Back to Blog

JETSON/RK3588+FPGA+AI+GMSL Embodied AI Visual Perception

#AI#FPGADev#AutonomousDriving#MachineLearning

Scenario Requirements: Embodied AI faces challenges in dynamic and complex real-world environments, simultaneously achieving "precise understanding of 3D space," "real-time correlation with body movements," and "flexible adaptation to scene changes." Currently, mainstream vision solutions are distributed in the robot's head and wrist. Head vision requires global perception to achieve environment modeling, object recognition, and interaction understanding; while wrist vision requires close-range fine perception to support grasping and positioning, operational precision control, and object state judgment. The two work together to meet the demands of global decision-making and local fine manipulation.

Robot Stereo Camera:

■ Supports 5MP high-definition resolution, 60fps frame rate, with an observation range covering 0.25m-10m, suitable for robot head-mounted observation; ■ Equipped with an industrial-grade image sensor, offering stronger environmental adaptability and anti-interference capabilities, stably outputting high-quality images; ■ Built-in IMU module to enhance motion vision accuracy; ■ Adopts 2-in-1 GMSL transmission technology for more efficient and convenient integration.

Low-Speed Unmanned Vehicle Omnidirectional Perception

Scenario Requirements: Low-speed unmanned vehicles need to achieve high-precision environmental perception and positioning capabilities in complex and dynamic open road environments (such as industrial parks, campuses, airports), while strictly controlling costs to adapt to large-scale commercial applications in low-speed scenarios like parks, logistics, sanitation, and mining areas.

Product Application: Sienovo's High-Dynamic-Range GMSL Series, targeting low-speed autonomous driving scenarios such as park logistics, sanitation, and inspection, introduces cost-effective 1MP/3MP surround-view GMSL cameras. These cameras feature compact size, lightweight design, low power consumption, and ultra-low latency. They effectively support unmanned vehicles in performing precise driving, autonomous parking, and intelligent cleaning tasks in complex environments, enhancing operational reliability and efficiency. ■ High-Definition Imaging: Optional resolutions include 1MP (1280H x 960V), 2MP (1920H x 1280V), and 3MP (1920H x 1536V), with a smooth frame rate of 30fps. ■ High-Bandwidth Transmission: Utilizes GMSL 2 communication protocol for data transmission, offering high bandwidth and low latency. ■ High Dynamic Range: 120dB high dynamic range, capable of presenting clear details even in backlit environments. ■ Automotive-Grade High Protection and Reliability: Features IP67 high protection performance, adaptable to harsh working environments.