RK3588 Smart Retail and Self-Service Kiosk Solutions: Technical Selection Guide

Bottom line: RK3588 is a mainstream choice for self-service terminal mainboards — 8-core CPU, 6 TOPS NPU, dual independent displays and multiple camera interfaces let one chip handle display, recognition, payment and interaction, cutting total system cost by 20–30% versus a "mainboard + separate compute module" approach.

Self-checkout stations, smart vending cabinets, self-ordering kiosks and face-payment terminals share remarkably similar mainboard requirements: drive a high-resolution display, accept cameras for recognition, interface with payment and printing peripherals, and run 24/7. This is exactly where RK3588 excels. This article breaks down selection criteria by device type.

1. Configuration requirements for four mainstream terminal types

Device type	Display	Cameras	Typical configuration
Self-checkout	21.5–27" main + customer-facing secondary	1 (scan / face)	RK3588 + 4/8 GB RAM + 32/64 GB storage
AI vending cabinet	10–21" touchscreen	2–4 (product recognition)	RK3588 + 8 GB RAM, NPU for recognition
Self-ordering kiosk	Main + secondary independent display	1 (face optional)	RK3588 + 4 GB RAM + rich peripheral I/O
Face-payment terminal	8–15" portrait	1 RGB (+ optional 3D structured light)	RK3588 + liveness detection; payment certification required

2. Why RK3588 rather than RK3568 or an industrial PC

Option	Strengths	Limitations
RK3588	Strong 8-core CPU, 6 TOPS NPU, dual independent display, multiple MIPI	Slightly higher cost than RK3568
RK3568	Lower cost and power	Only 1 TOPS NPU — struggles with AI product recognition; mostly no dual display
Industrial PC	Mature ecosystem, expandable compute	High power, large footprint, higher cost, more failure points (fans, disks)
Repurposed tablet	Cheap	Insufficient I/O, not built for continuous operation, poor industrial peripheral compatibility

Decision rule: AI recognition (product or face) → RK3588; display and interaction only → RK3568 suffices; CUDA ecosystem required → industrial PC.

3. Hardware design considerations

3.1 Dual independent display (standard for self-service)

The operator screen plus customer screen in a self-checkout unit is implemented directly with RK3588's multiple VOPs — no extra display chip needed, and the two outputs can show different content. Common pairing: MIPI-DSI (main) + HDMI or eDP (secondary). Budget DDR bandwidth when both refresh at high resolution simultaneously.

3.2 Cameras and recognition

• Barcode scanning: MIPI or USB camera with classical decoding — low cost, fast (<100 ms)
• Face recognition / liveness: requires IR or 3D structured light; NPU handles detection and feature extraction; mind privacy compliance
• Product recognition (vision cabinet): 2–4 cameras with NPU classification/detection; accuracy is the key metric — validate with real samples first

3.3 Peripherals and interfaces (where design skill shows)

• Payment: USB (scanner, card reader), serial (POS), optional secure element
• Printing: thermal receipt printer over serial/USB — mind paper feed logic and cutter timing
• Weighing (cabinets, fresh produce): RS232/RS485 scale integration
• Door control (vending): multiple GPIO + motor driver + door sensor feedback, with power-loss protection
• Network: gigabit Ethernet plus optional 4G/5G module with automatic failover

3.4 Stability and operations

• 24/7 operation: fanless design, wide-temperature components, watchdog, self-recovery
• Scheduled power on/off: follow business hours to extend panel life
• Remote management: OTA updates, log reporting, remote screenshot and reboot (essential for chains)
• Power-loss protection: protect the filesystem (read-only root plus a data partition is recommended)

4. Payment security and compliance (confirm early)

Face-payment terminals have specific compliance requirements, usually driven by the payment institution or acquirer. Confirm the checklist with your payment provider before starting hardware design:

• Biometric data: comply with personal information protection regulations; typically extract features locally and never store raw face images in plaintext
• Payment certification: face-payment devices generally require security certification from payment networks or providers
• Keys and encryption: store keys in a secure element, not ordinary flash
• Notice and consent: the UI needs a clear privacy notice and user authorization flow

Self-checkout units doing barcode payment only (no face) face simpler requirements and are not subject to the biometric provisions above.

5. FAQ

Q1: RK3588 or RK3568 for a self-service terminal?
It depends on whether you need AI. Barcode scanning with dual display can use RK3568 to save cost; product recognition, face recognition or multi-camera analytics need RK3588. If AI features might be added later, go straight to RK3588 to avoid a board redesign.

Q2: How accurate is product recognition in an AI vending cabinet?
It depends on SKU count, visual similarity and capture conditions. In practice, fewer than 100 visually distinct SKUs can exceed 95%; similar packaging (same brand, different flavors) needs stronger lighting and multiple angles, or a weighing-based cross-check.

Q3: Does it support face payment?
Technically RK3588 can run face detection and liveness algorithms, but face-payment devices must pass security certification from a payment provider — confirm the path with your acquirer. We supply the hardware platform and algorithm adaptation; certification is led by the payment provider.

Q4: Android or Linux?
The vast majority of self-service terminals use Android — faster UI development, mature touch interaction, better peripheral driver support. Linux suits headless or light-interaction use cases (recognition boxes). For localization requirements, UOS or Kylin are options.

Q5: How long from design freeze to mass production?
Building an interface board and enclosure around an existing SoM typically reaches small-batch pilot in 6–10 weeks; a fully new mainboard takes 10–16 weeks. Projects with custom enclosures and certification (CCC etc.) take longer. Final timing depends on complexity.

6. Typical deployments

• Convenience store self-checkout: dual screen + scanning + card payment + receipt printing, one unit replacing a staffed lane
• Unattended retail cabinets (offices, scenic spots): vision plus weighing cross-check, scan to open and auto-settle on close
• Fast-food self-ordering: dual independent displays + touch + payment + queue-system integration
• Hospital / bank self-service terminals: identity verification + printing + dual-screen guidance
• Hotel self check-in: face verification + ID card reading + key card dispensing

7. Hongyin Tech's capabilities and how we work

Shenzhen Hongyin Technology specializes in SoMs, mainboards and full hardware-software customization on Rockchip (RK3588/RK3576/RK3566/RK3568) and HiSilicon platforms, with complete delivery experience in self-service terminals, smart retail, commercial displays and industrial equipment.

• Hardware: SoMs, interface boards and enclosure design, including peripheral interfaces and power
• Software: Android / Linux / domestic OS customization, driver porting, AI algorithm deployment, OTA updates
• Delivery: schematics/PCB, BOM, driver source, production test plans and technical support
• Response: requirements assessment within 1 business day

Contact: phone +86 13728726926 | email sales@hoyin-tech.com | website www.hoyin-tech.com | address: Room 301, Building A, Fanrong Science Park, Sanwei Community, Hangcheng Street, Bao'an District, Shenzhen, China.

If you are planning a self-service terminal project, tell us the device form factor, screen size, peripheral list and whether AI recognition is needed, and our engineers will propose a concrete hardware configuration and timeline.