RoboSense (02498) announced that two core chips of its digital LiDAR have passed AEC-Q102 automotive-grade reliability certification, making it the world's only technology company to achieve automotive-grade standards for self-developed chips across the entire chain of digital LiDAR transmission, reception, and processing. The successful automotive-grade certification of RoboSense's self-developed SPAD-SoC chip and two-dimensional addressable 2D VCSEL chip not only marks the maturity of RoboSense's new generation LiDAR chip technology, but also represents the first global breakthrough in automotive-grade certification for digital LiDAR core chips.

RoboSense's self-developed SPAD-SoC reception and processing chip (left) and two-dimensional addressable 2D-VCSEL chip (right) both passed AEC-Q series automotive-grade certification.

AEC-Q certification serves as the global authoritative standard in the automotive electronics field, established by the Automotive Electronics Council (AEC) founded by Chrysler, Ford, and General Motors to create a common component qualification and quality system standard. It verifies product quality and reliability through over a hundred rigorous testing systems including extreme high and low temperatures, temperature cycling, accelerated life testing, mechanical shock, and EMC testing. The testing standards are stringent, requiring extreme temperature testing at -40°C and 125°C to ensure long-term stable operation of chips under various harsh conditions.

Traditional analog LiDAR requires seven key components to complete the construction of three major systems: transmission, reception, and processing. RoboSense's innovative digital architecture requires only two core chips to achieve the same functionality. Meanwhile, under the digital architecture, LiDAR line performance can exceed analog LiDAR by more than 10 times.

RoboSense's SPAD-SoC is the world's first chip to integrate SPAD reception array with data processing SoC, meeting AEC-Q certification and achieving mass production application. In terms of performance, this chip is equipped with a super WAVE Engine, featuring quad-core APU and dual-core MPU, which can improve effective target recognition rate by 99%. Combined with a high sampling rate of 44 billion times per second, it ensures more complete signal collection and effectively prevents missed detections. Its built-in 384-Core signal processing core efficiently processes signals in parallel, while providing 0.25cm echo output resolution, exceeding industry levels by 3 times.

In signal processing, its wide dynamic response range can effectively detect low-reflectivity objects and resist environmental pollution such as rain, fog, and dust, achieving fine differentiation detection of objects in complex scenarios and breaking through near-distance blind zone limitations. In packaging technology, RoboSense globally pioneered the integration of SPAD and SoC through 3D stacking into a single chip, adopting copper-copper bonding technology to provide each SPAD with independent processing circuits, achieving lossless and efficient signal transmission. The high integration advantage also reduces LiDAR system volume by more than 50%.

RoboSense's 2D VCSEL chip is the world's only mass-produced two-dimensional addressable VCSEL emission chip. This chip adopts two-dimensional addressable array scanning technology, successfully suppressing high-reflection expansion and solving the world-class industry challenge of single-photon crosstalk. The 2D VCSEL chip supports flexible matrix partition scanning, further promoting significant improvements in laser emission efficiency with energy utilization rates up to 70%, while simultaneously reducing power consumption and being more heat-friendly.

In integration technology, RoboSense's 2D-VCSEL chip is highly integrated through dual-chip SiP co-packaging design, combining the VCSEL laser array with the drive unit into one, significantly reducing PCB footprint by 70%. System parasitic parameters are greatly reduced, allowing chips to interconnect directly on substrates, improving emission circuit performance. Laser pulse width is stably controlled within 2 nanoseconds from the discrete 5 nanoseconds, significantly improving ranging accuracy.

Behind the leading performance advantages of self-developed chips lies the support of RoboSense's three original patent technologies:

• Crosstalk Immunity Encoding (CIE) technology provides independent encoding for each transmission signal, breaking through environmental interference such as rain, fog, and strong light, improving anti-sunlight noise by 95% and anti-interference crosstalk by 90%.

• PicoSecond Timing Technology (PTT) improves each transmission deviation from nanosecond level by nearly a thousand times to picosecond-level phase accuracy, improving ranging accuracy by 15 times to 3cm while reducing power consumption.

• Adaptive Fusion Algorithm (AFA) can adaptively extract echoes, expanding LiDAR measurement range and filtering optimized signal optimal parts, successfully reducing blind zones by 70 times.

RoboSense's original patent technologies and core chips jointly construct the underlying technological barriers of digital LiDAR. Compared to traditional analog LiDAR, RoboSense's digital LiDAR has generational advantages including high resolution, strong anti-interference capability, and compact size, with breakthrough leading product performance.

Currently, based on full-stack chip architecture, RoboSense's E platform and EM platform digital LiDAR matrix have fully entered automotive-grade mass production stage, comprehensively covering upgrade demands across multiple scenario applications including Robotaxi, ADAS, and service robots.

As flagship products in the matrix, RoboSense's high-performance LiDAR EM4 and E1 strong combination has become the preferred choice for L3 and L4 autonomous driving LiDAR applications. To date, EM4 has secured main LiDAR designations from multiple leading North American and domestic Robotaxi customers, while E1 has become the preferred solution for Robotaxi solid-state blind-spot LiDAR large-scale operational deployment, receiving formal designations from 8 global leading companies. Additionally, the "strongest partnership" formed by EM4+E1 is becoming the preferred configuration for new generation Robotaxis.

Pony.ai's 7th generation vehicle model is equipped with 4 E1 units for large-scale road testing, while the Robotaxi jointly created by Didi and GAC Aion uses 6 E1 units to build a near-field perception protection network.

EM platform series products have cumulatively received designations for 45 vehicle models from 8 OEMs, with the true 192-line EMX receiving exclusive supply designations for up to 32 vehicle models from a certain global leading new energy vehicle company.

Currently, Zeekr 9X and IM LS6 have been launched and delivered equipped with RoboSense EM platform customized 520-line digital LiDAR.

Zeekr 9X equipped with RoboSense EM platform customized 520-line digital LiDAR.

From L2-level assisted driving to L4-level autonomous driving, RoboSense will continue to lead LiDAR underlying chip technology innovation, providing higher quality, higher performance, and more safe and reliable high-precision perception protection for full-scenario driving safety.