Edge AI and Robotics Set New Standards for In-Venue UX

The hospitality industry is exploring robotics to navigate persistent labour shortages with deployments now moving beyond the kitchen and into front-of-house, customer-facing roles. This trend could show a growing demand for localised data processing in busy public venues.

At the T-Mobile Arena in Las Vegas, hockey fans are being served by ADAM, a robot bartender that shows how automation can manage more than just simple, repetitive tasks. ADAM, which stands for Automated Dual Arm Mixologist, is a working application of edge AI computing within a dynamic hospitality environment.

The system was developed by the Las Vegas-based Richtech Robotics using Nvidia’s Isaac libraries. It was designed to address workforce challenges while creating unique customer interactions at the NHL venue.

“The hospitality industry faces substantial labour challenges, and ADAM is our answer to meeting those needs while elevating the customer experience,” says Matt Casella, former President of Richtech Robotics as of 2 December this year.

He adds: “With Nvidia’s Isaac platform, we’ve developed a solution that’s scalable, consistent and frankly, creates memorable moments for fans.”

The role of edge AI in real-time robotics

ADAM’s ability to interact safely and efficiently in a public space relies on its capacity to process vast amounts of data locally without relying on distant cloud servers. This is known as edge AI, a decentralised computing structure that reduces latency and enables faster response times.

For a robot bartender, this speed is crucial for pouring drinks accurately and reacting to its environment in real time.

The robot operates on an Nvidia Jetson AGX Orin, a compact edge AI computing platform capable of 275 trillion operations per second.

Before its first deployment, Richtech Robotics extensively trained ADAM in a virtual bar. Using Nvidia Isaac Sim, a robotics simulation framework, Richtech Robotics created a digital twin of the robot’s workstation.

This virtual environment included digital cups, utensils and varied lighting conditions, allowing the AI to learn how to handle real-world scenarios before installation. This simulation approach was used to generate synthetic data, which helped train the AI to identify objects even when faced with challenging visual conditions like glare or reflections from arena lighting.

Achieving low-latency processing

Using Isaac ROS 2 libraries, ADAM processes camera feeds, detects objects and calibrates its workspace continuously. For the system to be effective, its perception and reaction times must be almost instantaneous.

The robot’s perception system was built with the TAO Toolkit and optimised with TensorRT, enabling it to identify cups, measure liquid levels and adjust its movements with a latency below 40ms.

This low latency means ADAM can spot a misplaced cup or detect when foam reaches the rim of a glass and correct a pour without delay.

Such a rapid response time, which is processed locally on the device, is fundamental to the system’s operational success and a key use case for technologies that can deliver high-speed data processing.

Isaac Lab, Nvidia’s open-source robot learning framework, was then used to refine ADAM’s skills, including pouring and shaking beverages.

Expanding automation from hospitality to logistics

The technology underpinning ADAM is not limited to hospitality. Richtech has also developed Dex, a mobile humanoid robot designed for factory and warehouse applications. Unveiled at Nvidia’s GTC DC technology conference, Dex combines a wheeled autonomous platform with dual-arm manipulation for tasks such as machine operation, parts sorting and material handling.

Similar to ADAM, Dex was trained using a combination of real-world and synthetic data generated in Isaac Sim. The system runs on the Nvidia Jetson Thor, a robotics processor designed for the high demands of real-time sensor processing in industrial environments.

This expansion into logistics could signal a wider industrial trend towards integrating advanced robotics that require powerful onboard processing.

The initial deployment in a high-traffic public venue appears to be a success.

"The response at T-Mobile Arena has been phenomenal – people love interacting with ADAM,” Matt says.