Radar detection software improves accuracy by using less computing

Edge-sensing processing company Teraki has released its latest radar detection software that it says pinpoints stationary and moving objects with increased accuracy and lower computational power. The true traffic solution works on Infineon Technologies’ ASIL-D compliant AURIX TC4x microcontrollers.

Autonomous Driving (AD) and Advanced Driver Assistance Systems (ADAS) rely on accurate sensing of the vehicle’s surroundings to navigate safely. The role of advanced sensors and algorithms is to enhance awareness and ensure safety. Radar has become an essential technology for cost-effective signal processing but there are limitations that must be overcome.

For example, interference can severely compromise radar detection performance, leading to false detections in difficult multi-target situations, which also have high processing requirements. In addition, the accuracy required for reliable radar classifications includes more data points per frame and angular resolution of less than 1°, if stationary and moving objects are to be detected and classified correctly.

Teraki said the machine learning (ML) approach aims to solve this challenge by working with raw data and playing a denoising and cognitive role in slicing information from radar, identifying targets amid noisy environments, clumps, and other interferences while minimizing processing. edge ability. The processing pipeline uses ML to reduce the data required to make accurate discoveries while improving the quality and density of data points for individual discoveries. This ML detection provides more points per object, resulting in fewer false positives, and thus increased safety, compared to other radar processing techniques, such as CFAR (Continuous False Alarm Rate).

Teraki Infineon . block diagram
In Teraki’s hybrid approach, the company said it overcame radar processing challenges by combining traditional signal processing with machine learning. (Photo: Infineon Technologies)

Teraki’s ML-based algorithm, ported using Infineon’s AURIX TC4, claims to reduce radar signals after the first FFT achieves 25 times lower error rates for the same RAM/fps missing objects. Compared to CFAR, the rating is up to 20 percent higher in accuracy, and correct detections are up to 15 percent higher. With this release, Teraki said it is improving its high-end hardware chip architecture, ensuring real-time processing performance on the AURIX TC4, easing computing demands by consuming 4 or 5 bit rates instead of 8 or 32 bits without compromising on F1 – grades. This leads to poor memory required.

Daniel Richart, CEO of Teraki, said, “We’ve improved our software to achieve more with fewer resources. Our solution allows stationary and moving objects to be detected from radar signals and correctly categorized. In addition, it enables customers to detect obstacles at greater distances. This provides AD- and ADAS applications provide more reliable information and therefore better awareness of conditions that lead to safer decision-making. Ultimately, we ensure safety by reducing inference time and processing power required at the edge.”

Product Marketing Manager for Automotive Microcontrollers, Marco Casol, said that the performance of the automotive radar system has increased significantly over the last generation of products. He added, “Edge AI processing is one of many innovations that have helped us drive this increase in radar performance. Infineon’s new parallel processing unit (PPU) is now implemented in Teraki’s unique radar algorithms to showcase the radar performance of the next generation of AURIX TC4x hardware. From Infineon”.


Related content:



Leave a Comment