Lead Innovation

R&D in the fields of In-Vehicle Architecture•Autonomous Driving• V2X connected system consulting

Almost all of the vehicles have been so far composed of decentralized control architecture, which functions in safety, infotainment, powertrain, and drivetrain domains In the middle of the 2020s, the E/E architecture will shift to centralized composition, which will be connected with every domain and be handled with the integrated control in order to deal with a wide range of transportation/travel/location/charging services and manage with electrified components.

Regarding the safety domain, ADAS Level 2+ has become the mainstream from Level 2, and levels 3 and 4 will be deployed to specific highways and parking lots equipped with HD dynamic map. In harmony with this evolution, V2X data-based ADAS/AD control algorithm has been implemented, and in the middle of the 20’s, Software Defined Network (SDN) will be adapted to dynamically change data transmission patterns and the system structures.

With regard to safety-related sensors such as on-board camera, millimeter wave radar and LiDAR will increase their resolutions from approximately 2M pixels to 4K, and deep learning such as CNN and other advance AI model for image/video data has been hired in order to analyze complicated driving circumstance and then forecast the behavior and trajectory route of vehicles and pedestrians. In millimeter-wave radars, CMOS chip which integrates short-range and long-range detecting capabilities into one solution and 4D by combining 3D and data integral are making progress.

ADAS management based on behavior forecast and trajectory plans

The creation of suitable trajectory plans in accordance with the precise behavior forecasting of other vehicles and pedestrians, and the calculation algorithm to avoid potential accidents, are increasingly in need these days.

In order to realize the integrated control over infotainment/cockpit, ADAS/AD, body/in-cabin, connected/V2X-related ECU, semiconductors equipped with arithmetic processing capabilities ranging from a few ten trillion to several hundred trillion or even more FLOPS will be used. Furthermore, 1 Gbps and multi-Gbps for data transmission by the Ethernet will be also utilized as a backbone network in vehicle, and they will coexist with the next-generation CAN and existing SerDes signal protocols to build an ultra-high-data-rate network.

New BEV architecture in the middle of the 2020s

In the middle of the 2020s, the in-vehicle system structure and data transfer patterns will be totally changed; for example, the dynamic management of data transmission in harmony with external services and applications will become active.

Roadmap of cutting-edge technologies and advanced applications in BEV and other xEVs

We have been researching and engaged in the advanced development projects related to E/E architectures including central computing, zone-ECU, Ethernet, SDN, V2X connection with LTE/5G-NR/DSRC, integrated antenna, and cloud-coordinated vehicle control. By utilizing the know-how, we navigate automotive OEMs, tier1s, and tier2s to the most appropriate direction through our continuous analyses on the latest trends.

Development of accident prevention system by animal behavior analysis

We have been engaged in the variety of research and analysis projects on automotive and transportation sectors, and by utilizing these know-how, we are now developing a system that protects animals caught by traffic accidents—many different animals are increasingly involved in traffic accidents all over the world.

Development of a system which analyses the walking and running-away patterns of specific animals on a variety of road environments

  • We are developing a unique system for preventing traffic accidents involving animals. It is composed of such process steps as; 1) analyzing walking trajectory and running-away patterns on/around main roads for deer such as moose, elk, wapiti, caribou, reindeer, and bears; 2) matching between the road structure and the conditions on how deer and bears tend to encounter accidents that are analyzed in advance; and 3) providing ADAS ECU with its dataset.
  • The method of the pattern-learning is aggregating a wide range of images and videos of major deer and bear species, including their ordinary walking scene on/around roads and the behavioral changing patterns when vehicles are approaching. Then, their big-data are analyzed through a unique deep-learning mechanism to identify the feature points of the behaviors. On top of that, we set up high resolution cameras and millimeter-wave radars on specific roads, which record multiple deer and bears real-time as well.
  • In order to differentiate by season and time of the day (e.g., breeding season, when they take a group behavior, evening, and at night) and road structures, we have classified the animals’ walking and running-away trajectory patterns into their different conditions.

Factor analyses for forecasting animals’ behavior

  • In order to forecast how animals start to move and run away when vehicles approach or when they are caught in the headlight, we are analyzing how they react to a potential accident; especially how their eyes, ears, nose, hair and foot move, and if they start to tear.
  • Then we generate their trajectory patterns by road structure/condition, by season and by time period.

Matching analysis between predicted walking and running-away patterns of animals, road structures, and seasonal factors

  • The recognition process steps on the vehicle side are as follows: 1) The on-board sensors detect the types of objects; 2) the object is identified as a ‘moose’; and 3) ADAS-ECU extracts the dataset of the moose and read their walking/running-away patterns on the map.
  • After that, the ECU warns a car driver about the risk of accident through the in-vehicle voice function or HUD, and then, the ECU makes a forecast on an accident, calculates, and controls the suitable vehicle speed and the driving lane so that the accident can be prevented.