Autonomous Parcel Delivery Robots in an Urban Community in Singapore

Advancements in AI, sensor, and connectivity technologies have led to the maturation of service-oriented intelligent robots, enhancing the quality of daily life through applications such as security, logistics, and caregiving. In a community in Singapore, the integration of Advantech’s AIR-030 Edge AI Inference Systems into delivery robots has markedly improved the efficiency of parcel distribution by property managers in the community. This enhancement reduces the complexity of their tasks, allowing residents to conveniently order and receive parcels. As a result, the overall quality of life has been elevated.

Challenge

Delivery robots are intelligent devices with various capabilities such as visual recognition, motion control, and wireless connectivity. These functionalities necessitate the integration of multiple devices like cameras, sensors, and wireless modules, demanding connection support for various I/O and expansion slots. These robots require high operational efficiency and a robust thermal design to prevent system failures due to overheating during highperformance tasks, which could otherwise degrade battery life and shorten the device’s lifespan. Therefore, embedding a highperformance, stable, and versatile computing system that supports a wide range of peripheral devices is imperative for delivery robots. Specializing in the manufacture of delivery robots, the customer faced challenges with their existing model, which was based on x86 architecture. These challenges included high power drawn, inadequate I/O support, high power consumption and heat dissipation issues impacting battery lifespan, and a lack of AI inference capabilities resulting in poor responsiveness. To overcome these obstacles, they adopted Advantech’s AIR-030 Edge AI Inference Systems to improve the robots’ obstacle detection and mobile safety features while increasing the battery life to extend the operational lifespan of the robots.

Solutions

The AIR-030 Edge AI Inference System is a high-performance system equipped with NVIDIA Jetson AGX Orin SOM, delivering up to 275 TOPS of AI performance with configurable power consumption between 15 and 50 Watts, making the system highly suitable for both training AI visual models and performing inferences.

To accommodate diverse functionalities, the AIR-030 provides extensive options for connecting multiple peripheral devices. This includes a variety of I/O ports, such as LAN, DIO, COM, CANBus, USB 3.2, and USB Type C, making it easy to connect to cameras, imaging sonar, visual sensor fusion, 3D lidar, and motion control sensors. For instance, connecting a camera allows for the acquisition of real-time video feeds, while integrating 3D LiDAR enables delivery robots to perform tasks such as positioning, mapping, and 360-degree obstacle detection. The robot can intelligently plan paths and maneuver around obstacles (be it humans or objects). Moreover, ultrasonic sensors act as a secondary obstacle detection method, halting all motor functions if an obstacle is detected within the predefined range (typically 0.5–1.0 meters ahead of the robot).

Regarding system stability, the AIR-030 features a rugged design and is certified with IEC 61000-6-4 for heavy industrial applications, ensuring stable operation and resilience in demanding industrial environments. This includes operating within a wide temperature range from -10°C to 60°C, thus mitigating overheating issues during highperformance computing. Furthermore, the interior of the delivery robot must have sufficient space for parcels, leaving little room for other components. Thus, the AIR-030’s compact size is an ideal solution for delivery robots.

Results

The implementation of these delivery robots has streamlined community mail and parcel delivery operations. In the community management office, property managers can easily track residents’ online delivery requests and deliver them with the robots, which autonomously navigate to the defined collection point at the specified time. Residents are notified to collect their parcels via an app and can unlock the robot’s compartment using a QR code. After collection, the robot autonomously returns to its base station, ready for the next dispatch. These smart delivery operations not only save time for management personnel but also ensure swift parcel delivery, allowing residents to enjoy a more sophisticated lifestyle.

Beyond community management, delivery robots can also assist retail outlets with smart delivery services by helping address delivery staff shortages, increasing productivity, strengthening supply chain logistics, and ultimately lead to enhanced customer satisfaction.

Looking to the future, the functionality of delivery robots will inevitably be expanded as retail logistics applications evolve. The AIR-030’s extensible and agile I/O design ensures it can be quickly adapted to accommodate many more peripheral devices for emerging applications. Combined with Advantech’s Edge AI SDK and Robotic SDK, engineers can rapidly develop multiple AI models to realize complex robotic functionalities in delivery robots, enabling highly customizable AI applications that businesses can deploy to enhance the customer experience.