Within enclosed campuses, factories, and smart communities, demand for safe, precise, and unmanned last-mile logistics delivery continues to grow. The Prompt delivery robot integrates multiple functions including autonomous navigation, cargo handover, continuous video security, and personnel identification. Its core challenges lie in achieving reliable all-weather perception in complex dynamic environments, precise interaction with specific recipients, and maintaining 360-degree security monitoring while in motion. This demands a visual system that not only “sees” but also ‘understands’ and “recognizes” accurately. Our integrated 3D facial recognition vision module, deeply optimized for smart locks and security applications, perfectly addresses the dual demands of mobile robots for secure delivery and active monitoring in dynamic, unstructured environments. Featuring integrated 3D depth sensing, powerful localized AI computing, dual-spectrum (IR+RGB) all-weather imaging, and a compact, low-power design, it serves as the core “visual brain” for next-generation intelligent mobile service robots.
I. Stereoscopic 3D Depth Perception with Wide Field-of-View Coverage: Enabling Precise Navigation and Obstacle Avoidance in Dynamic Environments
Robots navigating campuses encounter dynamic environments including pedestrians, vehicles, and temporary obstacles. Relying solely on 2D vision fails to accurately gauge obstacle distance and volume, increasing collision or entrapment risks. Simultaneously, robots require a broad forward field-of-view for proactive path planning.
This module employs a highly integrated dual-camera infrared solution. The two cameras form a fixed optical baseline, enabling real-time disparity calculation to directly output precise depth information and construct a 3D point cloud of the robot's forward environment. With a horizontal field of view (FOV) of 76.8° and a diagonal FOV of 88°, it provides extensive forward perception, allowing the robot to detect pedestrians or obstacles approaching from the side in advance. 3D depth perception grants the robot genuine spatial understanding, enabling it to not only recognize “an object ahead” but also determine “this object is 1.5 meters away, approximately 0.5 meters tall, and likely a parcel box.” This allows for smarter, safer obstacle avoidance and path planning decisions, ensuring stable operation in complex, dynamic environments.
II. Powerful Localized NPU and Multimodal Biometric Recognition: Ensuring Reliable, Privacy-Secure Interactive Delivery
A core function of the Prompt robot is accurately delivering packages to designated recipients. This requires the robot to swiftly and precisely identify individuals in natural postures while maintaining high anti-counterfeiting capabilities to prevent fraudulent claims. Simultaneously, considering privacy and network dependency, the identification process must be completed offline on the device itself.
This module embeds a high-performance NPU (Neural Processing Unit) and integrates proprietary advanced algorithms, supporting dual biometric recognition of facial features and palm prints. When the robot approaches the recipient, the module completes rapid facial matching within 1.1 seconds, achieving an accuracy rate of TAR: 98% @ FAR: 1 in 1 million, effectively preventing misidentification. More importantly, it possesses robust liveness detection capabilities (TAR: 98% @ FAR: 1 in 100,000), effectively resisting attacks from photos, videos, or even 3D masks. This ensures the cargo compartment unlocks only when a “real person is present.” All feature data processing and comparisons occur locally within the robot, eliminating the need for cloud uploads. This safeguards user privacy while ensuring reliable functionality in areas with poor network signals (e.g., underground garages, factory workshops). This provides critical technological assurance for unmanned, trustworthy “face-to-face” parcel handover.
III. IR+RGB Dual-Spectrum Imaging with Superior Environmental Adaptability: Supporting 24/7 Uninterrupted Security Monitoring
The robot must perform uninterrupted video surveillance and personnel detection tasks under any lighting conditions—daytime, nighttime, backlighting, shadows, and more. Traditional RGB cameras fail in low light and are prone to overexposure in bright light.
This module innovatively integrates dual IR (infrared) and RGB (color) sensors, equipped with an efficient 850nm wavelength infrared illuminator. During daylight or well-lit conditions, the RGB camera delivers full-color imagery capturing rich environmental details. At night or in extreme darkness, the active IR illuminator activates, enabling the IR camera to produce clear, noise-free monochrome images. This dual-spectrum collaborative mode ensures the robot achieves “24/7 clear vision.” Furthermore, its algorithms are specifically optimized for complex lighting conditions (low light, strong light, backlight) and support Wide Dynamic Range (WDR). This enables balanced exposure and visible details in high-contrast scenarios like entrances/exits or shaded areas under trees, providing a stable and reliable visual stream for continuous video security analysis and event tracing.
IV. High Integration, Low Power Consumption, and Industrial-Grade Reliability: Meeting the Demanding Embedded Requirements of Mobile Robots
As the core sensor for mobile robot platforms, the module must be compact, low-power, and capable of withstanding continuous vibration during movement, outdoor temperature and humidity fluctuations, and potential environmental interference (such as static electricity).
Featuring an integrated dual-camera co-substrate design, the module achieves extreme compactness for easy integration into the limited space of a robot's head or mast. Its average power consumption is below 3.5W, with an average operating current of just 200mA@5V, minimally impacting the robot's precious battery life. Constructed to industrial and security standards, the module operates across a wide temperature range of -25°C to 70°C. It has passed rigorous electrostatic discharge (ESD) testing, including ±8KV contact discharge and ±16KV air discharge, ensuring exceptional environmental adaptability and interference resistance. This guarantees stable, reliable operation throughout seasonal changes, inclement weather, and extended continuous operation in outdoor campus environments.
In summary：
This integrated 3D vision module delivers a comprehensive visual solution for intelligent delivery robots like Prompt. It combines binocular depth perception for real spatial awareness, localized AI and multimodal biometrics for secure identity verification, dual-spectrum all-weather imaging for continuous monitoring, and a highly integrated, low-power, and reliable design tailored for demanding mobile embedded scenarios. This solution unifies "environmental perception, navigation and obstacle avoidance, personnel recognition, and security monitoring.“ Its profound capabilities will transform delivery robots from mere ”mobile cargo containers" into autonomous mobile entities with environmental understanding, capable of safe and intelligent human interaction while continuously safeguarding surrounding security. This significantly advances the technological upgrade and large-scale application of unmanned logistics and integrated security management within enclosed areas.
