The Feasibility of Using RK3588J to Build Robot Control Board

Under the impetus of the smart AIoT technology wave, intelligent robots are making inroads across various industries with an unstoppable momentum. They range from attentive service robots in hotels and exhibition venues to street-cleaning environmental protectors, from meal delivery solutions to precise smart lawn mowing operations. These robots also excel in interactive gaming challenges, complete disinfection processes, automated industrial logistics—particularly represented by AGV (Automated Guided Vehicles)—and even drone inspections soaring the skies and low-speed unmanned retail vehicles offering convenient services. They are quietly transforming our lifestyles and industrial landscapes.

What, then, is the core force driving these robots to operate efficiently and display such astounding capabilities? This article delves into Dusun IoT‘s latest robot control board solution based on the Rockchip RK3588, showcasing its potential as the optimal choice for these applications.

What Features Are Required for an Robot control board?

As the central brain of a robot, an intelligent robot control board must possess a series of key features to ensure efficient and smart operations. Here are the main characteristics required:

NPU Integration

• Support for Offline Large Models: The control board must be capable of handling complex AI tasks, including the ability to run large neural network models even without stable internet connectivity.
• Object Recognition Algorithms: It should either have built-in or support the loading of multiple algorithms for object recognition, facial recognition, and scene understanding to enhance the robot’s perception capabilities.
• Monocular Distance Measurement Algorithm: Support for distance measurement using a monocular camera improves the robot’s spatial awareness.
• Flexible Computation Scaling: Provides flexible computational power expansion, allowing the adjustment of the NPU’s resources according to task requirements.

Also read: Using RK3588 for AI Camera Development Because of Built-in Tri-core NPU

Operating System Support

• Support for Linux and Android Systems: Offers complete support for both Linux and Android operating systems, providing developers the freedom to choose the most suitable platform for development.
• User-Friendly Android UI Design: For Android-based robots, it offers tools and frameworks for designing user-friendly interfaces.
• Linux Adaption for ROS/ROS2: Fully compatible with ROS (Robot Operating System) and ROS2, offering robust ecosystem support for robot application development.

Also read: DSGW-380 AI-based Industrial Gateway Supporting ROS Robot Operating System

Intelligent Voice Interaction

• Multi-Microphone Array: Configured with multiple microphone arrays to enhance the clarity and directionality of voice capture.
• Offline Voice Wake-up: Supports offline voice recognition and wake-up functionality, allowing the robot to be activated with specific wake-up words even in the absence of network connectivity.

Case study: Home appliance control enabled by AI offline speech recognition modules with multi-microphone array

AMP Architecture Support

• Strong Real-Time Performance: Ensures high responsiveness and real-time processing of critical tasks by the robot control board.
• High Stability: Employs stable hardware and software architecture designs to reduce system failure rates and improve the reliability of robot operations.
• Lower Hardware Costs: Optimizes the Asymmetric Multiprocessing (AMP) architecture to balance performance and cost efficiency.

Also read:

what is a heterogeneous multi-core SoC?

The communication stack for memory sharing on a heterogeneous multi-core SoC

ROS Compatibility

• Standard ROS/ROS2 Adaptation: Seamlessly integrates with ROS and ROS2, enabling the robot to easily access various libraries, tools, and applications within the robot ecosystem.
• Lidar Compatibility: Supports various lidar sensors (such as laser radar and millimeter-wave radar), enhancing the robot’s perception capabilities in complex environments.
• Compatibility with Rockchip Structured Light: Highly compatible with Rockchip’s self-developed structured light modules, achieving more precise 3D perception and measurement.

Why RK3588 is Ideal for Robot Control Boards?

• Powerful CPU Power: With CPU performance reaching up to 93K DMIPS, the RK3588 provides robust data processing capabilities for robot control boards, ensuring that robots can quickly respond to various commands and tasks.
• Superior GPU Performance: With a GPU performance reaching 512 GFLOPs, the RK3588 excels in complex graphics rendering and multi-image processing, significantly enhancing the visual perception and interactive capabilities of robots.
• Advanced NPU Capability: Featuring a tri-core NPU with 6 TOPs of computing power (INT8), the RK3588 supports a wide array of deep learning frameworks. This enables highly efficient and precise AI applications, such as achieving an inference speed of 59.6 fps with the YOLO-v8n model, crucial for real-time processing.
• Proprietary Rockchip Modules: The integration of Rockchip’s self-developed structured light modules and RK1808 computing sticks offers specialized hardware support, boosting the overall functionality and performance of robotic systems to new heights.
• Industrial-Grade Durability: The RK3588 is designed to withstand extreme temperatures, operating reliably between -40°C and 85°C. This ensures that robots equipped with the RK3588 can function stably in a wide range of challenging environments, solidifying its suitability for various demanding applications.

RK3588 Block Digram for Robot Control Board

• Central Processor: The RK3588 serves as the main processor on the control board, leveraging its multi-core CPU, GPU, and NPU to handle various computational tasks efficiently.
• High-Speed Memory Support: LP4/LP4X/LP5 DDR controllers provide fast dynamic random-access memory (DRAM) to meet the data processing demands of the RK3588.
• Persistent Storage: eMMC 5.1 Flash is utilized for storing the operating system, applications, and user data, offering reliable and persistent storage capabilities.
• Display Interfaces: HDMI/eDP interfaces are used to connect high-definition displays, enabling a visual interface for human-machine interaction and command control.
• Camera Inputs: MIPI CSI supports up to 8 camera inputs for environmental sensing and object recognition functions.
• Internet and Sensor Connectivity: RGMII 0/RGMII 1 can be used for Ethernet connections or to interface with smart sensors such as LiDAR, supporting SLAM (Simultaneous Localization and Mapping) and obstacle avoidance functions.
• Power Management and Audio: PMU IO connects to power management integrated circuits (PMIC), DC-DC converters, and low-dropout regulators (LDO) to provide stable power supply. It also supports audio codecs for connecting microphones (MIC) or speakers (SPK).
• Expandable Storage: SDMMC offers SD/MMC card slots for additional storage expansion or data backup purposes.
• Wireless Communication and Interface Support: SDIO/UART/PCIE 3.0 interfaces allow for the development or connection of WiFi, Bluetooth, and 5G modules, enabling wireless communication capabilities. They also support CAN interface development for motor control applications.
• Peripheral Connections: USB ports allow for connections to depth cameras (such as structured light, ToF, and stereo cameras), radar devices, and a USB HUB for expanding additional USB devices.
• Sensor Integration: I2C/SMBus connections are used for temperature sensors, battery management units (BMU), providing battery health monitoring and temperature control.
• Motor and Actuator Control: CAN interfaces connect directly or through a CAN controller for precise control of robot chassis motors, joint controls, and speed regulation.
• Proximity Beacon Interfaces: Infrared and ultrasonic sensor interfaces connect via GPIO or other specialized interfaces, aiding in short-range detection and obstacle avoidance.

Also read:

Bluetooth beacon used for retail and shopping proximity marketing

IoT in retail

BLE asset tracking

FAQs on Robotic control boards

Why Does A Robotic control boards Need to Support Structured Light Modules?

• Superior Depth Restoration: Structured light modules leverage optical encoding technology along with Rockchip’s neural network (NN) deep algorithms to accurately reconstruct the 3D information of scenes. This capability is crucial for tasks such as environmental sensing, obstacle avoidance, and navigation in robotics.
• Wide Field of View: With a horizontal field of view exceeding 100° and a vertical field of view of 90°, structured light modules can cover a broader area, enhancing the robot’s vision range and perception capabilities.
• Extensive Detection Range: Ranging from 0.3 meters to 3 meters, the detection range satisfies various robotic needs, from close-proximity interactions to long-distance navigation.
• Rich Interfaces: Offering various interfaces like USB-C and Ethernet, these modules facilitate easy connection and data transmission with other devices or systems, boosting the robot’s flexibility and expandability.
• Scalable Computational Power: Supporting a range of processor platforms from RV1126 to RK3588, these modules allow the selection of appropriate computational configurations based on specific robotic application requirements, achieving a balance between performance and cost.

How Do Voice Recognition Algorithms on Robot control board Empower Robotic Applications?

• Noise Reduction and Audio Capture Optimization: In noisy environments like subways, marketplaces, or inside vehicles, voice algorithms utilize AI combined with Active Noise Reduction (ANR) technology to effectively suppress transient noises (such as impact sounds and babble). This enhances the accuracy of speech recognition. For outdoor environments, audio capture feedback suppression technology can be employed to counteract interference from noises like car horns and wind.
• Multi-Microphone Array Audio Capture: By leveraging spatial information, techniques such as beamforming are used for voice enhancement, noise reduction, and localization, enabling robots to accurately recognize user commands even in complex environments.
• Voice Wake-up and Keyword Recognition: The control board supports high-precision voice wake-up and keyword recognition functions, achieving a human-machine interaction accuracy rate of over 95%. It supports multiple languages and recognizes hundreds of keywords, thereby enhancing the robot’s interactive experience and intelligence level.

How Does NPU-Accelerated Neural Network Algorithm Work on Robotic control boards?

• Localized Deployment of Large Models: The NPU built into high-performance processors like the RK3588 supports the local deployment and efficient inference of large-scale neural network models. For example, it can handle models ranging from 1.4 billion to 7 billion parameters locally, achieving a processing speed of 6 tokens per second, thereby meeting the real-time processing needs of robots for complex tasks.
• Monocular Distance Measurement: Utilizing vision algorithms and the computational power of the NPU, robots can quickly compute the distance and angle information of visual obstacles, providing precise data support for obstacle avoidance and navigation.
• Face Detection and Recognition: Supports face detection and recognition in various complex scenarios, including handling blurred faces and different angles. Additionally, it can perform face quality assessment, gender, and age recognition, enhancing the robot’s application value in security, service, and other fields.
• Multi-Target Recognition and Perimeter Detection: By accelerating neural network algorithms with the NPU, robots can quickly identify and monitor a variety of targets, including humans, faces, vehicles, non-motorized vehicles, and pets. This provides strong support for intelligent decision-making and action in various application scenarios.

Introducing to Dusun IoT’s RK3588 SoMs (with Carrier Board) and AGV SBC Solutions

As a long-term partner of Rockchip, Dusun IoT has been deeply involved in the field of intelligent robotic control boards, successfully utilizing the RK3588 chip to develop multiple high-performance solutions, including system on modules and embedded control boards. These advancements have empowered various industries to create high-performing AI robotic solutions like charging robot control board.

Dusun IoT offers customized hardware support and professional technical services, catering to the diverse needs of their clients. We look forward to collaborating with more partners to jointly drive innovation and development in smart robotics technology. Inquiries are welcome as we work together towards a smart future!

NXP system on modules

RK3588 SoM System on Module Overview

Dusun IoT’s RK3588 SoM Board is a high-performance, low-power embedded CoM designed around the Rockchip RK3588 processor. This board integrates the powerful computing capabilities of the RK3588 to provide robust hardware support for fields such as intelligent robotics, machine vision, and edge computing. The compact design and rich interfaces of the core board facilitate secondary development and system integration based on specific needs. Utilizing the RK3588 SoM Board, users can effortlessly deploy complex AI algorithms and deep learning models, enhancing the intelligence and competitiveness of their products.

RK3588-Based AGV and AMR Control Board Overview

Dusun IoT’s AGV and AMR Motheroard is a high-performance control board specifically customized for Automated Guided Vehicles (AGVs), developed based on the RK3588 Core Board. This control board leverages the RK3588’s exceptional performance, integrating various sensor interfaces (such as LiDAR and cameras), actuator control interfaces, and communication modules. It enables intelligent control over AGV functions like precise localization, path planning, obstacle navigation, and task execution.

The AGV Control Board, with its powerful data processing capabilities and AI algorithm support, allows AGVs to autonomously operate in complex and dynamic environments, enhancing production efficiency and safety. Moreover, the board supports remote monitoring and diagnostic functions, facilitating remote management and maintenance of AGVs for users.