Publié : 17 mai 2016
Mahendra Muli, Director of Marketing & New Business Development, dSPACE Inc.
The big transition to ADAS and autonomous vehicles is influencing everyone in the automotive industry. While we still continue to work on challenges in powertrains, electrification, battery systems and many other technologies, ADAS and Autonomous Vehicles seem to be at the center of attention.
ADAS is the leading edge of autonomous systems. Some ADAS systems are already in production, and the lines are being pushed. We are getting closer to the autonomous world as more and more companies are working to develop and bring these systems to the road. Let’s look at how disruptions, from many sides, are starting to influence the automotive technology development process.
Earlier this spring, I attended SAE World Congress in Detroit, Michigan, USA. During one of the presentations, the speaker asked the audience, “Who of you here are part of the automotive industry?” Nearly everyone in the room raised their hand, but then he proved all of us wrong when he stated that we are really part of a “new mobility industry” – one that includes other new solutions from electrified aircrafts to Hyperloop. The overall automotive industry definition, itself, is being redefined.
This redefined mobility industry will include a myriad of technologies and business models to transition to, from an ownership-based personal transport to a services-based model. For the potential value of autonomous technology in the services-based business model, it is receiving a very high value. An autonomous driving technology start-up company in California – Cruise Automation, was bought by GM for $1B dollars. There are many others who are hoping to have a similar big payoff.
The collaboration between Google and FCA to develop a self-driving Chrysler Pacifica Hybrid minivan is another example of the shake-up that is occurring within the automotive industry that pairs a technology giant with a traditional automotive OEM.
The disruption of the automotive industry is real and it is happening now. The new requirements rising out of this disruption are having a direct impact on automotive product development and engineering, and it is accelerating the pace at which this technology development is expected to happen.
Another disruptor affecting the automotive industry is robotics. We had the industrial robotics revolution in the 90s. A lot of progress was made. However, robots that manage our lives remained a distant dream. The use of robotics is coming back now with vengeance as artificial intelligence (AI) is being applied to autonomous vehicles. To underline the importance of this experience, the entire Carnegie Mellon robotics team was hired away by Uber to design autonomous vehicles.
Another disruption taking place at the technical level is the consolidation of the electronic control unit (ECU) architecture.
Traditionally, a separate ECU/subsystem is developed for a particular feature, and it is more of a norm to have a distributed architecture. However, as ADAS functionality evolves, there are an increasing number of sensors, subsystems, features and requirements. These features require knowledge of the overall vehicle environment and should come from a single source of truth. Sensor and data fusion are required to produce such information.
Centralized Domain Controller for Autonomous Vehicle Applications
There is one more disruptor that is having an impact on the conventional automotive development process, but in a positive way. I’m talking about virtual validation technology.
Virtual validation is a positive disruptor because it benefits the whole ECU development process by enabling early ECU software tests and allowing the reuse of the critical HIL testing environment. You can run PC-based simulations to validate, verify and test ECU software for any sort of scenario, making the development task even more efficient. Such technologies are a must for testing vehicles comprehensively to confirm, statistically, that the autonomous driving technology is superior to human driving.
dSPACE Virtual Validation Platform – VEOS
Studies done by RAND corporation to estimate the amount of driving time required to statistically prove that autonomous driving is better than human driving indicate that for proving autonomous driving to be 20% better would require 500 years of driving 24/7 with a fleet of 100 vehicles. This is a big challenge that requires us to define new approaches for testing and tools like virtual validation to perform accelerated studies.
HIL testing is, without doubt, the most critical milestone in vehicle software integration and testing. However, with autonomous and ADAS features, it is no longer simply the component models that matter. Also deployed are various operating scenarios (e.g. vehicle, fellow vehicles in traffic, road conditions, environmental conditions, communications -- including infrastructure information exchange, etc.), which need to be simulated.
The simulated information is no longer simply sensor and bus communication, but also includes video inputs, map data, V2X communications, pedestrians, non-moving items in the vehicle environment, such as construction zones, signs, etc. Thus, the conventional HIL setup is being augmented with significant addition of simulation solutions. It also requires new test procedures and evaluation criteria.
HIL System for Testing GPS and V2X Communication
Autonomous technology development can be broadly classified as sensor (and information) fusion and artificial intelligence (or advanced decision-making algorithms). Both of these components require large memory, computational (number crunching) power, and very fast computation. This requires advanced multi-core or GPU-based platforms.
The software algorithms are also coded in different programming languages such as C++ and CUDA, rather than the conventional C programming. A lot of this programming, especially with the new technology companies, is being done using a conventional human programming method, rather than benefitting from the advances made through graphical model-based design. However, we should remain vigilant to not lose the benefits offered by the graphical methods. It will require tools that will provide a graphical environment, yet auto generate code in higher level languages, and being able to execute on the new hardware platforms.
To fill this technical gap for our customers who are working on ADAS and autonomous vehicles, dSPACE recently formed a new collaboration with Intempora SA. We will be distributing their flagship product RTMAPS and integrating it closely to work with dSPACE hardware and software platforms.
One of the key requirements in testing vehicle software for ADAS and autonomous features is the time-correlated replay of all sensor data. RTMaps enables easy communication and information fusion and makes this capability available into the dSPACE environment.
With dSPACE ControlDesk, we already have the feature to correlate data from many different sources to the same time axis. Once recorded, ControlDesk can then replay this data in the virtual scenario, including the 3-D animation, videos, GPS, and other informational data gathered/simulated by the V2X solution. This is a tremendous capability and a critical one necessary for the development of ADAS systems.
In addition to the synchronized replay of sensor data, there is also a need for sensor data simulation, information presentation, reproducibility and repeatability. The information from traditional and non-traditional sources has to be presented coherently, with inherent capabilities, to simulate potential errors, and connectivity issues for ad-hoc networks, etc. dSPACE SCALEXIO systems, together with Automotive Simulation Models (ASM), ControlDesk and various other solutions, provide these unique abilities.
When you consider the vast amount of change that ADAS and autonomous technology is bringing to the automotive industry, you may ask yourself, did we see this change coming or were we blindsided? My answer would be a firm “no.” dSPACE has been working to develop various technologies, specifically for ADAS, into a very comprehensive solution.
Let us consider toolchain development over the last 5 years. Starting from 2011, dSPACE already had anticipated the upcoming growth in software. We had introduced many technologies, including embedded PC for MicroAutoBox, Virtual Validation (VEOS), and a completely new platform for HIL Testing – SCALEXIO, among other things.
In 2012, we added new options for SCALEXIO and a new tool – SYNECT, to manage large amounts of tests and test data. In 2013, we introduced the solution for Ethernet communication, a new rendering engine and many enhancements to MotionDesk. In 2015, we added V2X communication capabilities, and for ControlDesk, new instruments for displaying maps, videos, etc.
As you can see, dSPACE has continuously evolved its tool-chain to help you take on the challenges of disruption. But here is the most important point. dSPACE strives to gather most of the requirements and provide standardized platforms, processes and tools. However, it is possible that you may have additional needs. That is where our Engineering Team can help and fill that gap with a specific solution for you. Please feel free to identify those gaps to us. Maybe there is already a solution ready that will migrate into the product? If not, we will make one available.
Finally, in conclusion, I will say that it is time for us to sit up and take notice of the changes around us, adapt and evolve!
I wish you the very best for your ADAS and Autonomous vehicle projects. I sincerely hope that you will try to use the tools discussed in this blog and make the development process easier.
Faire avancer l'innovation. Toujours à la pointe de l'évolution technologique.
S’abonner à nos newsletters, gérer ses abonnements ou se désabonner