Author: Zen, PANews
With the emergence of electronic navigation maps, human travel and transportation methods have undergone tremendous changes. Traditional static maps have become a real, dynamic and real-time perceptible digital world.
According to Harley and Woodward, two of the most famous cartographers of our time: "A map is a graphical representation that helps spatially understand things, concepts, conditions, processes or events in the human world." Therefore, high-definition maps (HD Maps) with higher richness, accuracy and freshness will naturally become an indispensable new digital infrastructure for future autonomous driving and smart cities, especially for the highly automated (L4) and fully automated (L5) autonomous driving technologies. It is an indispensable "fuel".
However, high-precision maps have problems such as high costs and slow updates, making them difficult to implement on a large scale. According to the White Paper on High-Precision Maps for Intelligent Connected Vehicles, the mapping efficiency of centimeter-level maps is about 100 kilometers of roads per car per day, and the cost is about 1,000 yuan per kilometer, which is a huge amount of money. On the other hand, in 2023, the total mileage of pure electric new energy vehicles in China alone will exceed 160 billion kilometers. One can't help but imagine that if these drivers can participate in collecting data and images without adding extra energy, and even get paid for it, instead of being taken away by monopoly companies, can the dilemma of high-precision maps be broken? And the decentralized map network Hivemapper is doing just that.
Mapping 28% of the world's roads in two years
Launched in November 2022, the Hivemapper network is an open, global, decentralized mapping network based on the Solana blockchain that uses AI technology to build maps based on the daily traffic of thousands of drivers. In the two years since its launch, Hivemapper has mapped 28% of the world's roads, capturing more than 28 million kilometers of street-level images per month on average, detecting and mapping thousands of map features, and growing 5 times faster than Google Street View. Although Hivemapper has faced hardware shortages in the past year, its map coverage is still expanding much faster than any other crowdsourced mapping project, even faster than projects that only require smartphones.
Currently, three giants including HERE Technologies, one of the top ten map makers in the world, rely on Hivemapper to keep maps updated, and leading companies in the automotive, logistics and other industries have also begun to use it. Hivemapper's outstanding performance has also attracted the attention and praise of many well-known investment research institutions such as A16Z, Binance, Blockworks, Coinbase and Messari. For example, in the (2024 Cryptocurrency Status Report) released by a16z, Hivemapper is used as a typical example of the application of the new force of the DePIN project to the real physical world.
Hivemapper's stable and rapid growth is due to its choice of development direction and its professional team. Project founder and CEO Ariel Seidman started his mapping career at Yahoo Search and Maps more than ten years ago and realized that traditional solutions were too expensive. Since then, Seidman has been committed to building a new global map. His first project, Gigwalk, is a location-based data collection company that collects data through mobile devices and allows people to earn extra income. It has also been shortlisted as one of the top ten mobile innovation companies by the well-known business magazine Fast Company. Evan Moss is the co-founder and CTO of Hivemapper. As a senior developer, his fields include mapping, automation of quantitative human-computer interaction systems, computer vision, data science and machine learning, remote sensing, graphics programming, distributed systems, etc.
In terms of development funds, according to public information, in April 2022, Hivemapper announced the completion of an $18 million Series A financing round, led by Multicoin Capital, with participation from Craft Ventures, Solana Capital, Shine Capital and others. Together with Spark Capital, Founder Collective and Homebrew, which participated in previous rounds of investment, Hivemapper has raised at least $23 million in total. In addition, in July 2023, asset management company VanEck announced a deal with Hivemapper to invest directly in Hivemapper's HONEY token. In an announcement from VanEck, it stated that "in the long run, Hivemapper may be able to take market share away from existing map providers by providing customers with better and cheaper products."
The “trilemma” of crowdsourcing
Since the official launch of Google Street View in 2007, the entire mapping industry has followed suit, using a fleet of vehicles to capture 360-degree panoramic images and allowing users to view these images through maps as if they were actually there. This traditional method is very reliable and time-tested, but it has the fatal flaw of being extremely expensive. Even industry giants like Google can only collect real map data once every one or two years.
Therefore, high-quality crowdsourced data is the ideal solution. Existing crowdsourcing solutions can be roughly divided into three options: smartphones, relying on professional equipment such as dashcams, and relying on sensors directly built into the car. However, just as blockchain has an "impossible triangle" consisting of scalability, decentralization, and security, the rich images, consistent data quality, and scalable coverage required to build high-precision maps also have a "trilemma."
Take smartphones as an example. Their low threshold can bring in a wide range of participating users and save developers from hardware design and manufacturing. However, they lack cartographic-level accuracy, have no stereo camera options, and have poor AI model effects. Although they can detect map data, they cannot accurately locate it. So far, no smartphone-based project has been able to continuously refresh map data on a large scale.
In addition, although the configuration of smart driving cars equipped with ADAS modules highly overlaps with the hardware requirements for map data collection, their map data lacks high-definition images to verify its authenticity and accuracy. Most automakers are reluctant to provide access to vehicle images and limit the bandwidth for uploading map data to the cloud via in-vehicle remote communications.
To solve the "trilemma", Hivemapper Inc., the developer behind the Hivemapper network, launched its innovative brand Bee Maps, which focuses on navigation maps, and designed and launched Bee, the "ultimate map miner", which can also serve as an advanced dashcam to ensure the safety of drivers in daily driving. It is worth mentioning that after the introduction of Bee Maps, the Hivemapper network is now positioned as an open source platform that allows any developer to build on it, and Bee Maps is the first case of using this platform to create a product.
AI+DePIN realizes innovation of crowdsourcing model
To achieve a "real-time map with almost zero latency", map makers must not only be able to efficiently extract edge intelligence, but also have to find a way to quickly deliver this information to map users. Although cloud computing costs continue to fall, it is still costly to continuously collect images for every road in the world. Therefore, in actual operations, cutting-edge mapping companies have begun to use AI to replace repeated image reviews and manual map updates, which not only saves labor costs, but also significantly reduces the amount of images stored on centralized servers and helps protect data and privacy.
The Bee dashcam designed by Bee Maps has better positioning, more consistent data quality and a simpler user experience, making it better suited to running Map AI than most devices on the market. After months of prototype testing, Bee will start mass production in the fourth quarter of 2024, and tens of thousands of devices are currently on pre-order.
Unlike ADAS modules in modern cars, Map AI provides images and standard stereo cameras to achieve more accurate and computationally efficient reconstruction of 3D map features. It can dynamically process the right number of frames to extract useful map data, and all selected frames are privacy blurred before Map AI processing. All detected people and vehicles are irreversibly blurred to protect the privacy of bystanders.
In addition, Map AI uses a custom-optimized YOLOv8 computer vision model to detect relevant map features such as traffic lights and road signs from privacy-treated keyframes. And by comparing the two images taken by Bee's stereo camera, Map AI uses reprojection technology to estimate the depth map of each pixel of the central RGB camera.
After combining the depth map with Bee's high-precision GPS coordinates and the orientation read by the IMU, Bee can estimate the precise latitude, longitude and size of the relevant map features detected by computer vision, and send this precisely located map feature data packet to the Hivemapper network, so that the end customers of the map data can detect changes in near real time. In the final quality assurance process, human review will be involved to ensure that the map features have been correctly classified, located and deduplicated.
In terms of other hardware and software, the combination of stereo cameras and DepthAI software is one of the key differences between Bee and other crowdsourced map data collection platforms. As long as the precise position and orientation of the dashcam is known, the depth camera can be used to triangulate the position of objects in the real world. By combining the left and right images, Bee is able to calculate the estimated difference between the same pixels on each image (disparity map) to estimate the distance of each pixel from the dashcam. In addition to the lens, Bee's mapping performance also comes from a customized version of the Luxonis OAK-SoM computer vision module, which is designed for running low-power, high-performance, AI and depth perception tasks in real time. It is a workhorse that can generate consistent map data 24/7 on the highway.
Hivemapper, HONEY and the 10 billion market behind them
Unlike most projects on the market that focus on niche markets or have unproven business models and value, Hivemapper Network is backed by a market for enterprise maps worth tens of billions of dollars a year and a market for geospatial services worth hundreds of billions of dollars a year. As its products are now used by some of the world's most advanced mapping and technology companies, the market size is growing and it is expected to take a share of this huge market.
HONEY token is the native token of the Hivemapper network, built on the Solana blockchain, and is currently listed on many mainstream exchanges such as Coinbase, Kraken, Gate, BingX, and Solana's leading DEX Raydium. In addition, MEXC just launched the HONEY/USDT trading pair at 3 pm on November 14.
Unlike many project tokens that lack practical scenarios, HONEY tokens can be used to create economic incentives in the network while balancing the needs of two groups in its ecosystem. Specifically, contributors who participate in and help build maps can receive HONEY as a reward; businesses and developers use map data to support their products and services, and whenever they use data from the Hivemapper network, HONEY will be destroyed.
According to the MIP-15 proposal, 75% of the HONEY redeemed for map points each week will be permanently destroyed, and 25% will be recast as consumption rewards, with a maximum weekly consumption reward limit of 500,000 HONEY. Any excess will be permanently destroyed. This means that when there is enough income, the token supply will become deflationary.
“Hivemapper’s HONEY rewards are impressively capital efficient,” a recent report from Blockworks Research also noted. “Given the multi-billion dollar size of the mapping market, the upcoming Bee dashcam, and Hivemapper’s relatively low HONEY emissions, we believe the net on-chain consumption of HONEY tokens is foreseeable.”
