Introduction to decentralized geospatial digital twins
The case for a new way to produce and consume near-real-time 3D models of our world
Long before we started building CIRCUM, our team struggled with traditional geospatial data providers. Although absolute leaders in the field, Google Maps and Open Street Map could not provide the level of detail in 3D we were looking for in our projects.
We started to dream of a way to get detailed 3D geospatial models, from everywhere in the world. We even went on to wish for a near-real-time data source. And of course, we wanted it absolutely uncomplicated and inexpensive to use.
Then, because dreaming only gets you so far, we started building CIRCUM.
The basic idea for CIRCUM is simple: At every given moment in the world, a huge number of captors are scanning their surroundings to provide information to a great diversity of actors and needs. Our approach is to create a protocol able to attract and merge the stream of everybody’s 3D datasets. This protocol, called CIRCUM, would then offer a single data source for 3D geospatial models.
If you’re already seeing many apparent issues with this plan, you’re right. The purpose of this post is to introduce insights into what we propose to resolve these issues.
CIRCUM works at the edge of the state of the art of two worlds: blockchain technology and 3D surface reconstruction. Because of the new possibilities that CIRCUM would open up and the philosophy behind its design, we believe this project could have a major impact on the geospatial digital twin market and beyond.
Merging and meshing algorithms for LiDAR point clouds
When we started to discuss the idea that led to CIRCUM, the first question to arise was the kind of data we should focus on. There are a few different remote-sensing technologies but we quickly settled on LiDAR.
Operational since the ’60s, LiDAR technology has recently gained significant interest. It is heavily used by automotive constructors to make more autonomous vehicles, as well as by the construction industry and researchers. As a result, the last few years have seen the emergence of low-cost acquisition systems such as the Ouster LiDAR (10,000€) or the integration of a LiDAR module into the iPhone 14.
However LiDAR datasets aren’t optimized for modelization and need to be processed to be used in 3D engines and to support texture. There is also a huge heterogeneity issue that makes the merging of datasets highly challenging.
To tackle this issue, we formed a partnership with the LaSTIG, the french research laboratory on geographical information science supervised by Gustave Eiffel University, the EIVP, and the IGN-ENSG. This ensemble is both an operator of LiDAR collection programs and an international leader in 3D surface reconstruction research. LaSTIG’s research is open source, as Circum technology will be and our interests overlap in many ways. This partnership is reinforced by the personal involvement of some LaSTIG researchers in the project CIRCUM
In the future, research around 3D surface reconstruction algorithms should become a central part of CIRCUM’s activities and even more so of the company founded to carry out this project, EXTRA. Indeed, EXTRA’s business model is based on providing algorithms to the protocol and other pay-as-you-go services for geospatial digital twin projects.
The first iteration of the CIRCUM protocol then consists of fusion and meshing algorithms that process LiDAR data through decentralized computing. In simple terms, it takes LiDAR point clouds and gives 3D meshed models.
The first questions are out of the way, but that’s only the beginning of our problems.
Solving the fragmentation of the datasets through decentralized economic incentive
Now, the approach of merging LiDAR data collected all around the world has an obvious issue. These datasets aren’t openly available. Even though open data politics have progressed in recent years, most LiDAR collectors still keep their databases for their own usage or sell them for a pricey fee. For an app looking to consume 3D maps, paying for this data may not be the issue, but finding hundreds of data collectors, getting in touch with them, and sorting out commercial details absolutely is.
This fragmentation of data providers is probably the most crucial problem CIRCUM needs to solve. A lot of iterations have led us to a design using blockchain to create a relationship between the respective masses of data providers and data consumers. In this case, blockchain allows us to manage the creation and circulation of value between both sides of the market by representing this value as a token. The token is earned by data providers for their work and bought by data customers to grant them access to the data created. By playing the role of interface between the two, the token brings scalability to the system. It also creates a great network effect since a token earned by a provider for data produced in Nevada can be used by a consumer to access a model in Spain. Thanks to this, any new participant in the protocol creates value for the entire network. Each new datasets increase global coverage and accuracy, attracting new consumers and usages, which in turn attract new data providers. This network effect will allow the protocol to invest in its own future by incentivizing the development of geospatial data outside of western urban areas. Indeed, rural and developing areas are often left out of LiDAR data collection programs due to the cost of collection. However, it is known that geospatial digital twins are instrumental in development and land use strategies. We refer to this aspect of the protocol as the “principle of constant expansion” — see at the end of the article for the full list of the protocol’s governing principles. CIRCUM is designed to incentivize its contributors to keep expanding the extent of territory covered by its data, the accuracy of these data, and their recency.
We’ve discussed the reward for data contributors but CIRCUM actually mobilizes other resource providers that make sure that the protocol can work on its mission: algorithms, computing power, storage, and human arbitration are provided by contributors as well and rewarded with tokens. As such, we can summarize the mechanism of the protocol by saying that it borrows resources, produces a new source of value and sells it, then redistributes the income to those who provided resources in the first place.
The question of how many tokens are distributed to which contributors and how frequently is highly critical for the protocol’s health. That is going to be one of the main focuses of our blockchain team in the future.
Protecting the protocol against false data
Data collectors can earn money by contributing to the protocol. But how do we ensure that they do not provide false data?
This consensus was a central issue during the design phase of the protocol. At the end of this phase, one of our beliefs is that we can never know for sure if a data is true. Due to the remote nature of our geospatial data, verification of the data itself is not feasible on a large scale. Since we cannot trust the data or the permissionless contributors, we choose to build objective incentives for contributors to behave as expected. More precisely, we make it retroactively costly to feed false data. We achieve this by combining two mechanisms known as “proof-of-stake” and “proof-of-reputation”.
Proof-of-stake forces providers to take a stake in the protocol — i.e. buy tokens and lock them into the platform — before being able to contribute data. We mentioned earlier that we can’t know for sure if the data is true, but within a few days or weeks, the protocol will be able to deduce with a certain certitude if a particular dataset is false. It will be able to cross-reference the datasets with the information provided by other actors, identify certain patterns and behaviors, and even open up a bounty for local arbitrators to settle the issue. These arbitrators would have to prove their position, and therefore their legitimacy to chip in, using proof-of-location such as the one developed by the FOAM’s team.
Every contributor with a stake in the protocol also gains a reputation score. Providing datasets that the protocol deems voluntarily unfaithful or too poor in quality will lower the contributor’s reputation score. Not only will he stop receiving tokens for the corrupted data, but if his reputation score falls below a certain threshold, the tokens that the provider has staked will be destroyed. He will also be forbidden to provide new data until he renews his proof-of-stake by buying and locking new tokens.
We see this consensus based on stake and reputation as the first and most general layer of protection. It will be a delicate balance but one we are really excited to work on. Once this first finalized version of CIRCUM will be complete, we plan on releasing it on open source under the management of a foundation. Community and foundation efforts would then contribute to the development of new layers of protection. For example, the protocol could become able to cross-reference with off-chain data to gain insight or to detect datasets produced using stable diffusion.
Data consumers profile
Who needs such a protocol anyway? Quite a lot of people actually. We classify them into 3 categories: First, those looking to make mobility more autonomous. We have already mentioned automotive constructors on this subject, but it concerns even more drones and other unmanned aerial systems. Accessing an up-to-date, detailed, and pre-processed data source for collision avoidance is an important topic for the development of drone use in our cities. Indeed, such a data source would allow the optimization of the use of sensors and the computing capacity of drones as well as their payload. A bunch of research teams are working on this subject, see this article published by Jeremy Castagno and Ella Atkins from the Department of Aerospace Engineering of the University of Michigan.
Then, the second type of use case for CIRCUM is the one that led us to it in the first place: to build better user experiences. A lot of applications using maps could upgrade to 3D, provided that it would be cheap and efficient enough. Just to think about the troubles Zenly went to to get a 3D map on its application. Gaming is a big deal as well, with games like Pokemon Go or Minecraft having led the way.
Finally, we regroup a lot of use cases under the idea of improving connectivity between humans and their environment. This applies to the problems of smart cities, from neighborhood legibility to mobility, garbage removal, 15-minute city, and event planning. The construction industry is also a big consumer of geospatial digital twins and a protocol like CIRCUM could give a powerful boost to the BIM tools. Another unexpected use case is the deployment of 5G network. This very promising technology is restricted by a small broadcasting range and a bandwidth that can be blocked by large structures or even trees. This makes the selection of sites for the construction of the tower stations particularly delicate and costly. These sites then require constant monitoring. These are problems for which CIRCUM could provide valuable services.
CIRCUM’s development principles
Decentralized: None of the resources, data, or architecture elements of the protocol will be controlled by a single actor.
Permissionless: Any actor with the necessary resources will be allowed to provide them to the protocol and any actor wishing to consume the protocol’s data will be allowed to do so.
Unstoppable: The protocol will be designed to gather the resources necessary to achieve its mission for as long as the blockchain on which it is based operates.
Constantly expanding: By design, CIRCUM encourages its stakeholders to continually expand its scope, whether by adding new locations where no data was previously available, adding new levels of detail, or updating existing information.
Credibly neutral: Trying to achieve a credible status of neutrality with respect to the data that will be entrusted to the protocol and the uses of that data. Apply, for example, to the level of detail, which will not be constrained, or to the equality between data providers regardless of the position of the data they provide.
Composability: Will allow anyone to use the protocol as a brick of a new project.
Open source: The code source of the protocol will be free to access in order to allow everyone to improve its functioning or to apply it to another use case.
Non-extractive: The protocol will run at cost and neither the Extra company nor any other actor will charge a commission on its operations.
We welcome feedback from the community and are open to collaborations, so please get in touch.
