The ultimate goals of Advanced Driver Assistance Systems (ADAS) are driver comfort and safety. Automated Driving (AD) is recognized as the way to achieve both by saving the driver from tedious driving tasks and reducing drastically the risk of road deaths and injuries, which are mostly caused by human error.
In AD, the vehicle assumes control in certain conditions, with the provision for the driver to take over control whenever he or she wants. The AD vehicle obeys all traffic laws, observes speed limits, returns to its lane after overtaking on motorways, and does not slow down other vehicles.
Such functionality requires the usage of multiple sensors such as cameras, radars, lidars (using lasers for measurement), and ultrasonic sensors around the body of the car. These provide data for algorithms in the ECU to analyze the road and the environment and make decisions. They are also fed with additional data coming either from the car’s existing sensors (such as car speed and tilt, current steering angles) or from sources outside the car like GPS information and maps.
SAE International, the association of automotive, aerospace and commercial vehicle engineers, has established several levels of autonomous driving. These levels range from 0, where the driver alone controls the vehicle, up to level 5, a fully automated driving system controlling all aspects of dynamic driving, regardless of the roadway and environmental conditions (i.e. from the motorway to dense urban cities).
Currently Level 2 is deployed in vehicles, and some OEMs are about to launch Level 3 (2018-20). Level 3 entails , the car being driven by electronics and algorithms under controlled conditions. The driver only needs to supervise the functions and react very quickly (within 10 seconds) in case of an inappropriate system reaction. The next level will be level 4 functions, where no driver supervision will be required – the driver can simply do something else. OEMs have made announcements for such systems to appear from 2021 onwards.
Figure 1 (From ADAS towards AD) shows an example of how existing Level 1 functions are combined to create Level 2 functions and gives an abstracted roadmap for functions to come for Level 3, Level 4 and Level 5 autonomy.
In order to design systems that deliver such levels of autonomy, OEMs need to equip the cars with more and more sensors.
Figure 2 (Automated Driving Architecture) illustrates an example setup using camera, radar, and lidars as well as maps that are connected to an automated drive ECU. This will perform sensor fusion, free space detection and situation assessment while taking decisions for the maneuvers to be performed and commands to be sent to the actuators.
Renesas is contributing to Level 3 and Level 4 systems, using the R-Car SoC and RH850 MCU, and working on the technologies required to address the next generation of functionalities.
Integration of these highly complex setups in the automated vehicle's system requires the ability to design a modular solution that scales across different vehicles and that always meets embedded constraints in terms of performance, low power, safety and security.
The R-Car Gen3 SoC family provides a scalable and reusable platform ideal for dealing with such constraints. Thanks to this family approach, software development from one device can be re-used on another device to scale performance and power consumption. This helps manage the complexity of multiple ECU requirements for Tier1s, for example by tailoring their platform according to OEM requirements.
The R-Car Gen3 family is a complete and scalable line-up of automotive SoCs based on TSMC’s cutting edge 16nm finFET process. In addition to power consumption performance, this silicon process offers high performance due to higher clock speed.
The line-up is based on the latest Arm-v8A processor architecture (Arm® Cortex-A57/Cortex-A53). It is a high performance 64-bit architecture which comes with specialized instructions for floating-point, SIMD and even security processing. Based on this, the R-Car H3 features an outstanding performance of roughly 40,000 Dhrystone Million Instructions per Second (DMIPS).
These Arm processors are complemented with hardware accelerators for computing vision, such as the IMP-X5 image recognition engine that supports machine learning and other artificial intelligence algorithms.
Other co-processors, such as GPUs, are also provided to get the best trade-offs between performance and ease of use depending on the application requirements. The R-Car H3 also features advanced security measures, ranging from specific non-volatile memory to store secret keys, secure boot to make sure the code that will run is authorized, as well as lifecycle management and secure communication. Such features are key for the end user’s ability to trust the autonomous car and make sure that hackers cannot compromise people’s safety. Last but not least, the R-Car H3 has been designed according to ISO26262 in order to meet the requirements for safety up to ASIL B in demanding applications such as sensor fusion.
However Renesas portfolio is not only limited to automotive safety capable SoCs. It also covers the required high safety controller that is mandatory to issue up to ASIL D commands to the actuators in an AD application. This controller is based on Renesas RH850 architecture in the name of the RH850/P1H-C. This device provides up to 1344DMIPS for ASIL D applications and offers extensive security support by the integration of an dedicated HSM (High Security Module). Renesas offers multiple evaluation and support platforms for different development phases. To test algorithms and to transition into embedded hardware that is refined up to series production, Renesas offers software support and works with partners to develop an integrated automotive system.automotive safety capable SoCs.
R-Car H3 and M3 Starter Kit Premier is designed to accelerate development based on the R-Car H3. It provides OpenCL and OpenGL ES 3.1 support in order to give software developers an easy transition from their algorithms on PCs to an embedded platform.
Renesas is also working with TTTech to deliver a reference solution for Automated Drive ECU called RazorMotion.
RazorMotion integrates a full automotive design, taking into account reliability, thermal, integration requirements and also a safety concept capable for ASIL D. It is combined with a middleware technology based on time-triggered architecture which streamlines the complex ECU integration process. The benefit for OEMs & Tier1s is faster development of their product both from a hardware and a software perspective.
Renesas is using such solutions in order to identify and understand the difficulties its customers will meet when developing autonomous cars. It demonstrated its Skyline initiative at CES 2016 and 2017, enabling autonomous driving functions based on the HAD Solution Kit combined with camera, radar, lidar & vehicle-to-vehicle and vehicle to-infrastructure communication.
Through these endeavors, Renesas can understand better the synchronization and integration issues faced by OEMs and Tier1s when dealing with multiple software providers in a constrained timeline. The integration also enables tests of state-of-the-art algorithms from its partners and covers the required steps to move towards the next automated drive functions.