With the development of artificial intelligence and the widespread application of mobile robots in various industries, the market demand for SLAM technology is growing stronger.
So what exactly is SLAM technology? In which fields is SLAM technology mainly applied? These increasingly concerning questions will be answered one by one by the WanKe Intelligent Driving Industry Research Group today.
SLAM: Simultaneous Localization and Mapping
(1) What is SLAM technology?
Simultaneous Localization and Mapping (SLAM) is a concept where a robot starts from an unknown location in an unknown environment and, during its movement, locates its position and orientation by repeatedly observing map features (such as corners, pillars, etc.) and incrementally builds a map based on its position, thus achieving the goal of simultaneous localization and mapping.
(2) In which fields is SLAM technology mainly applied?
1. Robot Localization and Navigation: Map Modeling. SLAM can assist robots in tasks such as path planning, autonomous exploration, and navigation. Domestic companies like Ecovacs, Tami, and the newly launched Lanbao vacuum robot can efficiently create indoor maps and intelligently analyze and plan cleaning environments by combining SLAM algorithms with LiDAR or cameras, successfully entering the realm of intelligent navigation.
Domestic company SLAMTEC is a major provider of this technology, with its name derived from the phonetic sound of SLAM. Its main business is to research solutions for autonomous localization and navigation of service robots. Currently, SLAMTEC has reduced the price of key 2D LiDAR components to around 100 yuan, which undoubtedly further expands the application prospects of SLAM technology.
2. VR/AR: Enhancing Visual Effects. SLAM technology can construct more realistic maps for visual effects, rendering the overlay of virtual objects based on the current perspective, making it more realistic and seamless. Representative products in VR/AR such as Microsoft’s Hololens, Google’s Project Tango, and MagicLeap have all applied SLAM as a means of visual enhancement.
3. Drone Field: Map Modeling. SLAM can quickly construct local 3D maps and, when combined with Geographic Information Systems (GIS) and visual object recognition technology, can assist drones in identifying obstacles and automatically planning avoidance paths. The Hovercamera drone, which once went viral in the US, has applied SLAM technology.
4. Autonomous Driving: Visual Odometry. SLAM technology can provide visual odometry functions and integrate with GPS and other positioning methods to meet the precise positioning needs of autonomous driving. For example, Google’s self-driving car based on LiDAR technology and the Wildcat autonomous vehicle modified by Oxford University’s Mobile Robotics Group have both successfully completed road tests.
(3) Detailed Explanation of SLAM Algorithms:
Summary and Introduction of Companies Involved in SLAM Localization Algorithms
Autonomous/Assisted Driving
1
Baidu
Main Products: Autonomous Driving Software
Baidu’s intelligent vehicles are paving the way for the future. Based on SD maps, ADAS maps, high-precision maps, artificial intelligence, and big data, it provides autonomous driving system solutions and HMI human-machine interaction platforms to domestic and foreign car manufacturers; collaborating with car manufacturers, Tier 1 suppliers, chip manufacturers, and service providers to create a new ecosystem for smart cars. The Apollo platform is an open, complete, and secure platform that helps partners in the automotive industry and autonomous driving field quickly build their own autonomous driving systems.
Apollo is an open, complete, and secure platform that helps partners in the automotive industry and autonomous driving field quickly build their own autonomous driving systems. Road Hackers is Baidu’s open platform for autonomous driving, providing access to a wealth of driving data in China and demonstrations of deep learning-based autonomous driving algorithms.
2
Horizon Robotics
Main Products: Intelligent Driving Solutions
Based on self-developed computing platforms for intelligent driving solutions. Currently, Horizon has formed a product matrix for intelligent driving consisting of the Matrix autonomous driving computing platform, ADAS, DMS, AR HUD, Face ID, etc., providing solutions for different levels of autonomous driving such as L2, L3, and L4. Horizon is also the only AI technology startup in China that has established partnerships with top automotive Tier 1s and OEMs in all four major automotive markets globally.
Multi-level, Horizon deeply couples core hardware such as the Matrix autonomous driving computing platform with the Horizon intelligent driving software platform, providing multi-level visual environment perception solutions for intelligent driving. High performance, supporting fine-grained, structured semantic perception of complex scenes, highly scalable, and modular 3D semantic environment reconstruction.
Low latency, based on edge AI computing, enabling efficient computation and data processing on the local end. High precision based on deep learning for environmental perception, target detection, recognition, and motion prediction, using structured deep neural networks for efficient and comprehensive computation.
3
TuSimple
Main Products: Autonomous Trucks
Effective perception distance of up to 1000 meters, ensuring that trucks have a 35-second reaction time on highways, helping the system perform safer and more efficient driving tasks. The perception system can correctly identify under various weather conditions, enabling the commercialization of autonomous driving transport services, allowing TuSimple to continuously generate revenue through L4-level (SAE) autonomous driving solutions.
In the US, TuSimple’s autonomous trucks complete 3-5 cargo transports daily, serving thirteen end customers, and will open a new transport route from Arizona to Texas, expanding its fleet of autonomous trucks to 50 this year.
4
Tencent
Main Products: Intelligent Driving Solutions
Tencent’s autonomous driving business center leverages its accumulated capabilities in high-precision mapping, integrating the company’s advantages in the Internet of Vehicles, big data, cloud platforms, deep learning, multi-source information fusion perception and positioning, and high-precision map production and application.
Focusing on “one core, two products, and three platforms,” Tencent builds a leading development platform that integrates autonomous driving product R&D, testing, and evaluation verification, providing complete autonomous driving technology solutions and open platform support capabilities, acting as a “software and service provider” to offer inclusive, innovative, and sustainable smart solutions through “digital interfaces and toolboxes” for the industry.
5
Four-dimensional Map Technology
Main Products: Autonomous Driving System Solutions
Autonomous driving maps are an indispensable part of autonomous driving, playing a crucial role in perception, positioning, and planning decisions.
High-precision positioning based on real-time dynamic differential RTK technology, achieving centimeter-level precision for autonomous vehicles.
Smart Horizon. Maps enable vehicle control, comprehensively enhancing ADAS functions and capabilities through software and hardware solutions.
Smart Perception. Smart Collector, a high-precision, low-cost, self-algorithmic intelligent perception system.
6
Yushijia Technology
Main Products: U-Drive™ Intelligent Driving System
Providing users with a comprehensive, upgradeable intelligent driving system. AI algorithms based on leading visual capabilities for multi-perception recognition, centimeter-level fusion positioning, and planning control.
Automotive-grade hardware, integrating AI algorithms into intelligent driving controllers, compatible with various sensors such as cameras, LiDAR, and millimeter-wave radar. Cloud-based intelligent services, providing remote vehicle monitoring, driving, and fleet scheduling, vehicle-road collaboration, and autonomous driving simulation services based on secure cloud data communication technology.
7
Zongmu Technology
Main Products: ADAS Advanced Driver Assistance Systems
ADAS advanced driver assistance systems perceive the surrounding environment and vehicles through sensors and cameras, providing reliable alerts and prompts to drivers, making driving safer.
During product design, a professional vehicle product design process was adopted, ensuring high shock resistance, low interference, low power consumption, high-temperature resistance, and impact resistance. Custom fisheye lenses cover the vehicle’s surroundings with 360-degree blind spots. The compact camera is designed for vehicle applications, facilitating installation and perfect matching with vehicle models.
8
Didi
Main Products: Intelligent Driving
Intelligent driving will significantly enhance safety and travel efficiency, providing a strong supplement to existing driver services. Collaborative traffic between vehicles and roads will change the way vehicles travel.
Didi is actively laying out high-level intelligent driving and V2X technology research and development. The Didi platform generates a large amount of data daily, providing a solid foundation for training intelligent driving algorithms. Currently, Didi has R&D centers and testing fleets in both China and the US and will continue to increase investment in intelligent driving.
9
Momenta
Main Products: Environmental Perception
High-precision maps and driving decisions.
Environmental perception: Achieving high-performance recognition of multiple lanes, traffic signs, signals, and drivable areas under conditions of darkness, backlighting, adverse weather, and lack of clear lane markings.
Pedestrians: By detecting pedestrians and recognizing human feature points, it can understand pedestrian postures and behavioral intentions while accurately estimating the distance between pedestrians and vehicles.
Vehicles: Every vehicle on the road can be restored to its 3D bounding box, robustly detecting the direction of the vehicle, accurately estimating distance, and performing real-time positioning on high-precision maps.
Data-driven driving decisions: Similar to establishing an intelligent driver with 100 billion kilometers of driving experience. Crowdsourced road tests allow us to obtain massive driving trajectories in high-precision semantic maps. By learning from driving trajectories, algorithms can make driving decision plans based on current environmental perception and high-precision map information. Thus, data rather than rules can be used to solve the algorithm’s failure in rare road conditions.
10
New Mind Technology
Main Products: Autonomous Driving Solutions
Using multi-sensor fusion (cameras, LiDAR, millimeter-wave radar, ultrasonic radar, and GPS/IMU) to create a complete autonomous driving solution, featuring fast response, high customization, and strong scalability, capable of achieving L3/L4 level autonomous driving.
Based on deep learning, the vision-based perception system applies cutting-edge image signal processing (ISP) technology and computer vision technology for target recognition and detection, utilizing highly optimized deep neural networks to achieve precise environmental perception under all-weather, multi-road conditions, and various complex weather.
11
Beefly Technology
Main Products: Autonomous Freight Trucks / Automotive AI Chips / ADAS Advanced Assistance Driving
Autonomous freight trucks. Beefly provides comprehensive solutions for L2-L4 autonomous freight trucks, applying AI technology to make freight driving safer, more convenient, and more efficient.
Automotive AI chips. Based on its own AI technology, Beefly has developed a fully independent intellectual property Chinese “chip” aimed at providing “hard” support for rapidly entering the AI era with high performance, low power consumption, and low cost.
ADAS advanced assistance driving. Lane departure warning, forward collision warning, fatigue driving warning, Beefly’s ADAS is installed in the driver’s cabin, using front-facing and cabin cameras to recognize the road conditions ahead and driving behavior, assisting the driver with forward collision and lane departure warnings; it can also monitor the driver’s fatigue level and driving norms.
12
Jimu Intelligent (Wuhan)
Main Products: ADAS Driving Assistance Solutions
As a professional ADAS driving assistance solution provider, Jimu Intelligent has independently developed professional computer vision technology that can accurately recognize vehicles, lane lines, pedestrians, traffic signs, and other road information, helping drivers perceive the surrounding driving environment, alerting potential collision dangers, and reducing collision accidents, while also providing core technology and services for the popularization of autonomous driving.
Warehouse Robots
1
JD.com
Main Products: Unmanned Vehicles
Unmanned vehicles are focused on creating intelligent transport equipment within JD.com’s intelligent logistics system, based on core autonomous driving technology, developing and producing various series and models of autonomous unmanned vehicle products according to user needs in different scenarios.
For logistics transportation and distribution scenarios, generating autonomous freight trucks and delivery robots; for warehouses, factories, parks, and communities, generating security inspection robots; for office building scenarios, generating service robots. Relying on commercial market demand and user needs, it achieves commercial operation services through efficient, low-cost, and enhanced user experience operational models.
2
SF Technology
Main Products: Indoor AGV
Automation and robotics have realized the application of various technologies in logistics scenarios, greatly reducing labor costs and serving as an important upgrade for traditional logistics to intelligent logistics.
3
Alibaba Damo Academy
Main Products: Intelligent Logistics Vehicles
High-precision positioning: Exploring high-precision intelligent positioning technology based on high-precision maps, constructing accurate mappings of the physical and digital worlds, providing precise traffic elements, POI, and other scene information for autonomous vehicles.
Researching cloud-based multi-source fusion positioning technology, building a cloud platform for integration and processing, connecting vehicles with people and vehicles with vehicles, forming a dynamic positioning network. This network provides positioning services for autonomous driving while supporting precise data collection and feedback based on accurate positioning, enabling online updates of high-precision maps, and comprehensively enhancing the accuracy, reliability, and intelligence of positioning systems.
Decision Planning: Decision planning is divided into driving decision systems and motion planning systems. The driving decision system analyzes a large amount of measured high-precision traffic flow data to explore the essential mechanisms of human driving behavior, excavating optimal behavior models and parameters to ensure the optimization of online driving behavior. The motion planning system searches thousands of possible operating trajectories under limited computing resources and selects the optimal trajectory based on surrounding environmental information to achieve smooth autonomous driving.
Intelligent Control: Mainly includes lateral control and longitudinal control. Lateral control mainly studies the path tracking ability of autonomous vehicles, i.e., how to control the vehicle to travel along the planned path while ensuring the safety and smoothness of the vehicle’s travel. Longitudinal control mainly controls the vehicle’s speed tracking ability, controlling the vehicle to cruise at a predetermined speed.
Autonomous Driving Simulation Platform: Composed of traffic intelligent agent systems, scene editing generation systems, virtual world rendering systems, and large-scale server deployment systems. The traffic intelligent agent system extracts realistic behavior models from a large amount of traffic data, simulating the behavior of vehicles, pedestrians, and other objects;
The scene editing generation system supports the editing of various extreme scenarios and the generation of similar scenarios, achieving pressure testing of autonomous driving algorithms; the virtual world rendering system visualizes the simulated scenes, enhancing the intuitive understanding of problem analysis;
The large-scale server deployment system improves the scale and efficiency of simulations, achieving hundreds of millions of kilometers of simulation testing annually; the simulation platform saves a lot of time and costs for autonomous driving testing, greatly reducing testing risks, randomly simulating various extremely complex scenarios, achieving comprehensive testing coverage.
Data Platform: Production data platform, data production process, and monitoring. To collect rich data for perception to improve the robustness of the entire autonomous vehicle. Researching how to establish efficient labeling systems to reduce labeling costs, improve efficiency, and increase the accuracy of labeling results;
Researching how to more effectively monitor the operational status and issues of vehicles, and how to visualize data more reasonably; exploring the process of combining real data and simulation data to provide structured, semi-structured, and unstructured data integration, collection, and indexing.
Vehicle-Road Collaboration: Through the collaborative cooperation between vehicles and roads, constructing a new model of “vehicle-road-cloud” integration, researching and developing roadside multi-sensor perception, multi-modal data fusion and processing, low-power edge computing, and medium-short distance direct communication technologies, creating new roadside intelligent devices to provide real-time data services for traffic tools and comprehensive traffic information foundational data for the cloud.
4
Beijing Xizhi Technology
Main Products: Intelligent Logistics Automation Solutions
Applicable Scope: The Xizhi picking system is suitable for warehouse picking operations of small to medium-sized, multi-category items, with warehouse areas reaching tens of thousands of square meters, targeting various industries such as e-commerce, retail, 3PL, pharmaceuticals, footwear and apparel, food, daily necessities, industrial, and automotive manufacturing.
Functional Features: Stable lifting and handling, with a maximum load of 1,000 kg. Powerful navigation and obstacle avoidance, visual combination navigation methods, and millimeter-level positioning. Flexible intelligent scheduling, intelligent path planning, and congestion control algorithms. Advanced automatic charging, intelligent autonomous charging, and 24-hour uninterrupted operation.
5
Megvii Technology
Main Products: Intelligent Warehouse Logistics Solutions
The intelligent warehouse logistics solutions from Megvii Technology, provided by its wholly-owned subsidiary Ares Robotics, can offer comprehensive solutions for e-commerce, supermarkets, and large warehouse logistics industry users, including “goods-to-person,” “orders-to-person,” “vehicles-to-person,” and “robots + warehouse systems + data analysis,” continuously helping the supply chain enhance overall efficiency and create more value for industry clients.
6
Hikvision
Main Products: Intelligent Warehouse Solutions
To address the high labor costs, low management efficiency, and high order error rates in traditional warehouse operations, Hikvision has independently developed an intelligent warehouse robot system, providing high-efficiency intelligent warehouse solutions.
The solution involves three core modules: Hikvision intelligent warehouse robots, robot scheduling systems (RCS), and intelligent warehouse management systems (iWMS), with the core concept of replacing humans with machines, allowing workers to perform necessary operations at the workstation terminal to meet various warehouse business needs, significantly improving the efficiency and management level of warehouse operations in e-commerce sorting centers, 3C manufacturing, traditional manufacturing, medical, food, and automotive manufacturing industries.
Supporting business functions such as hot and cold analysis, shelf position adjustment, shelf organization, and intelligent inventory, the operation terminal interface supports visual management.
Service Robots
1
Pudu Technology Co., Ltd.
Main Products: Catering Robots
Starting work easily. As a new employee in the restaurant, “onboarding” is no problem for the Happy Delivery robot. It does not require ground tracks for assistance and does not need training; it can complete map construction and officially start working within as little as 2 hours. It has no work injuries or sick leave, and features like remote fault repair and software upgrades make it easy for restaurant operators.
2
Gaoxian Robotics
Main Products: Building Delivery Robots
Autonomous positioning and navigation module.
The Gaoxian Conbox autonomous positioning and navigation module series products possess top-notch environmental perception, high robustness, precise positioning, open platforms, and comprehensive documentation tools, allowing for seamless and rapid integration into various downstream terminals, widely applied in logistics, security, banking, supermarkets, government affairs, retail, and other fields, holding over half of the domestic market share, playing a key role in achieving typical scene applications for robots and building industrial ecosystems.
3
UBTECH Robotics
Main Products: Humanoid Robots
UBTECH’s positioning and navigation system (U-SLAM) can load different positioning solutions based on different scenarios, allowing the application subject to establish surrounding environment maps and perform self-positioning, intelligently selecting the optimal path to the target point for navigation behavior control based on environmental changes.
4
Furuiwei Intelligent Robots
Main Products: Home Service Robots
Integrating global R&D resources, continuously promoting the integration and innovation of robotics technology with artificial intelligence, mobile internet technology, cloud computing technology, and IoT technology, leading the construction of smart homes, smart communities, and smart cities, providing intelligent solutions and advanced lifestyles for people.
Not just in form, but deeply considering human-machine interaction; not just in appearance, but fully considering interaction in lying, sitting, and standing positions; combining expressions, sounds, and actions to make human-robot interaction more natural.
5
Xiaomi
Main Products: Vacuum Robots
The new generation of Roborock vacuum robots is equipped with 13 types of sensors, including Roborock’s self-developed high-precision LDS laser ranging sensor, capable of scanning rooms at a speed of 5360 times per second. Obstacles on the floor will form light spots in the sensor, which will then accurately map the room using its unique SLAM algorithm and reasonably plan the cleaning route.
Above the LDS laser ranging sensor, the Roborock vacuum robot is also equipped with omnidirectional pressure sensors, which sense slight collisions from 360° at the top, preventing it from getting stuck under low sofas or cabinets, and also protecting the laser sensor. Wall sensors allow the vacuum robot to clean wall edges at about 10mm distance, and infrared recharging technology allows it to continue cleaning after recharging.
6
Ecovacs Robotics
Main Products: Vacuum Robots
Ecovacs Robotics, as one of the earliest developers and manufacturers of service robots globally, focuses on the independent R&D, design, manufacturing, and sales of service robots.
In the more than ten years since its establishment, it has successfully launched a complete line of household robot products, including the Deebot vacuum robot, the window-cleaning robot, the air-purifying robot, and the housekeeper robot UNIBOT, as well as a series of commercial robot products led by the public service robot Wangbao. Ecovacs Robotics’ more than ten years of cultivation in the service robot market has also won unanimous recognition from users.
7
Ping An Technology
Main Products: RPA An Xiaofeng
RPA An Xiaofeng is a process automation robot independently developed by Ping An Technology, providing the lowest cost and fastest solutions to replace manual operations by simulating human actions.
With Ping An Cloud RPA An Xiaofeng, enterprises can utilize virtual employees to efficiently handle intensive repetitive tasks on user interaction interfaces of different business systems, freeing up human resources and effectively improving enterprise efficiency.
The execution robot offers various usage methods, enabling full-process unattended operation, fully realizing the value of automation. It can provide manual operation, timed operation, remote control, and intelligent scheduling.
8
Zhen Di Robotics
Main Products: Underwater Robots
The PowerDolphin little dolphin is equipped with a unique dual-joint rotating camera, with a wide-angle lens reaching 132°, and can rotate to meet the shooting needs from different angles above and below water, with a rotation range of -150° to +70° (defining the horizontal position as 0°, rotating towards underwater as negative), providing users with a new shooting perspective of 220°.
9
Orion Star
Main Products: Service Robots
Orion Star is born for “truly useful” robots, a company with full-chain AI technology in the service robot industry, with an internet gene for robotics. The company has developed a full-chain AI technology for robots, including chips + algorithms (brain), full perception visual recognition (eyes), voice full-chain technology (mouth), microphone arrays (ears), seven-axis consumer-grade robotic arms (hands), and indoor autonomous navigation platforms (legs).
After more than two years of rapid development, Orion Star has released its self-developed robot platform OrionStar OS and a series of self-developed products, including the Orion Star AI intelligent service robot “Little Secret,” the Little Leopard AI tutoring robot, the intelligent new retail robot “Little Vendor,” and the robotic arm coffee kiosk “Little Coffee.”
10
Zhuimi Technology
Main Products: Vacuum Cleaners
Zhuimi Technology has a series of new invention technologies in high-speed digital motors, monocular machine vision, SLAM (Simultaneous Localization and Mapping), and multi-cone cyclone separation technology, leading the world.
These technologies have also been successfully applied to cordless vacuum cleaners, robotic vacuum cleaners, and other products.
Drones
1
DJI
Main Products: Drones with Visual Sensing Navigation Systems
Guidance is a brand new visual sensing navigation system that can perceive nearby obstacles, allowing the aircraft to actively avoid them. It has a powerful processing core, equipped with five sets of visual ultrasonic combination sensors, along with advanced visual algorithms, providing users with a safe flying experience. High-precision visual positioning Guidance adopts high-precision stereo visual algorithms, achieving centimeter-level positioning accuracy near the ground. It can provide positioning information under complex terrain and high-speed flight conditions.
The effective height of the visual positioning system is up to 20 meters, with all-around obstacle perception. Guidance can monitor environmental information in multiple directions in real-time and perceive obstacles, allowing the aircraft to avoid potential collisions during high-speed flight.
2
JD.com
Main Products: Drones
JD.com drones aim to create a three-level drone + aviation logistics system covering rural areas across the country, intending to achieve connectivity in every village and county. This system starts with end layout, gradually establishing trunk and branch logistics networks, ultimately constructing an integrated intelligent logistics network for two-hour logistics life circles, enhancing the shopping experience for consumers.
JD.com drones are building a three-level drone + aviation logistics system, establishing trunk, branch, and end drone logistics distribution, gradually constructing an integrated intelligent logistics network.
3
SF Technology
Main Products: Drones
Based on solving logistics transportation issues in various special scenarios (characteristic economy, medical cold chain, emergency delivery, special logistics, etc.), multiple models and related supporting software and hardware have been successfully developed to meet different operational needs, including multi-rotor drones, vertical take-off and landing fixed-wing drones, operational control systems, communication systems, and drone express delivery cabinets.
On March 27, 2018, it obtained the first domestic drone aviation operation license, dedicated to providing high-quality logistics delivery services for remote areas with inconvenient land and water transportation, offering general or customized drone products and comprehensive solutions for specific industries.
4
EHang
Main Products: Drones
EHang Intelligent is a high-tech innovation enterprise integrating R&D, production, sales, and service of intelligent aircraft, headquartered in Guangzhou, China. It has been selected by the international authoritative media Fast Company as one of the “World’s Most Innovative Companies” and one of the top three global drone companies.
Upholding the mission of “allowing humans to fly freely like birds,” EHang Intelligent provides simple, intelligent, safe, and efficient aircraft products and solutions for customers in various industries, including autonomous flying vehicles, smart city command and dispatch centers, industry application networked drones, automated drone clusters, and drone logistics delivery.
As a pioneer in many innovative technologies and application models in the global civil drone industry, EHang Intelligent continuously explores the boundaries of the sky, making flying accessible to the beautiful life of smart cities.
5
Yiwat Ewatt
Main Products: Drones
Logistics transportation drones provide industry application solutions. Drone delivery can be efficient and fast, especially in underdeveloped traffic areas, where drones can utilize air traffic to shorten delivery distances and improve efficiency. At the same time, drones can serve as aerial platforms, carrying special equipment to complete high-intensity tasks such as inventory checks in logistics warehouses and park inspections.
Augmented Reality (AR)
1
Baidu
Main Products: AR Effects
Baidu’s developed SLAM technology features low power consumption and high real-time performance, supporting various data sources such as monocular, binocular, RGBD, and semantic SLAM. Baidu’s semantic SLAM technology combines SLAM technology, image recognition technology, and object recognition technology to construct accurate semantic maps, empowering augmented reality and achieving an AR interactive experience of “what you see is what you get.”
Helping developers quickly integrate AR SDKs, efficiently produce and distribute AR content. This mainly includes a technology open platform for AR application developers and a content open platform providing content production and distribution services.
2
Huawei
Main Products: HUAWEI VR
HUAWEI VR is a platform for Huawei VR content developers. Developers use the Huawei VR SDK for development, and after completion, upload it to the Huawei VR application store, where consumers with HUAWEI VR2 glasses can directly purchase and download it after passing the review.
Currently, the hardware form of HUAWEI VR is VR2 glasses + Huawei series phones, with specific supported phone models listed in the development guide. The VR2 glasses themselves are 3DoF headsets, paired with a 3DoF functional controller, allowing developers to directly port other platforms’ 3+3 content, and slight modifications can easily port applications without controllers.
3
ArcSoft
Main Products: 720°/3D AR/VR Technology
Acquisition and display of 3D VR content is the main way to reflect immersion in virtual reality technology. Of course, a wide-angle or no blind spot full-angle display effect is better; however, solving dizziness remains the key issue, while high performance is always a tough nut to crack.
ArcSoft utilizes over a decade of research results in core algorithms and engines for computer vision to achieve a robust and fast 3D content acquisition and display engine, with clear rendering effects, globally consistent content, no ghosting, no object and structure misalignment, and smooth and comfortable viewing experience. The current engine can support 720-degree full-angle views while achieving clear depth effect display at ultra-close distances of 10 cm.
4
Lenovo Research Institute
Main Products: AR Solutions
Augmented Reality (AR) is a technology that calculates the position and angle of camera images in real-time and adds corresponding images, videos, and 3D models, overlaying the virtual world on the real world for interaction. It allows users to see real-world information while simultaneously displaying virtual information, with both types of information complementing and overlapping each other.
It mainly includes three major features: virtual-real combination, real-time interaction, and three-dimensional registration. Based on these typical features, Lenovo Morning Star AR is committed to enhancing enterprise productivity, providing end-to-end AR solutions for industry clients.
5
Lianfengtai
Main Products: AR Solutions
Design • The HiAR G100 adopts a simple integrated design, incorporating Eastern aesthetics into technological beauty, achieving a seamless connection of form and spirit. It features a smooth and lightweight design, as well as a natural and breakthrough appearance.
Wearing • The HiAR G100 features an ergonomic design that provides an extremely comfortable and natural experience for every user, as if experiencing “customization.” The detachable and adjustable headband components and innovative elastic clamping support provide a balanced and comfortable wearing experience. Optional myopic lenses support personalized adjustments for interpupillary distance and myopia degree.
Optics • The HiAR G100 employs an optical see-through solution with zero latency and no dizziness, allowing users to navigate seamlessly between the virtual and real worlds. The free-form optical waveguide projection system balances comfortable viewing angles with lightweight design, achieving brightness of up to 1000 nits, presenting stunningly bright images.
Performance • The HiAR G100 is innovatively equipped with Qualcomm® Snapdragon™ 820 processors, featuring 4GB dual-channel LPDDR4 flash memory, significantly improving computation speed, graphics processing speed, and data transmission rates compared to previous models.
Vision • The HiAR G100’s vision capabilities far exceed human eyes, featuring a 13-megapixel camera and a 5P blue glass filter lens with an 82.2-degree wide-angle lens, allowing it to capture high-quality images from distances of 5 meters or even further. Its unique PDAF phase focus function can complete focus in just 0.1 seconds.
Chip Cameras
1
SLAMTEC
Main Products: Modular Autonomous Positioning Navigation Solutions
SLAMWARE is a single modular autonomous positioning navigation system that integrates laser-based simultaneous localization and mapping (SLAM) and accompanying path planning functions.
The host system can utilize the communication interface provided by SLAMWARE and combine the high-precision robot positioning information obtained in real-time with the environmental map data autonomously constructed by SLAMWARE for flexible and diverse functional expansions. Robots can achieve autonomous positioning, automatic mapping, path planning, and automatic obstacle avoidance, effectively solving the problem of autonomous walking for robots.
2
Speeding Technology
Main Products: RS-LiDAR-Algorithms Perception Algorithms
RS-LiDAR-Algorithms is a software development kit specifically developed for autonomous driving environmental perception, including functions such as localization, curb/drivable area detection, lane marking detection, obstacle detection, dynamic object tracking, and obstacle classification and recognition.
We developed this SDK to facilitate autonomous driving project developers for secondary development based on this kit, accelerating your autonomous driving projects.
3
RoboSense
Main Products: Robot SLAM Algorithm Boards
The LS SLAM module is based on LiDAR, integrating autonomous localization and mapping, as well as accompanying path navigation and obstacle avoidance functions, used for robot autonomous navigation and positioning system solution modules. “LiDAR + SLAM” allows service robots to move freely.
4
Little Smart
Main Products: Binocular Camera Solutions
Providing a mature binocular stereo vision system suitable for various robots, helping robots achieve 3D stereo map modeling, planning navigation routes, obstacle avoidance, and other functions. Mature whole machine R&D experience.
Industry-leading binocular hardware solutions. IR active light solves the problem of recognizing white walls and textureless objects. Industry-leading VSLAM / VIO mature algorithms support.
5
Hesai Technology
Main Products: Environmental Perception Sensor Fusion Systems
Hesai Technology focuses on developing laser sensors, with its product line including LiDAR for autonomous driving and robots, as well as laser remote sensing systems for safety inspections in the energy industry.
Hesai Technology’s self-developed Pandar40 LiDAR has been installed in dozens of top autonomous driving companies’ vehicles in Silicon Valley, Detroit, Pittsburgh, and various locations in Europe and China. Among the 62 high-tech companies in California that have obtained public road testing licenses for autonomous vehicles, over 40% are already paying customers of Hesai Pandar40.
6
MoVision Technology
Main Products: Autonomous Driving Chips
MoVision Technology aims to “unlock the AI brain of cars, achieving safer and freer autonomous driving” and is committed to integrating advanced AI technology with successful experiences in automotive electronic systems, bridging the technological gap in traditional automotive intelligence;
Implementing international-level visual perception, fusion, localization, path planning, and control algorithms on embedded low-power automotive-grade chips, achieving autonomous driving under different scene conditions.
7
Speeding Technology
Main Products: 3D Sensor Vision Solutions
The vSLAM (visual simultaneous localization and mapping) algorithm is the core technology of the team. The team’s self-developed vSLAM algorithm can fuse data from various sensors (LiDAR, inertial measurement units, odometers, ultrasonic, etc.) to obtain stable and accurate position and posture information while helping robots and other intelligent devices acquire 3D spatial environmental information, enabling them to have autonomous movement, path planning, and scene understanding capabilities.
After multiple iterations of the Qfeeltech vSLAM algorithm, it maintains leading positioning and posture effects compared to other algorithms in the industry under various international public datasets.
The Auto Navigating (autonomous navigation) algorithm is one of the engineering technologies developed by the team based on specific application scenarios. The team’s self-developed Path Planning algorithm uses visual images and LiDAR data as primary inputs, combining AMCL (particle filter) localization algorithms and A* (global planning), DWA (local planning) path selection strategies to achieve more accurate positioning results than single sensor solutions. In response to special events, this algorithm can achieve faster real-time strategy responses and select more suitable moving paths.
The Object Tracking (object detection and tracking) algorithm is one of the research technologies based on traditional vision and depth vision technologies. The team’s self-developed Visual Tracking algorithm can achieve faster and more accurate dynamic object recognition, detection, and tracking results in traditional RGB video streams using the MLD architecture, which is currently widely applied in fields such as autonomous following of drones.
Additionally, this algorithm can automatically perform foreground extraction and object segmentation based on RGB-D video streams with depth information, resulting in better detection and tracking results.
The Scene Perception (image scene understanding) algorithm is one of the technologies the team is researching in the current robotics industry context. The team aims to develop vertical application image understanding algorithms by leveraging the preliminary technical accumulation of depth vision technology, combined with mature deep learning algorithms based on CNN (convolutional neural networks) and industry data. The innovation of this algorithm lies in utilizing the advantages of depth images over traditional images to achieve more precise image language understanding and cognition through the acquisition of higher-dimensional feature point sets.
8
Chipron Technology
Main Products: Dynamic Vision Sensor Chips
Chipron Technology possesses globally leading dynamic vision sensor chip technology, dedicated to creating efficient image processing systems. Dynamic vision sensors achieve revolutionary changes from the source of data input, enabling pixel-level active perception and transmission, reducing data redundancy at the front end to provide structured information, improving system operation efficiency, and lightening the overall module burden at the back end.
Traditional image sensors are based on frames for imaging, where the photoelectric current detected by the pixel array is integrated in capacitors within a fixed time period (frame time). During each frame time, the voltage level reached by each pixel is transmitted out of the chip in a sequential manner.
Therefore, when using traditional image sensors, all pixel information is transmitted during each frame time, regardless of whether there are changes on individual pixels that would require information transmission.
In contrast, DVS chips consist of independent light-sensitive pixels, each of which independently determines whether it is activated. When a pixel is activated, it simultaneously sends out three sets of information (X, Y, A, T) regarding its position, gray value, and time.
This individual triggering mechanism differs significantly from the traditional full-frame triggering mechanism, giving dynamic vision sensors a clear advantage in real-time analysis scenarios such as SLAM.