NVIDIA Unveils Isaac AMR Platform

At CES 2022, NVIDIA unveiled the Isaac autonomous mobile robot (AMR) platform. The end-to-end platform is designed to speed up AMR deployment and improve efficiency.

The NVIDIA AMR system is designed for large-scale deployments in big facilities, where mapping can take a long time and fleet management is usually complex. Isaac AMR works with a number of NVIDIA platforms, including Omniverse, Isaac Sim, DeepMap, Metropolis and ReOPT.

The platform uses NVIDIA Omniverse to create a digital twin of the facility where AMRs will be deployed. Isaac Sim can then simulate the behavior of the robot fleet, people and other machines. The platform will also generate synthetic data for training AI models.

Facilities are mapped with DeepMap, a startup the company acquired in June 2021. DeepMap created high-resolution mapping software for autonomous vehicles. NVIDIA uses the same software to map indoor and outdoor facilities for AMRs.

The DeepMap cloud software development kit (SDK) can speed up robotic mapping of large facilities from several weeks to several days, with centimeter-level accuracy.

Maps can be updated in real time as the environment changes. The DeepMap SDK allows robots to identify real-world objects represented by pixels on the map and respond to them accordingly.

NVIDIA Metropolis is a video analytics platform that works together with SmartMap. It provides access to cameras and sensors deployed throughout the facility. Metropolis also helps AMRs better understand their surroundings and avoid areas with heavy traffic.

The final part of Isaac AMR is the NVIDIA ReOpt AI software libraries. These libraries can be used to optimize vehicle route planning in real time, taking into account robot speed, battery life, vehicle size and weight, and facility layout. With ReOpt and Isaac Sim, multiple AMR interactions can be simulated before the robots are deployed.

The technologies that make up the Isaac AMR platform can be deployed and managed on EGX servers using NVIDIA Fleet Command.

According to Interact Analysis, the number of deployed mobile robots will reach 53,000 units by 2025. In 2020, that number was just 9,000.

“As more ROS developers use hardware platforms that include additional computing capabilities designed to offload the CPU, ROS is evolving to make it easier to efficiently use these advanced hardware resources. Working with an accelerated computing leader such as NVIDIA, and its extensive experience in AI and robotics innovation, will bring significant benefits to the entire ROS community”

Brian GerkeyCEO and co-founder of Open Robotics

«From intelligent automation in manufacturing to last-mile delivery robots, robots are becoming increasingly common in everyday life. However, developing industrial and commercial robotics can be complex, time-consuming and expensive. Unstructured environments are also common in many use cases and scenarios. The NVIDIA Isaac robotics platform can address these challenges with a comprehensive solution that helps reduce costs and simplify development and time to market», the company said. 

The Isaac Autonomous Mobile Robots (AMR) platform extends NVIDIA Isaac for developers building and deploying robotics applications, enabling mapping, site analytics, fleet and route optimization on NVIDIA EGX servers. The new platform will help improve and accelerate AMR applications for the logistics industry, from warehouses to retail.

Benefits of the accelerated platform for robotics and artificial intelligence:

1. Data generation. Collecting and labeling datasets takes time and is expensive. More diverse training data enables more accurate AI models. Training can be improved with NVIDIA Isaac Sim synthetic data.
2. Simulation and testing. Robots are trained and tested virtually in photorealistic and physically accurate environments using NVIDIA Isaac Sim, running in the NVIDIA Omniverse virtual environment.
3. Robot training. You can choose the optimal NVIDIA platform for training your AI model, from local workstations to scalable AI in the data center and cloud. NVIDIA can also help accelerate development with the TAO Toolkit, which includes pretrained production-quality models for use as-is or for fine-tuning with only part of the data needed to train from scratch.
4. Application development. High-performance robotics applications can be developed using hardware-accelerated SDKs such as Isaac ROS GEM for ROS-based robots, the DeepStream SDK for streaming video understanding, NVIDIA Riva for natural language processing, and the Isaac SDK for complete infrastructure.
5. Deployment and management. NVIDIA offers a complete continuous workflow for seamless deployment. Robotics applications can be moved from NVIDIA Jetson to EGX edge computers to create truly autonomous machines. Scaling services are also available in EGX Fleet Command.