The startup Coco Robotics entered into a partnership with Niantic Spatial, a division of the company behind the game Pokémon GO. The collaboration uses a vast bank of images captured by players to train robot navigation systems. The central objective is to achieve significantly greater accuracy than traditional GPS in dense urban environments.
The agreement, announced in March 2026, integrates Niantic’s Visual Positioning System (VPS) with Coco’s fleet of autonomous vehicles. Esses robots, which operate on sidewalks, have completed hundreds of thousands of food and supply deliveries. The technology seeks to overcome common GPS failures caused by interference from skyscrapers in cities.
Jogadores creates three-dimensional visual map with billions of images
Over the course of a decade, Pokémon GO players have pointed their smartphone cameras at various points of interest, gyms and streets while hunting virtual characters. Esse process generated an impressive accumulation of over 30 billion images. The photos included crucial metadata on location, angle, and light conditions.
Niantic Spatial used this material to train its robust spatial intelligence model. The result is a detailed three-dimensional map of the urban world, allowing devices to locate themselves with an accuracy of just a few centimeters. Diferentemente of conventional GPS, which depends on satellite signals, the VPS system compares images captured in real time with its vast accumulated database.
In areas with many skyscrapers, satellite signals can reflect off building facades, generating positioning errors of up to 50 meters. Essa inaccuracy is a constant challenge for autonomous vehicles. The VPS system, on the other hand, avoids this fundamental problem. Ele works effectively precisely where GPS fails most, in environments known as “urban canyons”, ensuring more reliable navigation.
Robôs’s Coco use cameras for centimeter navigation
The Coco Robotics fleet totals around a thousand units, which operate at approximately eight kilometers per hour on sidewalks. The robots are designed to carry up to eight large pizzas or four bags of groceries, making deliveries in several cities. Eles have already completed more than half a million deliveries, traveling millions of kilometers.
Coco robots are equipped with four cameras, which constantly record the surroundings. Essas images are then cross-referenced with the Niantic’s spatial model, allowing the vehicle to know not only where it is, but also which direction it is pointing. Esse process is vital to reduce the risk of stopping at the wrong address or accidentally blocking sidewalks.
Coco Robotics’s operations span several key locations. The partnership has demonstrated significant potential to improve urban delivery logistics.
- The fleet of approximately one thousand robots already operates in Los Angeles, Chicago, Miami, Jersey City and Helsinque.
- Cada unit is capable of carrying up to eight large pizzas or four grocery bags.
- Deliveries made exceed the half million mark, with millions of kilometers covered.
- The collaboration focuses on improving accuracy in locating collection points and customer doors.
- The implemented system combines the Niantic’s VPS with the existing GPS to ensure greater stability and redundancy in navigation.
Urban Desafios and the logic behind the technological partnership
Coco Robotics faced significant limitations with GPS technology in dense urban environments. Seus robots, which move more slowly, require extreme precision to stop at exactly the right point. Essa is crucial to avoid disturbing the flow of pedestrians and to ensure the correct completion of the delivery. The company’s Executivos highlighted that while humans intuitively adjust when GPS fails, machines still struggle with this critical aspect of navigation.
Niantic Spatial, in turn, dealt with a spatial problem that had similarities to the robotics challenge. Para To make a character like Pikachu move realistically on the smartphone screen and interact correctly with the real environment, it was essential to have an accurate understanding of the physical space. Esse The same reasoning applies directly to a robot that needs to move safely and efficiently on city sidewalks.
John Hanke, who is in charge of Niantic Spatial, openly compared the two challenges, highlighting their similarities. Ele emphasized that the partnership with Coco represents the first major practical test of Niantic’s spatial mapping technology in a delivery robotics context. Atualmente, engineers from both companies work closely to integrate the VPS system seamlessly into Coco Robotics’s current fleet of robots.
Contribuição of users without noticing when building database
Pontos of battles and gyms within Pokémon GO concentrated a considerable portion of the images captured by players. Milhões of these locations, spread across several cities around the world, received photos from multiple angles, at different times of the day and under varying weather conditions. Users sent this material when participating in specific game tasks, such as “Campo Surveys”, launched in 2020, which encouraged exploration and visual recording.
Muitos players didn’t even realize that by participating in these activities, they were contributing to a much larger spatial database. The game encouraged walking and scanning points of interest in exchange for virtual rewards. Over time and massive user adoption, this strategy resulted in dense, detailed coverage of many urban areas, forming an unprecedented database.
Niantic Spatial transformed these 2D images, initially collected individually, into highly complex multiview 3D models. Repeated training on a wide range of intersections, building facades and stores allowed the artificial intelligence system to generalize to nearby or similar locations. Essa’s generalization capability works even when there are few new photos, ensuring the map remains robust and functional in new areas.
Futuro from urban logistics gains live map and more precision
In initial tests, the combined technology has demonstrated remarkable stability, especially in difficult-to-navigate urban environments. Robots equipped with the VPS system can identify the correct side of the street and stop just a few centimeters from the customer’s door. Essa improved accuracy is crucial and promises to significantly improve the experience of those receiving food or shopping without leaving home.
Coco Robotics has already established partnerships with several fast food delivery services and other companies. The fleet of robots, easily identifiable by their pink color, travels along sidewalks in varying weather conditions. With the implementation of VPS, the company expects to drastically reduce delivery delays. Tais delays are often caused by deviations or location errors that traditional GPS can generate.
Especialistas in technology and logistics are closely following this advance, as the solution does not depend exclusively on satellite signals. Ela uses advanced computer vision to “see” and interpret the world in a way more similar to human perception. Além In addition, the visual map continues to dynamically update as new data is collected, including from the robots’ own routes, creating a continuous learning system.
Outros examples of user-generated data appear across different technologies. reCAPTCHA, for example, used user typing to digitize old books and later train autonomous cars with complex street images. Da Similarly, autonomous vehicle companies collect millions of miles driven by everyday drivers to continually improve their driving systems. No Pokémon GO, the environment scanning mechanism was always optional, activated by the player’s express choice. Niantic claims to maintain full transparency about data collection and use since 2019. Não there is no indication that the VPS system is being shared with government authorities at this time.
The partnership between Niantic Spatial and Coco Robotics paves the way for the development of a “living map” of the urban environment, which gains detail and precision over time. Cada new robot delivery can potentially feed the model with updated images and fine-tuning. Cities are constantly changing, with new construction, emerging businesses and frequent changes in traffic, requiring an adaptable mapping system.
Niantic Spatial envisions a true “Cambrian explosion” of practical applications in robotics. The company’s main focus lies in creating spatial descriptions that machines can understand and interpret autonomously. In the future, robots and other smart devices will be able to “read” the environment around them with greater independence, without relying excessively on external signals, such as those from satellites.
For now, testing with the Coco Robotics focuses mainly on last-mile deliveries, a critical segment of logistics. The initial results obtained promisingly indicate that the intelligent combination of historical game data with current computer vision technologies has the real potential to resolve significant bottlenecks in urban logistics.
