Google’s artificial intelligence lab recently published a new paper explaining the development of Robotic Transformer 2 (RT-2), a novel vision-language-action (VLA) model that “learns from both web and robotics data, and translates this knowledge into generalized instructions for robotic control.”

This technology promises to upend many industries, as robots will be able to perform increasingly complex tasks as real-world experiences are incorporated into the data driving their underlying models. 


But what does it mean for the geospatial industry?

 

In this blog I’ll discuss how RT-2, and models like it, lay the groundwork for technology that will not only improve rapid geospatial data collection and feature extraction, but will enable instantaneous insights and actions enabling the operationalization of geospatial data in real-time. 

 

I’ll do this by diving into two of the largest use cases for geospatial data: utility corridor management, and municipal asset management.

 


Utility Corridor Management

 

Back in March of this year I wrote on the ROI utilities could realize from EO data without total automation. I concluded that expensive geospatial data sets would not yield meaningful ROI for utilities until field operations reached sufficient automation. With the release of RT-2, it looks like that time is closer than I may have thought!


It is unlikely that robots enabled with RT-2 like AI will displace traditional aerial and satellite data sources in utility corridor management, but rather, will not only leverage but enrich these geospatial data types. This technology will yield real-time updates, insights, and operationalization of geospatial data, leading to the near-full automation of utility corridor management. 

 

By fusing vision, language, and action, RT-2 allows robots to perceive their surroundings, comprehend natural language commands, and execute tasks efficiently. RT-2 is a VLA in its early stages, as it and models like it mature the complexity of tasks robots are able to complete will be increasingly astounding. 

 

The deployment of VLA-enabled robots by utilities will shift geospatial data from being a valuable tool for utility corridor management to an essential component of automated corridor management. With detailed geospatial data of the physical world integrated into their data models, these robots will be able to navigate complex environments efficiently and carry out a wide range of tasks including identifying, triaging, and trimming overgrown trees in utility corridors.

 

As mature VLA-enabled robots are integrated into utility corridor management, their contribution will be transformative. Enabled robots will efficiently survey and monitor vast stretches of utility corridors, detect anomalies, respond to changing conditions promptly, and integrate this real-world data back into the utility’s system of record. With its ability to translate data into actionable insights in real-time, VLA-enabled robots will minimize downtime, optimize overall operational efficiency, and displace field workers prone to injury and error.

 

VLA-enabled robots and traditional satellite and aerial geospatial data sources will have a symbiotic relationship, where each component complements the other’s limitations. While EO data provides highly accurate and precise measurements of the physical world at scale, VLA-enabled robots will contribute by not only enriching the data with contextual information and real-time insights, but taking immediate action. This combination will usher in nearly fully autonomous utility corridor management.

 

Municipal Asset Management


It is highly likely that VLA-enabled robots will not only displace human-operated geospatial data collection methods in municipal asset management, but significantly reduce the time from data collection to work order creation and issue resolution. 

 

Municipal asset managers have recently begun adopting a host of geospatial data into their operational workflows, specifically street-level LiDAR and imagery data processed through ML models to extract, identify, and triage municipal assets from signs and sidewalks to roads and manholes. This data is currently collected in large part by companies such as Cyvl.ai using sensors mounted to human-operated vehicles.  

 

Modern street-level data collection has vastly improved historic asset management workflows, but it still takes significant time to collect, process, deliver, and operationalize the data. VLA-enabled robots will not only remove the need to deploy human-operated vehicles to collect data, but will also reduce the need for field crews necessary to operationalize the data. This will dramatically reduce the time from data collection to operationalization. 

 

As VLA-enabled robots are deployed to manage municipal assets, they will be able to collect vital geospatial data, produce insights, address issues, and create work orders for problems beyond their scope in real-time. This kind of vertical integration of services will render traditional methods of street-level geospatial data collection obsolete. 

 

With VLA-enabled robots handling geospatial data collection and field operations, municipal asset managers will experience a paradigm shift in efficiency and productivity. This transformation will free up resources that were previously allocated to data collection and work order creation, allowing asset managers to focus on other solutions. Geospatial companies currently engaged in street-level data collection should seriously consider integrating VLA-enabled robots into their long-range business plans. 

 

Because VLA-enabled robots will significantly reduce the time from data collection to work order creation and issue resolution, it is highly likely that they will displace human-operated street-level geospatial data collection methods. 

 

Conclusion

 

There are a ton of use cases left to consider! But no matter the use case, VLA-enabled robots stand to upend traditional data acquisition and operational workflows to some degree or another. I predict they will have less of an impact on the collection of aerial and satellite data sources, but will enhance the ability to operationalize this data significantly. On the other hand, these robots will likely displace street-level data collection by vertically integrating data collection, processing, and field operations. 

 

This technology will improve operational efficiency, adaptability, and free up resources to enable more problem-solving. But it will come with consequences for jobs.

Buckle up, because we’re a hop skip and a jump away from SkyNet….