RK3588 AMR Robot ODM Case Study: AI Vision, SLAM and ROS

Autonomous Mobile Robots (AMRs) require more than basic motor control. They need to process multiple sensor inputs, perform real-time AI inference, run SLAM algorithms, and make navigation decisions in dynamic environments.

For this project, KICKPI developed a high-performance RK3588 robot control board for an AMR application targeting smart warehouses, factories, and logistics centers.

The solution integrates AI vision, SLAM, ROS, LiDAR, depth cameras, IMU, multi-camera input, and high-speed wireless communication into a high-performance embedded computing platform.

1. Project Overview

Application	Autonomous Mobile Robot (AMR)
Industry	Smart Warehousing / Logistics
Processor	Rockchip RK3588
Operating System	Ubuntu / Linux
Project Type	Robotics Mainboard ODM
Key Technologies	AI Vision, SLAM, ROS, Multi-Sensor Fusion
Target Applications	Warehouses, Factories, Logistics Centers

Unlike traditional AGVs that mainly execute predefined movement commands, AMRs require greater computing capability for environment perception, localization, navigation, and autonomous decision-making.

For applications focused on automated material handling and industrial AGV control, see our RK3568 AGV Robot ODM Case Study.

2. Customer Requirements

The customer required a computing platform capable of supporting multiple sensors and autonomous navigation functions.

Key requirements included:

• Autonomous navigation
• LiDAR-based SLAM
• AI vision and object detection
• Dynamic obstacle avoidance
• Multi-camera input
• Depth camera integration
• IMU and encoder data processing
• 5G / Wi-Fi connectivity
• ROS support
• Real-time path planning and scheduling

The core challenge was to build a reliable embedded platform capable of processing these workloads simultaneously.

3. RK3588-Based AMR Computing Architecture

KICKPI selected the Rockchip RK3588 platform to provide the computing performance required for AI-enabled AMR applications.

The system integrates cameras, depth sensors, LiDAR and IMU data into the RK3588 computing platform.

The platform can also integrate 5G, Wi-Fi 6, Ethernet and CAN interfaces for communication between the robot, peripheral devices and cloud-based management systems.

4. Key Technical Challenges

Multi-Camera Data Processing

AMRs may need to process multiple video streams simultaneously, including:

• RGB cameras
• Depth cameras
• Front-view cameras
• Rear-view cameras

The hardware design therefore needs to consider CSI resources, USB bandwidth, DDR bandwidth and ISP capabilities to ensure stable multi-camera operation.

AI Vision Processing

The robot may use AI vision for:

• Human detection
• Object detection
• Obstacle recognition
• Object tracking

The RK3588 NPU can be utilized for AI inference, helping reduce the processing load on the CPU and improve the efficiency of edge AI applications.

SLAM and Multi-Sensor Fusion

For autonomous navigation, sensor data from LiDAR, cameras, IMU and wheel encoders needs to work together.

The system architecture supports the processing chain:

Localization → Mapping → Path Planning → Dynamic Obstacle Avoidance

This makes the platform suitable for AMR applications that require autonomous navigation in changing environments.

5. KICKPI RK3588 Robotics ODM Solution

For this AMR project, KICKPI developed the embedded computing platform around RK3588 with interfaces and features designed for robotic applications.

Key integration options include:

• Multiple CSI camera interfaces
• USB 3.0
• Gigabit Ethernet
• CAN
• Wi-Fi 6
• 5G / M.2 expansion
• IMU interface
• Motor-control interfaces
• Ubuntu / Linux
• ROS-based robotics applications

In addition to hardware design, the ODM development process can include system integration, BSP/software development, prototype testing and production support.

The hardware was also designed with attention to thermal management and long-term operating stability, which are important for robots deployed in warehouses, factories and industrial environments.

6. From Development Board to Custom Robot Mainboard

For robotics companies developing an AMR product, selecting the processor is only one part of the project.

The final robot controller needs to balance:

Computing Performance + AI Processing + Sensor Interfaces + Communication + Thermal Design + Reliability

A customized RK3588 platform can provide a foundation for integrating these functions into a dedicated robot control system.

KICKPI provides custom embedded board development and ODM services for robotics, industrial automation, AI edge devices and smart logistics applications.

Looking for an RK3588-Based Robot Controller?

If you are developing an AMR, autonomous robot, warehouse robot or industrial mobile robot, KICKPI can help with customized hardware and software development based on Rockchip platforms.

Contact KICKPI to discuss your hardware, software and customization requirements.