Enhance functionality on the same platform with heterogeneous architecture
Support for SoC, software, and reference

As a number of recent collisions involving large buses have occurred due to drowsy drivers, calls for mandatory installation of advanced driver assistance systems (ADAS) are growing louder.

The "Enforcement Decree and Enforcement Rules of the Traffic Safety Act," partially revised in April, will take effect on the 18th of this month. This mandates the installation of a Lane Keeping Assist System (LDWS) on large commercial vehicles, such as buses and cargo trucks, to warn drivers of unintentional lane departures, such as those caused by drowsy driving.

ADAS encompasses six systems: LDWS, Forward Collision Warning Systems (FCWS), Intelligent Parking Assist Systems (IPAS), Computational Navigation Systems (CNS), Rear Park Assist Systems (RPAS), and Pedestrian Detection Systems (PCWS). Because ADAS technology relies on cameras to recognize situations, performance varies by product.



Automotive companies want to leverage existing hardware components and software without compromising system performance. Heterogeneous architectures allow for increased functionality within a single platform. To address this, Texas Instruments (TI) offers a multi-OS, multi-domain software architecture.

TI offers the TDA driver assistance SoC as an ADAS solution. This product family provides open, scalable solutions based on a common hardware and software architecture for ADAS applications such as camera-based front, rear, surround view, night vision systems, multi-range radar, and sensor fusion systems.

The TDAx SoC features a heterogeneous architecture that integrates specialized hardware accelerators (HWAs) such as an Image Signal Processor (ISP), an Embedded Vision Engine (EVE), and a Digital Signal Processor (DSP). The TDA2x comprises dual Cortex-M4 cores, each with two ARM Cortex-A15s running at 200MHz, and four EVEs.

“This SoC allows for simultaneous execution of front camera applications and implementation of parking assistance applications, including intelligent 2D/3D surround view and rear collision warning,” said Jongpil Park, Senior Researcher at WT Technology FAE. “It is designed to execute low- to mid-level image processing functions.”

The TDA3x offers lower-level processing than the TDA2x and does not include an ARM Cortex-A15 processor. It can accept four camera inputs and has an EMIF (External Memory Interface) that connects to DDR memory. Safety-related modules supporting ASIL B level are also provided.

It is suitable for ADAS applications such as 3D surround view, smart rear camera, radar, camera mirror replacement, driver monitoring, and lidar without using GPU. It also provides a vision software development kit (SDK) that allows ADAS functions to be implemented in the same platform environment, from compact cars to large luxury vehicles, through various built-in software libraries.

Senior Researcher Jongpil Park said, “We provide the appropriate tools and software support, DSP, EVE example programs, and references necessary to realize ADAS functions,” adding, “Compared to other products, our performance is high for the price.”