Ghost Robotics Securely Updates Software at the Edge Using AWS IoT Greengrass

Powered by a robust 5G communications network, Ghost Robotics builds highly durable and scalable quadrupedal unmanned ground vehicles (Q-UGVs) using Amazon Web Services (AWS). Like many robotics companies, however, it needed a way to update its software at the edge quickly and securely to meet its customers’ data needs. Ghost Robotics didn’t have to look far for a solution, finding one among the AWS services designed for Internet of Things (IoT) applications.

Using AWS IoT Greengrass, an IoT open-source edge runtime and cloud service that helps developers build, deploy, and manage device software, Ghost Robotics can now quickly configure and update its robotics applications remotely and in person. This solution lets the startup support Haskell, an architecture, engineering, and construction firm, in creating a digital twin for a major football stadium.

Headquartered in Philadelphia, Pennsylvania, Ghost Robotics delivers advanced robotics solutions to support commercial and military applications. Its legged robots are designed to handle multiple types of terrain and have a 3-hour battery life compared to the 90-minute battery life of its competitors’ UGVs. These durable robots are powered by AWS Snowcone, an edge computing, edge storage, and data transfer device with 8 TB of usable storage purpose-built for use outside of a traditional data center.

Ghost Robotics has used AWS services to support its robotics solutions. The company uses Amazon SageMaker—a machine learning (ML) service used to build, train, and deploy ML models for virtually any use case—to ingest robotics data from its inference endpoints in the cloud, run data through custom ML models, and generate reports for its clients. However, Ghost Robotics realized that it needed a way to securely deploy updates to its robots at the edge. “We wanted to quickly configure and manage all our different applications,” says James Laney, autonomy and applications head at Ghost Robotics. “We also needed to be able to load these configurations on our robots, whether remotely or right next to the Q-UGV.”

To help the company meet this need, the AWS team proposed a solution using AWS IoT Greengrass. Using this service, the startup realized that it could not only provision its core robotics software but also deploy software for various use cases when needed.

In May 2021, Ghost Robotics began to test AWS IoT Greengrass while working with Haskell on a project for the Jacksonville Jaguars, a National Football League team. “We wanted to create a digital twin of the stadium to inform future expansions, renovations, and other projects,” says Hamzah Shanbari, manager of construction technology and innovations at Haskell. The construction company chose to work with Ghost Robotics because of its customer centricity and advanced Q-UGV technologies, which stood out in contrast to a competitor that Haskell had previously contacted. Ghost Robotics’ solution also offered the flexibility that Haskell required for the project. Using AWS IoT Greengrass, which integrates with ML models and makes it simple to update and implement software components, Ghost Robotics was able to deploy new software as the project occurred and quickly process the Q-UGV data. Using these capabilities, Ghost Robotics was able to increase its data delivery speed.

To create the digital twin, Haskell wanted to combine UGV scans with two additional sources of data: drones and terrestrial lasers. Ghost Robotics used its Q-UGVs to capture data from the stadium’s interior corridors, which the terrestrial laser and drone technologies were unable to access, and it used AWS IoT Greengrass to manage its robotics software for the project. “Using AWS IoT Greengrass, our scan data is stored on AWS,” says Laney. “With this functionality, we can quickly query the data collected by our Q-UGVs and help our customers to do the same.” Ghost Robotics was able to scan the stadium’s interior and deliver data to Haskell in 1 day, compared to the several weeks necessary to perform and collect data from a terrestrial laser scan.

Ghost Robotics can reliably store its Q-UGV data on AWS, centralizing access for its customers. After the stadium scan was complete, the data was copied off the robot’s external scanner and stored on Amazon Simple Storage Service (Amazon S3), an object storage service that offers industry-leading scalability, data availability, security, and performance. Haskell can easily access Q-UGV interior scans on Amazon S3 and reference them against external drone and terrestrial scans. This helps Haskell identify deviations and fix errors quickly, thereby increasing the accuracy of the digital twin.

By using AWS IoT Greengrass, Ghost Robotics gained the ability to securely deploy updates and new applications to its robotics software from any location. Haskell plans to receive its terrestrial laser scans in July 2021, and it will combine them with its drone and Q-UGV data to create a digital twin of the Jacksonville Jaguars stadium.

In the future, Ghost Robotics will continue to use AWS services to improve its Q-UGV technology. In particular, it is exploring ways to stream its data directly from the Q-UGV to AWS. Ghost Robotics will also train ML models using the data it collected through the Haskell project and use AWS IoT Greengrass to develop a fleet management system, reducing the need for manual upgrades. “Our Q-UGVs are essentially IoT devices with legs, and AWS IoT Greengrass is exactly what we need to power them,” says Laney.