framework, similar to Theano, but its performance is currently not competitive compared to the other studied frameworks.

&lt;div class="section abstract"&gt;&lt;div class="htmlview paragraph"&gt;The emissions legislation in the US and Europe introduces the need for the application of diesel particulate filters (DPF) in most diesel vehicles. In order to fulfill future OBD legislations, which include more stringent requirements on monitoring the functionality of those particulate filters, new sensors besides the differential pressure sensor are necessary. The new sensors need to directly detect the soot emission after DPF and withstand the harsh exhaust gas environment. Based on multi layer ceramic sensor technology, an exhaust gas sensor for particulate matter (EGS-PM) has been developed. The soot-particle-sensing element consists of two inter-digitated comb-like electrodes with an initially infinite electrical resistance. During the sensor operation, soot particles from the exhaust gas are collected onto the inter-digital electrodes and form conductive paths between the two electrode fingers leading to a drop of the electrical resistance. A controlled regeneration of the soot-detecting electrodes is conducted by heating the sensing element and burning off the deposited soot at temperatures above 600°C.&lt;/div&gt;&lt;div class="htmlview paragraph"&gt;The concept and the performance of the EGS-PM as well as a new concept for the OBD of diesel particulate filters using the EGS-PM are presented.&lt;/div&gt;&lt;/div&gt;

&lt;div class="htmlview paragraph"&gt;This paper describes the new Engine Management System (EMS) ME7.&lt;/div&gt; &lt;div class="htmlview paragraph"&gt;Torque and A/F demands for modern EMS result from both, internal functions (i.e. engine start, idle speed control, catalyst heating) and external systems (i.e. driver's request, transmission or vehicle dynamic control). With ME7 these demands are processed to the optimized actions of the actuators by a centrally coordinated torque and A/F management. The design of the functions is physically based to provide optimum portability and minimum calibration time. Examples are given for the physical manifold pressure model and the cylinder charge control of ME7 with electronic throttle control.&lt;/div&gt; &lt;div class="htmlview paragraph"&gt;The real time operating system „ERCOS” and a layer based software architecture enable the implementation of these functions in a flexible family of products for current and future systems.&lt;/div&gt; &lt;div class="htmlview paragraph"&gt;Topics, such as warm-up strategies for catalysts in conventional port injection systems, gasoline direct injection systems (with their switch-over strategies between stoichiometric and stratified operation), NOx catalyst control, and the requirements of future integrated drive train management systems, all require maximum flexibility and expandability.&lt;/div&gt; &lt;div class="htmlview paragraph"&gt;The introduction of the ME7 is an important step towards this future. The design represents a good basis for development sharing with customers and is also an important prerequisite for the vehicle management system CARTRONIC&lt;sup&gt;®&lt;/sup&gt;.&lt;/div&gt;

&lt;div class="htmlview paragraph"&gt;Bosch introduced the Common Rail Injection System for passenger car DI-Diesel engines in June 1997 for series production. Bosch is currently working on a fairly high number of follow-up projects with customers around the world.&lt;/div&gt; &lt;div class="htmlview paragraph"&gt;The major differences of Common Rail System compared to standard Diesel injection equipment are the free choice of the injection pressure and timing as well as the benefit of the now available pilot injection at almost all operating points.&lt;/div&gt; &lt;div class="htmlview paragraph"&gt;The objective of the CR-System is to reduce the combustion noise significantly, and to improve the torque curve, especially in the lower engine speed range. Simultaneously, the emissions must be kept in the first step within the limits posted by the European legislation for the year 2000.&lt;/div&gt; &lt;div class="htmlview paragraph"&gt;This paper presents combustion related application experience as well as the advantages to the customer in terms of torque, power, fuel economy and noise. The effects of EGR on the timing and quantity of the pilot injection with respect to combustion noise and smoke is presented. Also included are the Common Rail System-specific demands on the nozzle as a crucial component of the system.&lt;/div&gt; &lt;div class="htmlview paragraph"&gt;It is believed that the emission results achieved and the possibility of meeting even stricter emission standards shows the potential of the Common Rail System.&lt;/div&gt;

&lt;div class="htmlview paragraph"&gt;Direct gasoline injection is one major approach in reducing fuel consumption to fulfill the stages of CO2 reduction commitments in Europe from today until 2008. One effort is to unthrottle the gasoline engine during idle and partial load utilizing charge stratification. This may be realized by using different combustion concepts.&lt;/div&gt; &lt;div class="htmlview paragraph"&gt;This paper shows the analysis of mixture preparation for three different types of direct injected gasoline engines. Each engine was driven with two injectors which have two different atomization concepts. The engine types draw a clear dividing line between their combustion concepts. The injectors were analyzed in a pressure chamber, in an optical engine, and in an actual 1-cylinder engine. The formation of wall-film in wall-guided combustion systems will be discussed. Several important injector and engine parameters for fuel direct injection are pointed out. The experimental tools used are the Laser-Induced Exciplex Fluorescence (LIEF), Schlieren and light scattering techniques, as well as shadowgraphy.&lt;/div&gt;