By ATN Editorial Team
Summary: DiDi Autonomous Driving’s next-generation R2 robotaxi, co-developed with GAC Aion, incorporates Raytron’s Horus 640D thermal camera as part of a 33-sensor perception suite designed to provide 360-degree blind-spot-free coverage. The thermal imaging system is specifically intended to address detection gaps that affect visible-light cameras and LiDAR in low-light conditions, glare, fog and dust.
Key engineering takeaway: The Horus 640D uses a 12μm 640×512 automotive-grade infrared detector paired with a second-generation ASIC, detecting long-wave infrared radiation to identify vulnerable road users at up to 300 metres. A shutterless calibration algorithm eliminates the periodic image freeze associated with conventional thermal cameras, maintaining continuous output — a significant requirement for real-time autonomous driving perception pipelines.
Why it matters: Thermal imaging addresses a well-documented weakness in current autonomous driving sensor stacks: the inability of cameras and LiDAR to reliably detect pedestrians and other VRUs in degraded visibility conditions. As Level 4 systems move toward production deployment in uncontrolled urban environments, the ability to handle rare but safety-critical edge cases — such as detecting a person lying in a poorly lit road — becomes a key factor in demonstrating the robustness regulators and the public will expect.
Didi Autonomous Driving’s next-generation R2 robotaxi incorporates Raytron’s Horus 640D thermal camera.
Didi Autonomous Driving has unveiled the R2, its next-generation mass-produced robotaxi co-developed with GAC Aion, featuring Raytron’s automotive thermal camera as a key component of its perception system. Integrated alongside a sophisticated suite of 33 sensors, Raytron’s thermal imaging helps complete a 360° blind-spot-free perception architecture, enabling the autonomous vehicle to maintain reliable detection even in low-light and adverse weather conditions.
Why Thermal Imaging Is Essential for L4 Autonomy?
In the complex landscape of urban mobility, standard sensors—such as cameras and LiDAR—often encounter invisible barriers. Visible-light cameras can be compromised by the sudden glare of tunnel exits or oncoming high beams, while heavy fog, smoke, and dust can scatter LiDAR beams, significantly shrinking their effective detection range. By detecting long-wave infrared radiation rather than relying on reflected light, infrared thermal cameras can identify life signatures—pedestrians, cyclists, and animals—from hundreds of meters away, regardless of clothing color, camouflage, or total darkness.
Horus 640D: Raytron’s Breakthrough in Automotive Thermal Imaging
Powered by Raytron’s Horus 640D, the robotaxi can see beyond the reach of headlights. The thermal module integrates a 12μm 640×512 automotive-grade infrared detector with a proprietary second-gen ASIC chip, making it a compact yet robust automotive sensor. Moreover, Raytron’s self-developed shutterless algorithm eliminates the jarring “image freeze” that once plagued thermal cameras. This infrared thermal module enables the Robotaxi to detect vulnerable road users (VRUs) up to 300 meters away, giving autonomous systems precious extra seconds to sense, think, and act safely.
“In a real-world test in Yizhuang, dim lamplight left cameras blind to a drunk person lying by the road, while LiDAR captured only sparse, ambiguous points,” noted Zhang Bo, CEO of Didi Autonomous Driving. “The infrared sensor identified the heat signature at 100 meters, allowing the vehicle to decelerate safely. It resolves rare but critical ‘one-in-a-ten-million-mile’ risks.”
For more sensor technology news, click here.
