□ A research team from the Division of Mobility Technology at Daegu Gyeongbuk Institute of Science and Technology (DGIST) (President Kunwoo Lee), in collaboration with Gumi-based JuKwang Precision Co., Ltd., developed a “high-resolution angle estimation technology”. It dramatically reduces the computational load required for data processing while maintaining high radar detection performance for autonomous mobile robots (AMRs).
□ Radar, a key sensor for autonomous mobile robots, uses radio waves to determine the distance, speed, and direction of surrounding obstacles. As it operates reliably even under adverse weather conditions, such as rain, snow, and fog, radar is considered an essential sensor that complements cameras and light detection and ranging (LiDAR). However, conventional radar signal-processing methods, while computationally efficient, have difficulty accurately distinguishing between objects located close to one another. Although ultra-high-resolution algorithms such as multiple signal classification (MUSIC) can overcome this limitation, their computational load is too high to enable real-time processing on compact robotic systems.
□ To overcome this limitation, the research team developed a new algorithm called “Adaptive Grid-Refinement MUSIC.” While conventional methods thoroughly search the entire detection area using a fine grid from beginning to end, the newly developed technology rapidly scans the overall area at wider intervals and selects only the key regions where objects are likely to be present for a more precise search. This principle is similar to obtaining a broad overview of a map and using a microscope to examine only the areas that require closer inspection.
□ In particular, the research team implemented a smart computational architecture that performs the complex mathematical calculations required for high-resolution analysis only once and then reuses the resulting information, rather than repeating the calculations each time. As a result, the team succeeded in reducing unnecessary search computations by approximately 88% while maintaining detection performance comparable to the conventional algorithm. Simulations and real-world validation using 77-GHz automotive radar also demonstrated more than a twofold improvement in processing efficiency compared with existing approaches.
□ This achievement is particularly significant as an example of industry-academia collaboration that created synergies by combining DGIST’s advanced capabilities in radar signal processing with the practical expertise of a company in the Daegu-Gyeongbuk region. The research team at JuKwang Precision Co., Ltd. provided the robot platform and conducted real-world performance validation, while the DGIST research team oversaw the overall research process, including algorithm design and analysis.
□ “Although high-resolution radar algorithms offer excellent performance, their enormous computational load has made them difficult to apply to real-world robots and vehicles,” said Sangdong Kim, Principal Researcher in the Division of Mobility Technology at DGIST, “This study presents a practical technology that dramatically reduces the computational load by precisely searching only the target regions that require detailed analysis. We will continue to advance the technology so that the outcomes of our joint research with a local company can move beyond academic publications and be applied in real-world intelligent mobility settings.”
□ This research was supported by the Ministry of Science and ICT’s DGIST R&D Program and a program of the Ministry of Land, Infrastructure and Transport and the Korea Agency for Infrastructure Technology Advancement (KAIA). The study was conducted by Senior Researchers Bongseok Kim and Jeseok Kim, Senior Research Fellow Rock-Hyun Choi, and Principal Researcher Sangdong Kim of DGIST’s Division of Mobility Technology, along with Joong-Tae Kim, Director of the Research Institute, and Researcher Sihyoung Kim of JuKwang Precision Co., Ltd. The research findings were published in September in the internationally recognized Journal of Electrical Engineering & Technology .
Journal of Electrical Engineering and Technology
Adaptive Grid-Refinement MUSIC for Low-Complexity High-Resolution DOA Estimation in Automotive Radar
19-Aug-2026