Shenzhen UBTECH Robotics Corp., Ltd. is a global high-tech enterprise integrating R&D of artificial intelligence and humanoid robots, platform software development and application, and product sales. Starting in 2008 with the R&D of servo motors, the core power source for humanoid robots, UBTECH has gradually launched consumer-grade humanoid robots, commercial service humanoid robots, and the Jimu series of branded robots, successfully entering some Apple Stores worldwide. In 2017, UBTECH established a joint laboratory for intelligent service robots with Tsinghua University and an artificial intelligence research institute with the University of Sydney. It has made deep investments in areas such as humanoid robot drive servo systems, gait motion control algorithms, machine vision , speech/semantic understanding, emotion recognition, and U-SLAM (Ultra-Simultaneous Localization and Mapping), actively building a "hardware + software + service + content" robot ecosystem, striving to realize the dream of bringing robots into every household.
Alpha series robots
This robot is primarily intended for the home, school, and entertainment markets, offering flexible and versatile operation. It features PC-based 3D visualization motion editing software supporting human-computer interaction; it can also connect via high-speed Bluetooth to a mobile app for voice command operation, compatible with both iOS and Android. The robot's powerful core power source, a servo motor, provides high control precision and allows for flexible and adaptable movement, earning it multiple invention patents. Its low-power design results in short charging times and long battery life.
Figure 1. Humanoid robots Alpha 1, Alpha 2, and Lynx
Key technologies involved in the Alpha series robots:
(1) Digital servo control technology
The design of high-torque digital servos using coreless motors achieves a balance between performance and cost. As a core component of robots, this effectively controls the overall cost of the robot, enabling true commercialization. This includes the design of high-performance motors using rare-earth neodymium iron boron permanent magnets, achieving high torque density, high power density, low weight, and small size. Finite element simulation is used to optimize the stator and rotor structure, reducing back EMF harmonics, achieving sinusoidal waveforms, and improving motor efficiency and stability. Based on analytical and finite element methods, a cogging-free structure design is adopted, with windings attached to the smooth iron core surface, fundamentally eliminating cogging torque. The motor frequency conversion control algorithm includes general sensor equipment information acquisition and algorithm-based sensor sampling information processing. A corresponding drive control system is designed based on a vectorized motor control algorithm to control the motor speed, torque, and servo position.
(2) Multi-joint linkage and balance technology
Humanoid robots integrate multiple disciplines such as mechanics, electronics, materials, computers, sensors, and control technology. Multi-joint linkage and balance technology are key to the robot's movements, including dancing, walking, standing up, performing martial arts, somersaults, and various Olympic competitions. The Alpha robot uses 20 servo motors for motion control.
Figure 2 Schematic diagram of servo motor
(3) Internet-based voice cloud interaction technology
After a user issues a voice command, the robot's voice acquisition device collects the voice signal and preprocesses it, then transmits the information to the cloud (voice cloud server) via the internet. The processing performed on the cloud server includes: Automatic Speech Recognition (ASR), which enables computers to "understand" human speech and "extract" the text information contained within it; and Text-to-Speech (TTS) technology, which involves multiple disciplines such as acoustics, linguistics, digital signal processing, and computer science. It is a cutting-edge technology in the field of Chinese information processing, primarily addressing how to convert text information into audible sound information, that is, "making machines speak like humans."
(4) Intelligent sensing and regulation technology
Alpha Robot's intelligent perception technology includes: system storage space check – when the robot's storage space is insufficient, it announces through its stereo speaker, "Insufficient remaining space, unable to take photos, please clear my memory first"; face detection – if the robot is within its current field of view but does not detect a face, the system will control the robot's head to rotate to find the user to be photographed. If no face is detected within the head's rotation range, the robot will prompt, "I can't see you, please stand in front of the robot." When the robot detects that its battery power is low, it will issue a low battery warning and automatically switch from working mode to rest mode to reduce power consumption; when the robot detects that it is not connected to the network or the network connection is interrupted, it will prompt the user that the network is not connected and automatically reconnect, announcing the processing result, etc.
(5) User behavior intelligent learning technology
Alpha Robot can acquire user habit data, learn and analyze based on this data, and thus provide a better user experience. The robot acquires permission data or user habit data by integrating a statistical analysis SDK. Taking Android application statistical analysis as an example, download and unzip the SDK, and import it into the intelligent robot. Then, integrate basic functions: add a scene type setting interface, add permissions, and perform application startup statistics. If the application is allowed to return to the foreground for a certain period of time, it will be considered as two independent startups. For example, the user returns to the home screen or enters another program, and then returns to the previous application after a period of time.
Jimu Robot
Jimu robots are the only Chinese brand robots sold in more than 500 Apple stores worldwide, representing the highest level of educational robots in the industry. This series of robots is a high-tech carrier that combines knowledge from multiple disciplines such as mechanical structure, electronic technology, sensor technology, automatic control technology, and artificial intelligence, and is a representative of robots under the STEAM education concept.
Figure 3 Jimu Robot
The key technologies involved in the Jimu robot are as follows:
(1) Servo control system
Jimu robots integrate the achievements of multiple disciplines. Their core component, the digital servo motor, incorporates an MCU system, a servo control system, a planetary gear reduction system, a sensor feedback system, and a DC drive system. It features an independently improved PID algorithm, resulting in a compact size, high torque, and high control precision. Key performance indicators of the servo motor, such as the torque-to-volume ratio and control precision, surpass those of similar products from Japan, South Korea, Europe, and the United States, while costing less than one-third of the price. Motion execution utilizes the digital servo motor as the output, and the main control box uses an STM32 chip as the main control chip, along with a rich array of sensor modules including infrared, gyroscope, ultrasonic, and touch sensors. The modular sensors communicate seamlessly with the servo motor and main control unit in real time.
(2) Structural lightweighting technology
Jimu robots feature a modular, snap-fit design for their exterior components, allowing for seamless connections between hundreds of parts with high precision. The simple and easy-to-use parts enable users to quickly familiarize themselves with the product. The products utilize wear-resistant, self-lubricating materials and feature high-precision mold making.
The Jimu robot uses a hollowed-out rigid wheel structure to reduce the weight of the rigid wheel, as well as the weight of the crossed roller bearings; the outer part of the rigid wheel and the wave generator hub are made of aluminum alloy to reduce weight.
(3) Visualized readback action programming technology
Jimu robots utilize Bluetooth connectivity via mobile devices and feature easy-to-understand PRP (position, record, play) motion editing functions and logic programming modules. Users can see the corresponding modules displayed in Swift while performing logic programming. This is the first time that visual, replay-based motion programming technology and modular logic programming have been applied to Jimu, significantly lowering the programming barrier. Currently, no literature describes this technology.
(4) Interaction design technology for children
Operational visibility (a. Large card design for the model list and simple swipe gesture interaction increase the appeal of models to children. b. Utilize interactive elements and visual effects such as animations or projections to enhance children's cognitive abilities). Reduce user choices (children, unlike adults, do not think comprehensively about problems, so complex pages should be broken down into different steps, as true/false questions are always easier to score on than multiple-choice questions). Timely animation design and operation guidance (connecting different modules can be difficult, so operation guidance is needed to guide users). The app's "3D Dynamic Building" blueprint function guides enthusiasts to build endless creative possibilities through each decomposed step.
(5) Global community sharing and intelligent cloud technology
Community Module: All users can share their built models on this platform. A points reward system will be implemented for shared models, and these points can be redeemed for physical goods in the store. Users can like or comment on models or posts to earn corresponding points, which can be redeemed. Online-Offline Integration Module: Online registration for offline activities brings together different building block enthusiasts.
Intelligent Cloud: Retrieves the official model list, filtering by country/region and displaying different languages. File synchronization: Synchronizes models, programs, and actions created by the same user across different devices. Parts import: Synchronizes user kit parts data and intelligently determines whether the parts of the user's model are supported.
Cruzr robot
Cruzr robots are primarily used in homes, offices, and exhibition halls, integrating security, entertainment, greeting, and promotional functions. These robots are mobile and equipped with cameras, microphones, ultrasonic sensors, and smoke detectors. They support Wi-Fi/3G wireless communication and can automatically recharge. The robots feature automatic patrolling, human detection, human tracking, facial recognition, remote control, remote video recording, and smoke detection, among other security functions. They also offer interactive capabilities such as touchscreen control, voice control, appliance remote control, audio and video playback, and projector functionality.
Figure 4 Cruzr robot
The key technologies involved in the Cruzr robot are as follows:
(1) Complex robot system architecture design and integration
To achieve a variety of integrated functions such as natural customer service interaction and wayfinding, the Cruzr robot system is complex. In terms of hardware, it consists of numerous different sensors, interaction devices, power supply, motion control, and processing components. In terms of software, it needs to simultaneously collect vast amounts of information from multiple sensors, execute dozens of intelligent algorithms, process various events in real time including commands, interactions, and collisions, and respond to and switch between different tasks. As its functions increase, the robot system becomes large and complex. Developing and upgrading the robot through patching will only make the system increasingly complex and unstable. Cruzr takes a holistic approach, designing a standardized, efficient, and open robot system architecture to solve the problems caused by patching and upgrading the robot.
(2) Core technologies of computer vision
Traditional human detection methods using 2D images do not yet meet the accuracy requirements for application-level applications; traditional intrusion detection methods are also typically only applicable to fixed cameras. The Cruzr robot combines depth cameras, color cameras, and various other sensors, improving and utilizing breakthrough technologies from recent years to achieve low-cost, high-accuracy, and applicable functions such as human detection, intrusion detection, facial recognition, and navigation obstacle avoidance.
(3) Multimodal information fusion and intelligent decision-making
The Cruzr robot integrates various sensors such as cameras, human infrared sensors, ultrasonic arrays, smoke detectors, microphones, and obstacle sensors. The functional modes it needs to achieve are also diverse. How to comprehensively utilize this information to make intelligent decisions is also one of the key issues.
(4) Indoor positioning and navigation technology in complex scenarios
In the research of mobile robot technologies, navigation technology can be considered its core technology and a key technology for achieving true intelligence and fully autonomous movement. In indoor navigation, traditional laser navigation methods are expensive and difficult for ordinary users to accept. The Cruzr robot integrates indoor environmental information, obstacle information, robot body location information, and destination location information obtained from depth cameras, infrared cameras, ultrasonic sensor arrays, obstacle sensors, and encoders to establish a robot platform with learning capabilities, capable of localization, navigation, obstacle avoidance, autonomous movement, and automatic charging.
(5) Naturalized human-computer interaction and control technology
For robots, the ability to perceive their environment and interact with and control humans through sensors such as cameras and microphones, combined with various image and sound recognition technologies, is a prerequisite for intelligent robot actions. This allows for natural interaction and control between robots and both humans and the environment. The Cruzr robot integrates a 3G/Wi-Fi wireless communication module and features a user-friendly interface, enabling owners to remotely monitor their home and control the robot via computers, mobile phones, and other devices. It integrates multiple human-computer interaction modes, including voice interaction, touchscreen interaction, and motion interaction. The robot's ability to organically integrate voice recognition, facial recognition, gesture recognition, and robot control technologies in crowded and noisy public settings allows for customer location tracking and the provision of natural interactive services and actions.
For more information, please follow the Machine Vision channel.
