The automobile went through its digital transformation with the addition of electronic controls in virtually every system. With automated driving and predictive maintenance, the automobile is experiencing another digital transformation in which data-driven algorithms for implementing artificial intelligence are playing a key role. In this presentation, Roy Lurie, who leads the development of MATLAB®, shares advances in MATLAB for handling big engineering data, making analytics and deep learning easy and accessible.
Caterpillar, in collaboration with MathWorks, has developed a big data and machine/deep learning infrastructure. The infrastructure seamlessly provides for a web-based ground-truth interface, a database for storing and querying ground-truth metadata, and an engineering interface with tight integration with MATLAB products for machine learning, visualization, and code generation. The system automatically ground-truths and labels data, dramatically limiting the need for human supervision while reducing development time. The infrastructure is scalable in the number of users, the amount of
Control design and signal processing technologies are converging into common automotive platforms, accompanied by vast increases in system size and complexity. Tailoring MATLAB® and Simulink® to the application level provides a framework for automotive engineers to effectively use Model-Based Design. This presentation introduces new application-oriented tools from MathWorks for ADAS/automated driving and powertrain development as well as major advances in modeling, coding, verifying, and maintaining large systems.
One-Pedal Driving is a drive mode that offers a revitalized set of driver controls that take advantage of the capability of electrified powertrains. One-Pedal Driving combines the highest available level of coast regeneration with intelligent speed trajectory control, bringing the vehicle to a complete stop without using the brake pedal. Engineers at General Motors developed One-Pedal Driving for the 2017 Chevrolet Bolt EV on an accelerated timeline with Simulink® as a key enabler for rapid software development. Simulink provided means to predict control algorithm performance; quickly write, test, and iterate software; and generate code to integrate into controller software.
Application of Robust Statistical Analysis and Machine Learning Algorithms to OBD Threshold Determination
Join this session to learn how the quality of OBD thresholds can be improved with a robust statistical analysis of the passing and failing data used in threshold determination. Using a statistical method for threshold determination as a basis for comparison, Senior Calibration Engineer Tony Gullitti of Robert Bosch, LLC will present how logistical regression can be applied to the threshold determination. Time and scope permitting, the machine learning methods of Gaussian analysis and dimensionality reduction for threshold determination will be explored.
11:35 a.m.–12:00 p.m.
Crown Equipment designs and manufactures the majority of its forklifts and components, including electric motors, drive units, power units, cylinders, electronic modules, and wire harnesses. In addition, Crown develops vehicle application software in-house, and the Vehicle Systems Group supports software development for all of Crown’s North American and European manufactured lift trucks. In 2011 the group established a goal to transition to Model-Based Design using simulation and code generation tools from MathWorks. After many iterations and a few MATLAB® releases, Crown has a simulation environment that matches the vehicle performance and is building all vehicle systems application software including that for electric steering, traction control, hydraulic load handling systems, and vehicle diagnostics.
GE's Trip Optimizer product is an energy management system for heavy-haul freight locomotives. The system considers all factors that can affect the operation of the train and automatically controls throttle and dynamic brakes to operate the train smoothly and efficiently. This presentation covers the current state of the art of driver assistance technology in the rail industry before delving into the finer details of how the Trip Optimizer engineering team uses MATLAB products in each phase of the engineering life cycle. The session also covers a unique approach the team has taken to integrate the optimization library with the MATLAB code generation toolset.
Deploying Model-Based Design at an enterprise level requires more than just models. You also need infrastructure software and hardware to create a collaborative, efficient, and scalable environment for a large number of end users and use cases. In this presentation, Dave showcases both in-product capabilities and integration between MATLAB® and Simulink® and third-party technologies for supporting large teams engaged in developing complex automotive systems. Examples will include ways to share and discover reusable components in your organization, integrating Model-Based Design with third-party collaborative tools such as Github, and how MATLAB and Simulink take advantage of computing resources in the cloud for modeling and simulation.
Automated driving systems perceive the environment using vision, radar, and lidar, and other sensors to detect objects surrounding the vehicle. Join this session to learn how Automated Driving System Toolbox™ can help you:
- Visualize vehicle sensor data
- Detect and verify objects in images
- Fuse and track multiple object detections
Join this session for an introduction to deep learning with MATLAB®. Learn how to use a technique called transfer learning to solve challenging computer vision problems using deep learning. Get a walkthrough of a complete workflow including data preparation, visualization of ground truth, and training an end-to-end deep learning model for detecting ego-vehicle lanes in an image. Once the model is trained, the session will focus on discussing techniques to gain insight into performance and developing intuition on how to improve it further.
Vision and radar-based sensor fusion algorithms are commonly used in automated driving applications to track objects surrounding the vehicle. Join this case study to learn how to:
- Explore a baseline sensor fusion algorithm
- Improve the algorithm for new data set
- Synthesize data to further test the algorithm
The sensor fusion algorithm used in this case study is designed and tested in the context of a forward collision warning application.
MATLAB® has changed significantly over the last several releases. As the size and variety of your engineering data have grown, so has MATLAB capabilities to access, process, and analyze those (big) engineering data sets. In this session, we cover the newest MATLAB features and show you how to use these features to simplify your work, save time, and increase your productivity.
- MATLAB foundation updates (speed, graphics, and Live Editor)
- Data handling and language enhancements (importing data, data types, and big data)
- Working with MDF files directly in MATLAB
- Other toolbox enhancements
Join us for this session to see what you have been missing.
Data analytics is changing how automotive engineering is done. Traditional methods of running statistical models on stored data reach a limit. These store-before-analyze approaches to processing vehicle data are unable to keep up with the increasing data rate due to the rising number of connected vehicles coming online. This is particularly evident in engineering applications and use cases where insight is based on a combination of physics and data.
The solution is clear—combine physics-based models from in-product development with streaming analytics technologies and cloud-based compute and storage infrastructure. Integration of this array of engineering and IT solution can be a challenge, and engineers’ involvement plays a crucial role. In this presentation, Arvind shows a successful integration using MATLAB® through the “digital twin of vehicle powertrain” demonstrator.
With the wealth of available data from testing and connected vehicles, and recent advancements in big data and machine learning technologies, prognostics algorithms—algorithms that can predict component failures before they occur—are receiving lots of attention.
Two approaches are commonly used in developing such algorithms:
- Data-based: create predictive models using sensor data and machine learning approaches
- Model-based: compare sensor data with physics-based models of ideal behavior
In this session, you will learn how to implement both approaches in MATLAB® and Simulink®. We also discuss how such models can be taken into production, enabling companies to improve the reliability of their products and build new predictive maintenance services.
Powertrain Blockset™ is a new product launched in Release 2017a. In this session, you’ll get a closer look at just a few of the ways you can use Powertrain Blockset to accelerate your powertrain systems and controls development, including:
- Engine controller calibrations
- Fuel economy simulations
- Design optimization studies
- Hardware-in-the-loop (HIL) testing
- Multidomain simulation via Simscape™
There are different levels of fidelity when it comes to modeling electric machines. Some applications only require modeling the impact of saturation on machine parameters, and some applications analyze both saturation and the spatial harmonic components in the machine’s flux linkage and torque. This presentation uses a PMSM motor example to show you how to set up DoE in FEA tools and import the ANSYS FEA data to Simulink® for fast and accurate electric machine modeling.
In this master class, MathWorks engineers demonstrate an end-to-end workflow for parameterizing an engine system model, including test plan definition, data collection from a GT-Power CAE model, filling of the engine maps, and validation runs. They also demonstrate the use of a virtual engine for offline calibration of an engine controller in Simulink®. Engine system and controller models in Powertrain Blockset™ are used to demonstrate the workflows throughout this master class.