With an increasing need for advanced driver assistance systems (ADAS) to be a flexible and reliable platform, field programmable gate arrays (FPGAs) are becoming a logical fit for next-generation safety systems in automobiles.
FPGAs provide better flexibility over its counterparts in microcontrollers (MCUs), digital signal processors (DSPs), application specific processors and any combination of these architectures.
Multi-threading a boon to park assist
FPGAs are capable of multi-threading, which allows them to implement different functions in parallel. In surround-view park assistance application images captured by each camera are sent to a data fusion module. This data fusion module is responsible for executing different image processing functions to display the surroundings of a car on an LCD. Different image processing functions such as analytics, image warping, high dynamic range and object classification could be implemented in a single chip using an FPGA because of its multi-threading feature. This allows for fast booting of videos allowing FPGAs to be widely used for surround-view camera systems and night vision systems.
Revenue for FPGAs used in surround-view park assistance systems are expected to grow fivefold to reach $29 million in 2020 from 2013 with a compound annual growth rate (CAGR) of 27 percent. While the revenue for FPGAs used in night vision systems are expected to double to reach $33 million in 2020 from 2013 with a CAGR of 10 percent.
Most of the FPGAs have integrated ARM cores to allow for serial processing operation on top of parallel processing capabilities. Serial processing operation allows FPGA to be flexible in communicating over CAN, FlexRay and Ethernet transceivers.
Parallel processing and multi-threading capability are of great advantage in implementing ADAS functions. For instance, a system which implements lane departure warning (LDW) and pedestrian detection (PD) needs to process two different algorithms for LDW & PD. LDW would process algorithms for grayscale images and PD would process algorithms for RGB (Red-Green-Blue) images. With FPGA, independent processing channels could be achieved to implement functions for LDW & PD at the same time which is a benefit over the standard application specific processors.
Within FPGA redundancy for safety control functions could be adopted and programmed. Multi-parallel tasks also could be programmed via hardware. This flexibility allows for better performance for FPGAs over standard MCUs or application specific processors. Flexible and programmable split between serial and parallel processing allows for communication between modules over diversified protocols such as CAN, FlexRay and Ethernet.
Opportunities for FPGAs in ADAS
Considering the computational benefits of FPGAs, driver assistance systems hold a potential demand for FPGAs. ADAS is a dynamic market with increased number of applications being integrated in a car. With independent logic blocks, FPGAs could be used for implementing multi-function front-view camera applications. A one-box solution could be achieved as modern FPGAs could implement around 6 to 7 functions using a single chip. Long term penetration of multi-functions using FPGAs would be observed in Mid-range and High-end cars.
The processing power of FPGAs could allow them to be used in data fusion modules for analyzing data from different sensors such as radar, image, laser and ultrasound. Considering the drift from passive warning systems to active control systems, safety standards such as ASIL or ISO26262 could be achieved by implementing redundant functions using the logic blocks in FPGAs. Expertise in development of embedded software using hardware description language at the design level is a main challenge faced for the growth of FPGAs in driver assistance systems.