Advances in autonomous driving and advanced driver-assistance systems (ADAS) now enable vehicles to understand their surroundings in real time, a complex process based on the concepts of perception and sensor fusion.
Perception refers to the ability to detect and classify objects, recognize traffic signs, track pedestrians, and map the road. However, every type of sensor has its own physical limitations. Sensor fusion is the technology that bridges this gap by intelligently combining data from sensors using advanced algorithms.
In this article, we will explore how perception and sensor fusion are driving evolution and innovations in the three main categories of automotive sensors: radar, LiDAR, and cameras. We also highlight several radar, LiDAR, and camera products with advanced features to improve ADAS imaging and perception.
Radar performs in harsh environments
Automotive radar offers excellent performance in tough conditions, such as fog, rain, and darkness, where cameras and LiDAR are less effective. The introduction of 4D imaging radar has added elevation (height) to the standard range, velocity, and azimuth outputs. 4D radar can thus generate dense point clouds that were once available only with LiDAR. According to MarketsandMarkets, the 4D imaging radar market was valued at $390 million in 2025 and is expected to grow to $1.2 billion by 2030, at a compound annual growth rate of 25.2%.
Hirain Technologies has developed the LRR615, a long-range imaging radar built around Arbe Robotics Ltd.’s 4D imaging radar chipset solution. The radar system is the first to be equipped with a high-density waveguide antenna. This feature provides improved image clarity, detection sensitivity, and signal integrity.
Designed to be cost-effective and manufacturable on a large scale, the LRR615 can be paired with cameras in autonomous-driving systems, providing an alternative to LiDAR. After integrating, calibrating, and validating the device, Hirain is setting up the manufacturing process to achieve an annual production of 10,000 units.
Arbe’s chipset (Figure 1) is a multichip architecture that integrates three proprietary, automotive-grade integrated circuits that process ultra-high-resolution data. The chipset can process real-time data from 2,304 virtual channels, achieve 3-Tbits/s equivalent processing throughput, and provide over 10,000 detections at 20 fps.
Arbe’s chipset is based on GlobalFoundries’ (GF’s) proprietary 22FDX process technology. This platform, specifically addressing automotive radar, integrates RF, analog, and digital processing blocks on the same die. This reduces system costs and accelerates time to market. According to GF, using its technology, Arbe is the first in the industry to create a real-time, 4D image of the surrounding environment, achieving a 1° resolution.
Another relevant application of automotive radar is in-cabin monitoring. An example is Texas Instruments Inc.’s AWRL6844, a 57- to 64-GHz mmWave radar sensor designed for occupancy monitoring, including seat-belt reminder systems, child-presence detection, and intrusion detection (Figure 2).
The AWRL6844 is a low-power device integrating four transmitters and four receivers. This high-resolution sensing data is processed by specific AI algorithms running on a customizable, on-chip hardware accelerator and DSP. This single-chip solution improves detection accuracy, reduces processing time, and enables a safer driving environment.
Infineon Technologies AG offers a similar product, the XENSIV BGT60ATR24AIP 57.7- to 62.4-GHz radar sensor based on FMCW technology. Available in a compact, 8 × 8-mm2 package with antenna-in-package technology, the device features ultra‑low power consumption, high precision, and advanced sensing, making it suitable for contactless, high-accuracy in‑cabin monitoring.
LiDAR advances with better perception and identification
MicroVision Inc., a company specializing in advanced perception solutions that recently acquired the assets from Luminar Technologies, introduced its Tri-LiDAR architecture. This solution integrates two MOVIA S short-range sensors (placed on the front corners) with one forward-facing HALO long-range LiDAR, delivering continuous, 360° environmental coverage (Figure 3).
The company’s software platform performs the real-time fusion of the data coming from all the sensors and generates a single, high-fidelity point cloud. This enables accurate object detection, classification, and tracking, delivering a real-time perception system.
According to the company, Tri-LiDAR provides three main benefits: a reduction in the power consumption of each sensor, a reduction in the packaging of each sensor, and a cost reduction of the system.
Aeva Inc., a company specializing in sensing and perception systems, has licensed Cadence Tensilica Vision DSP IP to support the signal-processing tasks of its 4D LiDAR systems.
Tensilica Vision DSPs’ low-power architecture and Tensilica Instruction Extension language make Tensilica DSPs suited for applications in which real-time signal processing, low latency, and high efficiency are mandatory. According to Aeva, the flexibility and performance of Cadence’s Vision DSP technology will improve the perception and scalability of its solutions, addressing automotive and industrial applications.
Hesai Technology, a company specializing in 3D perception, announced the Picasso 6D Full-Color LiDAR SPAD-SoC, a solution that natively captures 3D spatial geometry (X, Y, Z) and 3D color data (R, G, B) on a single chip, eliminating the need for fusion of separate camera and LiDAR data.
By performing the sensor data fusion directly at the silicon level (the single-photon avalanche-diode SoC), the Picasso chip produces high-resolution, colorized point clouds simultaneously, allowing for better identification of objects such as traffic lights, lane markings, and construction zones. Hesai’s ETX LiDAR, upgraded to support up to 4,320 channels, will integrate this technology and will be available in the second half of 2026.
Lumotive, a company specializing in programmable optical semiconductors, has introduced a solid-state LiDAR platform that combines its commercially available LM10 Light Control Metasurface (LCM) with the ADS6311 Hawk sensor from Adaps Photonics.
The solution (Figure 4) delivers a 180° horizontal field of view and operates at 30 fps, eliminating blind spots, improving the tracking of fast-moving objects, and reducing the number of sensors required. Lumotive’s LCM technology electronically steers light at semiconductor speed, without encountering the limitations associated with mechanical scanners and fixed-channel VCSEL arrays. In addition to 180° horizontal coverage, the sensor offers up to 140° vertical coverage, configurable through software to optimize range, resolution, and frame rate.
Cameras shift to 8-MP sensors
Cameras are rapidly migrating to 8-MP sensors, representing the new baseline for ADAS front and side cameras. The higher resolution directly increases the detection range at highway speeds, enabling confident object classification at distances that 2-MP and 5-MP sensors cannot reliably achieve.
Omnivision introduced the OX08D20 8‑MP CMOS automotive image sensor based on its proprietary TheiaCel technology. The device is an improved version of the OX08D10 sensor for exterior cameras widely used in ADAS and autonomous-driving systems.
The OX08D20 image sensor features a 60-fps frame rate, 2× higher than its predecessor, OX08D10, enabling dual-use cameras. With 60 fps, the video flows more smoothly. This allows car manufacturers to save money, space, and wiring by using one camera to handle both background autonomous-driving tasks and real-time visual displays for the driver. The sensor also supports the latest cybersecurity standard MIPI CSE 2.0 (Camera Service Extensions v2.0) developed by the MIPI Alliance.
Omnivision’s TheiaCel technology has been designed to capture high-quality images even in extreme lighting conditions. It achieves high dynamic range (HDR) using a single exposure. In this way, the LED flicker experienced in traditional HDR sensors is eliminated.
Sony Semiconductor Solutions introduced the IMX828, the industry’s first 8-MP CMOS image sensor for automotive cameras featuring a built-in MIPI A-PHY transmission interface. Traditional automotive camera systems require externally mounted serializer chips to transfer data safely to the vehicle’s electronic control unit. By embedding the MIPI A-PHY interface directly into the sensor, Sony eliminates the need for this additional hardware.
This feature enables a reduction in board size and module power consumption and limits heat generation. The chip also integrates a proprietary error-handling circuit that resists external noise disruptions.
NXP Semiconductors’ 4K MIPI CMOS camera module (IMX-OS08A20) is a high-performance development tool for consumer, industrial, and automotive vision systems. It is built on OmniVision’s 8-MP OS08A20 sensor and adopts PureCel and Nyxel technologies to capture 4K Ultra-HD video at 60 fps.
The development kit includes the 8-MP sensor module, a MINI-SAS interface cable, and an adapter board. It is designed to plug directly into the NXP i.MX 8M Plus Evaluation Kit. The module’s high-resolution performance makes it well-suited for automotive in-cabin applications, such as driver-monitoring systems and occupant-monitoring systems.
The post Designer’s guide: Radar, LiDAR, and cameras advance ADAS appeared first on EDN.