FEATURE AUTOMOTIVE ELECTRONICS
DRIVING DATA DEVELOPMENTS
Michael Hendricks, director Automotive and Consumer Business Units, Intel Programmable Solutions Group, discusses the big-data challenge created by the development of autonomous vehicles
T
he concept of automated driving is fast becoming reality. Promising to
bring a whole new way to how we interact with our car, the benefits of fewer accidents, more efficient traffic flow and mobility for all are just some of the more notable gains. For the automotive manufacturer and
their suppliers, delivering against these goals present considerable challenges. Pressure has been building within the automotive electronics market for some time. Many of these factors, such as provisioning advanced driver assistance systems (ADAS), implementing car-to- anything / car-to-car C2X/C2C (V2X/V2C) infrastructure safety communication systems and the increasingly sophisticated infotainment and navigation systems, have driven the need for faster networking within the vehicle and “always-on” connectivity to the outside world. Alongside this, the tier 0/tier 1 suppliers have been trying to drive down component costs with initiatives such as electronic control unit (ECU) consolidation, simplifying the number of different networks used in a car and enhancing the passenger experience with even more features. Then there is the end game of
autonomous, or automated driving. There is no surprise that industry pundits are making the parallel of the car of the future being a data centre on wheels. Brian Krzanich, Intel’s CEO, in his recent blog1
AEC-Q100/200 for hardware and ISO- 26262 for both hardware and software components. Sensing its environment is crucial to an
automated vehicle, and it's highly dynamic. Being able to track multiple targets, by multiple means (visual, RADAR, LIDAR, SONAR) while moving forward at significant speed requires a constant stream of data to be acquired, processed and decisions made based upon it. This requires highly scalable sequential and parallel acceleration compute needs. Combining and aggregating data from the above and other sources, such as GNSS and future C2X/C2V, allows instant repeated decisions to be made with certainty and precision. Helping to speed the development of
electronic systems for automated vehicle applications is an approach that sees the integration of a number of different compute resources within a single platform. An example of this is the Intel GO development platform for automated driving. Two versions are available, one based around Intel Atom processor and the other based on Intel Xeon processors. Figure 2 below illustrates a functional
, estimates that an autonomous vehicle could generate about 4,000 GB of data every day – see Figure 1 above. Tackling the real-time big-data
challenges highlighted above is just the starting point for the systems developer. Balancing workloads, managing a hard real-time environment, processing huge amounts of visual data, all have their particular compute and scalability requirements, but the developer also needs to ensure that everything meets the differing hardware and software safety standards any automated environment is subject to. These include
8 JUNE 2017 | ELECTRONICS
block diagram of the Intel GO platform using the two Intel Xeon processors. Providing a comprehensive platform from which developers can prototype and optimise solutions for a host of automated driving solutions, it includes two Intel Xeon processors each with multiple channels of DDR memory per
Figure 1 – The data- generating autonomous vehicle
Michael Hendricks is director Automotive and Consumer Business Units, at Intel Programmable Solutions Group
https://newsroom.intel. com/editorials/krzanich- the-future-of-automated- driving/
References: 1
Figure 2 – Intel GO development platform using Intel Xeon processor
board. In addition, each Xeon CPU has an Intel Arria 10 FPGA expansion card that provides the capability of accelerate applications such as computer vision, sensor fusion and deep learning. With their parallel compute capabilities, each Arria 10 provides hard floating point digital signal processing (DSP) blocks with speeds up to 1,500 giga floating point operations per second (GFLOPS). A 16 port 10 gigabit Ethernet port (10 GbE) switch connects all the compute boards internally and provides 8 ports for external connections within the vehicle. Finally, an Infineon AURIX microcontroller (MCU) provides a means for developing functional safety routines in addition to managing connectivity via the popular automotive CAN and FlexRay bus interfaces. Complementing the hardware platform is a software development kit (SDK) that aids maximising the hardware resources. A number of computer vision, deep learning and OpenCL tool kits are included to speed the design of middleware and algorithms for the automated driving compute phases of perception, fusion and decision-making. The SDK also contains application optimised libraries in addition to open source compilers, performance and power analysers, and IDE debuggers that enable a full stack optimisation and rapid development using a functional safety compliance workflow. Selections of sample reference applications such as for lane change detection and object avoidance also assist developers to shorten the overall development process. In the near future the Intel GO platforms will benefit from the addition of a 5G-ready communications platform that will enable a host of connectivity options. Adopting a platform-based development approach to automated driving systems development will greatly assist engineers to establish and trial applications in a fully integrated manner, safe in the knowledge that software safety will not be an after-thought.
Intel Programmable Solutions Group
www.intel.com T: 01494 602000
/ ELECTRONICS
Page 1 |
Page 2 |
Page 3 |
Page 4 |
Page 5 |
Page 6 |
Page 7 |
Page 8 |
Page 9 |
Page 10 |
Page 11 |
Page 12 |
Page 13 |
Page 14 |
Page 15 |
Page 16 |
Page 17 |
Page 18 |
Page 19 |
Page 20 |
Page 21 |
Page 22 |
Page 23 |
Page 24 |
Page 25 |
Page 26 |
Page 27 |
Page 28 |
Page 29 |
Page 30 |
Page 31 |
Page 32 |
Page 33 |
Page 34 |
Page 35 |
Page 36 |
Page 37 |
Page 38 |
Page 39 |
Page 40 |
Page 41 |
Page 42 |
Page 43 |
Page 44 |
Page 45 |
Page 46 |
Page 47 |
Page 48 |
Page 49 |
Page 50 |
Page 51 |
Page 52 |
Page 53 |
Page 54 |
Page 55 |
Page 56
