Multiple Patents Filed and Published This Year, with a Focus on Autonomous Driving Reliability
From Sensor Integration and Malfunction Detection to Predicting Surrounding Conditions and AI Retraining
Targets 2029 Mass Production of In-House AI-Based Level 2++ Vehicles

Hyundai Motor Group is focusing on reliability to strengthen its competitiveness in autonomous driving. Most of the domestic and international patents it has filed or published so far this year focus on improving reliability across autonomous driving systems.


Following Tesla, global automakers are commercializing autonomous driving, but Hyundai Motor Group aims to begin mass-producing autonomous vehicles equipped with its own artificial intelligence (AI) in 2029. The timeline may be delayed, but the group appears determined not to compromise on autonomous driving safety.


Atria AI Collection of Edge Cases in Urban Autonomous Driving. Hyundai Motor Group

Atria AI Collection of Edge Cases in Urban Autonomous Driving. Hyundai Motor Group

View original image

A review of patents filed or published this year by Hyundai Motor Group, Hyundai Mobis, 42dot and Motional, among others, reveals the direction of autonomous driving technology development. The focus is on improving reliability across the system, including detecting malfunctions in essential autonomous driving sensors and supplementing driving information, as well as predicting the movements of surrounding vehicles and retraining AI with driving data.


Sensor integration and malfunction response technologies stand out first. The "Modular Sensor Device for Vehicles," disclosed by Hyundai Motor and Kia in China, describes integrating a camera, radar and lidar into a single module. Rather than installing each sensor separately, the system combines them in one device designed to be mounted wherever desired inside a vehicle.


In particular, the technology includes restoring the surface with a heating element if the sensor's protective cover is damaged by an external impact or wear. It also includes a function that automatically adjusts the detection areas of the radar and lidar according to vehicle movement. This approach aims to manage not only sensor recognition performance but also physical damage and detection errors that may occur while driving.


There is also technology designed to address situations in which sensors fail to function properly. Hyundai Mobis's "System and Method for Detecting Radar Sensor Occlusion" determines whether a sensor is obstructed by analyzing the direction and strength of the electromagnetic waves transmitted and received by the radar. The technology is designed to detect situations in which sensor performance declines because foreign matter has accumulated on the sensor surface or an external object is blocking it.

"Cars That Keep Going Even When Sensors Fail"... Hyundai Motor Boosts Autonomous Driving Reliability View original image

Hyundai Motor and Kia's patent for "Vehicle and Autonomous Driving System" describes a more specific structure for responding to sensor failures. If some sensors fail, the system uses information from other sensors that are functioning normally to generate driving data. If sensor-based autonomous driving is not feasible, it uses data received through a communications unit to generate alternative driving information. This helps prevent a malfunction in one sensor from interrupting the entire autonomous driving system.


This approach contrasts with Tesla's autonomous driving technology. Tesla's core approach relies on a camera-centered, vision-based perception system and training on large volumes of driving data. Hyundai Motor Group, by contrast, places greater emphasis on improving system stability through hardware designs that integrate cameras, radar and lidar, as well as a system for generating alternative information from other sensors and communications data when a sensor fails.


Technology that predicts future movements, going beyond simply recognizing surrounding conditions, also stands out. Motional's patent, "Trajectory Planning for Autonomous Vehicles Using Dynamic States," describes adjusting a vehicle's route by distinguishing fixed obstacles from moving hazards. The aim is to plan safer routes by taking into account the movements of surrounding objects whose positions and speeds change over time, as well as stationary objects.

Autonomous driving vehicles unveiled at the Hyundai Motor Group Autonomous Driving Media Day (from left: SDV Testbed, Gwangju autonomous driving demonstration project vehicle, and Data Collection Vehicle). Hyundai Motor Group

Autonomous driving vehicles unveiled at the Hyundai Motor Group Autonomous Driving Media Day (from left: SDV Testbed, Gwangju autonomous driving demonstration project vehicle, and Data Collection Vehicle). Hyundai Motor Group

View original image

42dot is developing technology that uses data collected on actual roads to improve AI performance. Its patent for an "Apparatus and Server for Retraining an Artificial Intelligence Model Implementing Autonomous Driving Mode" describes a system that collects false-detection data generated by vehicles operating in non-autonomous driving mode, then has a separate AI model annotate the data for use in retraining the autonomous driving model. Rather than simply recording errors that occur during actual driving, the approach converts them into training data to improve recognition performance.


42dot has also filed patents for technology that predicts the routes of surrounding vehicles and the possibility of collisions to determine whether to merge onto the main road, as well as technology that creates virtual stop lines based on pedestrian locations. It has also developed simulation technology that recreates roads and surrounding vehicles in a virtual driving environment to validate autonomous driving algorithms. In other words, the scope of its technology development includes not only perception and decision-making but also repeatedly testing a range of scenarios before the technology is deployed on actual roads.



Meanwhile, Hyundai Motor Group aims to apply autonomous driving technology developed in collaboration with Nvidia to mass-produced vehicles starting in 2028, while beginning mass production of its in-house Atria AI-based "Level 2++" vehicles in the second half of 2029.


This content was produced with the assistance of AI translation services.

© The Asia Business Daily. All rights reserved. Unauthorized AI training and use prohibited.

Today’s Briefing