RoboSense launches AC2 vision system at IROS 2025

Shanghai (Gasgoo)- At the 2025 International Conference on Intelligent Robots and Systems (IROS 2025), RoboSense unveiled the AC2 — the latest addition to its Active Camera lineup, marking another major step in its robotics vision technology roadmap.

Positioned as "the operational eye" of robots, the AC2 integrates a solid-state dToF (direct Time-of-Flight) sensor, binocular vision, and an inertial measurement unit (IMU). This combination allows the system to output either fused or independent streams of depth, image, and motion data, delivering a complete and coherent field of perception.

The AC2 introduces a redesigned solid-state dToF depth-sensing architecture that eliminates one of the biggest challenges in 3D vision—the decline in accuracy with distance. It achieves millimeter-level precision of ±5 mm across an 8-meter range. By incorporating a binocular vision system, the AC2 also enhances near-field accuracy within one meter. A built-in RGB camera with 1600×1400 resolution supplements depth data with high-definition color imagery, enabling more refined and detailed perception in complex environments.

To boost operational efficiency, the AC2 offers one of the industry's widest fields of view for 3D vision sensors—120° by 90°. Whether operating in fusion or standalone mode, the camera allows robots to perceive a larger workspace at once, improving recognition speed and task performance.

Despite its performance, the AC2 maintains a compact footprint of 100×30×45 mm, making it adaptable to various form factors—from humanoid robot heads and robotic dogs to the narrow end-effectors of mechanical arms.

Alongside the hardware release, RoboSense expanded its "AI-Ready" developer ecosystem to complement the AC2's capabilities. The update introduces open-source algorithms for pose estimation and human skeleton recognition—applications particularly relevant to service and humanoid robots. The AI-Ready suite includes AC Studio software tools, detailed documentation, and multi-scenario datasets, helping developers accelerate product development and reduce integration complexity.

The AC2 has already undergone preliminary validation and RoboSense has begun inviting academic and industry developers to participate in closed beta testing.

Earlier this year, RoboSense redefined its corporate positioning as a provider of "incremental components and solutions" for intelligent robotics. Its goal is to offer universal motion and manipulation solutions for robot OEMs, while developing scalable modules that reduce costs and shorten development cycles—ultimately helping accelerate commercialization across the robotics industry.

The company's robotics portfolio now spans multiple core technologies, including LiDAR, dexterous robotic hands, Active Camera vision systems, force sensors, high-power-density linear motors, and robot domain controllers.

RoboSense launched its first Active Camera model, the AC1, in March this year. The AC1 features a 120°×60° depth field of view and a 144°×78° RGB field of view, with a maximum range of 70 meters.

In the LiDAR segment, RoboSense has rolled out products such as the E1R and Airy. The E1R, powered by RoboSense's proprietary SPAD-SoC chip, features a 120°×90° field of view and 144 scan lines. Built to automotive-grade standards, it supports a wide range of applications in industrial, commercial, and embodied-intelligence robots. The company has partnered with Mammotion to co-develop a high-end intelligent lawn-mowing robot powered by solid-state LiDAR, with an initial agreement for 1.2 million units over the next three years.

The Airy model—touted as the world's first hemispherical digital LiDAR—measures just 60 mm in diameter, 63 mm in height, and weighs under 240 grams, enabling flexible integration into compact robot designs. Offering a 360°×90° field of view and a 120-meter coverage diameter, Airy is suited for humanoid, delivery, quadruped, and household service robots.