Niantic builds centimeter-level accuracy positioning system from Pokemon Go player data

While players are busy catching Pokemons, millions of fresh scans each week improve Niantic’s Visual Positioning System (VPS). The company has accurately scanned 10 million locations worldwide, and VPS has already provided centimeter-level accuracy in over a million locations.

Unlike current local 3D models, Niantic’s system is based on AI models.

“We have trained more than 50 million neural networks, with more than 150 trillion parameters, enabling operation in over a million locations,” Niantic said in a statement.

That means that a single location is covered by an average of 50 neural networks, each with around three million parameters.

The goal is to connect all the models into a single global large model, which Niantic calls the Large Geospatial Model (LGM). This model is capable of ‘spatial understanding’ of geographic locations, connecting scenes to millions of other scenes globally.

“With VPS, users can position themselves in the world with centimeter-level accuracy. That means they can see digital content placed against the physical environment precisely and realistically,” Niantic explains.

Neural geospatial models no longer represent locations using classical 3D data structures. Instead, they encode data in the learnable parameters, similar to how large language models store information.

“These networks can swiftly compress thousands of mapping images into a lean, neural representation. Given a new query image, they offer precise positioning for that location,” the company said.

A new way of determining location

The AI-powered spatial understanding helps predict unseen angles of structures and determine a location from derived predictions.

It’s fairly easy for humans to imagine how a church, a statue, or any other object might look from another angle, but not so for machines. Niantic believes that their spatial intelligence changes that.

For example, traditional 3D models may not be able to determine the exact position using a single picture of a church. With AI, even when having sparse local data, the model can extrapolate beyond it.

“On a global scale, we have seen a lot of churches, thousands of them, all captured by their respective local models at other places worldwide. No church is the same, but many share common characteristics. An LGM is a way to access that distributed knowledge,” Niantic said.

“An LGM would be able to make an intelligent guess about what the back of the building looks like, based on thousands of churches it has seen before.”

While traditional 3D models require a lot of data points, this nontraditional approach allows Niantic to create a map using “about 20 times less space.” For a given location in the world, the company has around 100 user-provided scans, uploaded by users when playing games like Ingress and Pokemon Go.

Applications beyond positioning

Niantic believes their new AI geospatial models are useful for more ‘than mere positioning.’ Interconnection with traditional generative AI models “will make sense of the world in ways that no single model can achieve on its own.”

Intelligent systems will be capable of perceiving, understanding, and acting upon the physical world, and the physical and digital realities will blend.

With VPS, users can see digital content placed precisely and realistically against the physical environment. The digital content can stay in the location and be shared with others.

“For example, we recently started rolling out an experimental feature in Pokémon GO, called Pokémon Playgrounds, where the user can place Pokémon at a specific location, and they will remain there for others to see and interact with.”

Beyond gaming, the company hopes that AI geospatial models will have widespread applications in areas such as spatial planning and design, logistics, audience engagement, and remote collaboration.

The models may also form a critical component in AR glasses, robotics, content creation, and autonomous systems.

“As we move from phones to wearable technology linked to the real world, spatial intelligence will become the world’s future operating system,” Niantic said.