Overview
In the rapidly evolving world of warehouse automation, fulfillment operations are under increasing pressure to deliver faster, more accurate order processing at scale. A leading automation provider in Europe sought to enhance its robotic piece-picking system with reliable, low-latency AI processing at the edge to support real-time vision-based decisions. By integrating Premio’s industrial edge AI computer, the deployment achieved high-performance inference, seamless multi-camera connectivity, and automation-ready integration within demanding warehouse environments.
Challenges
- Need for high-performance GPU computing to process real-time AI vision workloads
- Insufficient connectivity for multiple PoE industrial cameras within the picking cell
- Latency constraints impacting real-time object detection and robotic response
- Limited flexibility in mounting within compact robotic picking stations
- Requirement for reliable operation in continuous 24/7 industrial environments
Solution
- Premio’s Mid-Range Edge AI Computer powered by NVIDIA Jetson Orin Nano and NX modules (JCO-3000-ORN-B Series)
- NVIDIA Jetson Orin NX 8GB delivering up to 70 TOPS AI performance
- 4x PoE RJ45 ports supporting up to 120W power budget for machine vision
- Automation-ready I/O including USB, HDMI, isolated DIO, serial, and CAN Bus
- Fanless rugged design with wide temperature support and OOB management
Benefits
- Real-time accurate robotic picking decisions
- Simplified multi-camera system architecture
- Reliable 24/7 warehouse operation

Company Overview
The company specializes in advanced warehouse automation solutions, focusing on robotic systems that optimize order fulfillment workflows. It develops intelligent picking technologies that integrate AI, vision systems, and robotics to improve efficiency and scalability. With a strong focus on innovation, it continues to expand its capabilities in autonomous logistics and smart warehousing.
The Challenges
High Performance AI Processing Requirements
Robotic piece picking relies heavily on real-time image processing to identify, locate, and classify items within dynamic bin environments. The system required a powerful GPU-enabled platform capable of handling multiple AI inference workloads simultaneously. Without sufficient compute performance, picking accuracy and throughput would be compromised.
Multi Camera Connectivity Limitations
The picking cell required several industrial cameras positioned at different angles to ensure precise object recognition and depth perception. Traditional systems lacked the ability to efficiently power and connect multiple cameras without added complexity. This created integration challenges and increased system cost.
Real Time Latency Constraints
In automated picking workflows, even slight delays between image capture and robotic action can reduce efficiency or lead to mispicks. The system needed ultra-low latency processing at the edge to ensure immediate decision-making. Cloud-based or centralized processing approaches introduced unacceptable delays.
Integration Within Compact Robotic Systems
Robotic picking stations are often space-constrained, requiring compact and flexible hardware solutions. Bulky or difficult-to-mount systems limited deployment options and increased installation complexity. The solution needed to fit seamlessly within existing robotic infrastructure.
Continuous Industrial Operation Requirements
Warehouse automation systems operate around the clock, often in environments with dust, vibration, and temperature variation. Consumer-grade hardware could not meet reliability expectations. A rugged, industrial-grade solution was essential to maintain uptime and reduce maintenance.
The Solution
Premio’s JCO-3000-ORN-B Series
Premio’s Mid-Range Edge AI Computer (JCO-3000-ORN-B Series) was deployed as the central processing unit within the robotic picking cell. Installed directly at the edge, the system processes visual data from multiple cameras in real time.
NVIDIA Jetson Orin NX 8GB for AI Inference
Powered by NVIDIA Jetson Orin NX 8GB, the system delivers up to 70 TOPS of AI performance, enabling fast and efficient object detection and classification.
Integrated 4x PoE Camera Connectivity
The system supports up to four PoE-enabled RJ45 ports with a total power budget of up to 120W, enabling direct connection and power delivery to multiple industrial cameras. Cameras are positioned across the picking cell to capture bin contents, depth, and picking angles. This reduces cabling complexity while ensuring stable, high-speed data transmission for real-time vision processing.
Automation-Ready Integration with Rich I/O
Designed for seamless industrial automation, the system includes USB 3.2, HDMI, isolated DIO, RS-232/485 serial ports, and CAN Bus for direct integration with robotic controllers, PLCs, and sensors. The 16-channel isolated DIO enhances signal reliability by protecting against electrical noise in industrial environments. This automation-ready architecture ensures synchronized communication between vision systems and robotic actuation.
Rugged and Connected Industrial Design
Built with a fanless rugged enclosure, the system operates reliably in temperatures ranging from -20°C to 55°C and withstands shock and vibration in demanding environments. Integrated Out-of-Band management enables remote monitoring, diagnostics, and system updates for continuous operation. This ensures long-term reliability and simplified fleet management across warehouse deployments.
The Benefits
Improved Picking Accuracy
Real-time AI processing enables precise object detection and classification, reducing picking errors. This leads to higher order accuracy and improved customer satisfaction.
Streamlined System Architecture
Integrated PoE and rich I/O eliminate the need for additional hardware components, simplifying deployment and reducing system complexity. This results in faster integration and lower total cost of ownership.
Reliable Continuous Operation
Industrial-grade durability combined with remote management capabilities ensures stable 24/7 performance. With support from Premio’s Los Angeles-based team, the deployment maintains long-term operational efficiency.
Conclusion
By integrating edge AI computing into its robotic picking system, the automation provider significantly enhanced performance, accuracy, and scalability. The deployment enables real-time decision-making while simplifying system architecture. This approach establishes a scalable foundation for next-generation intelligent warehouse automation.