Update: Discover how AI and perception are revolutionizing mobile robotics in Episode 173 of The Robot Report Podcast. Join Amir Bousani, CEO of RGo Robotics, and Jacob Petersen, Chief Commercial Officer of Wheel.Me, as they delve into the innovations driving the future of autonomous mobility.
Introduction
In the rapidly evolving world of robotics, wheel.me stands out with its ambitious mission: to transform anything that moves into an autonomous mobile robot. This case study explores how wheel.me, in collaboration with NVIDIA and RGo Robotics, developed a cutting-edge solution that's reshaping the landscape of mobile robotics.
Wheel.me’s mission is to transform anything that moves into an autonomous mobile robot. For an AMR to deliver value, it needs to work everywhere, all the time, and make intelligent decisions. This is where RGo Robotics and NVIDIA come into the picture. RGo turns every AMR into an intelligent one by giving it human-level perception and the visual-spatial information needed to make smart decisions. NVIDIA provides powerful GPU-accelerated computing, the NVIDIA Isaac robotics platform , and advanced AI capabilities available at the edge to enable all of it to happen.
Phase One - build a robot that works everywhere, all the time
Wheel.me decided to build its next generation of robot, Genius 2, with the NVIDIA Jetson Orin platform and RGo Perception Engine, leveraging ROS as much as possible. Key considerations included time to market and getting the best and more adaptable perception solution, recognizing that perception is at the core of an intelligent AMR.
NVIDIA Jetson is a powerful platform for edge AI and robotics encapsulating a perception-centric approach that enables easy integration, through ROS, to multiple applications and tools like fleet management.
RGo Perception Engine is a software stack that uses a range of proprietary compute vision and AI algorithms (including visualSLAM) to provide real-time 6DoF localization information everywhere, even in dynamic and complex environments and outdoors.
Putting it all together quickly, into a deployable solution involved technical choices and decisions including:
Sensor support – including choosing the right sensors, connectivity and SW support.
Design architecture - whether to run native, YOCTO or as a modern container. The wheel.me team decided on the latter, enabling modern CI/CD container stack while making the most out of the resources available on NVIDIA Jetson Orin.
Shared resource management – Dealing with the resource management challenges of a platform that shares a single compute system between the perception, autonomy and application stack. This required ensuring proper resource availability for the soft-real-time compute-intensive computer vision, 3D and AI applications (such as RGo’s online Perception solution) while balancing higher level applications requirements in the shared compute environment.
The wheel.me team, with the help of the experienced team at RGo, managed to bring up a solution within a few months leveraging the ready-to-use NVIDIA Jetson Orin system-on-modules and JetPack software stack with native ROS support.
Migration from industrial PC to modern GPU-accelerated computing is never easy, let alone while adding intensive computer vision, 3D perception and AI workloads. However, NVIDIA’s perception-centric design and software support enabled a seamless transition and upgrade. Utilizing the modern container architecture, wheel.me and RGo worked seamlessly on the NVIDIA Jetson software stack for efficient real-time perception and autonomy.
The wheel.me platform is now running in production in the most demanding environments, turning everything that moves into autonomous intelligent robots.
Phase two – Take robot intelligence to the next level with AI
Once the basics of navigation everywhere, all the time, is solved, the wheel.me team is looking at how to take their platform to the next level, leveraging the latest advancements in AI.
Having built their robot on GPU-accelerated compute like NVIDIA Jetson, vision and AI perception software from RGo and ROS framework, wheel.me has the foundation layer for quickly adopting and deploying advanced AI capabilities into their solution.
Among the challenges that the wheel.me and RGo are looking to solve are site commissioning. The time between when the first robot gets into the warehouse and the last engineer/operator is out is considered one of the biggest bottlenecks of adopting automation at scale. Reliable and accurate perception in combination with advanced AI has the potential to significantly alleviate this pain. Among the solutions currently being explored:
Automating robot commissioning by leveraging 3D vision, semantics and learning algorithms from both RGo Perception Engine and NVIDIA Isaac Perceptor (including CUDA accelerated libraries like nvblox)
Connect through ROS to NVIDIA Omniverse and Isaac Sim, leveraging digital twin for planning and testing the commissioning
RGo already integrated its perception stack with some NVIDIA Isaac capabilities on the Nova Carter reference robot, showing promising results for achieving the desired goal of streamlined commissioning. The teams are now integrating the advanced capabilities onto the wheel.me robot to test in real customer environments.
Summary
The collaboration between wheel.me, NVIDIA, and RGo Robotics has yielded impressive results:
Rapid development cycle: The team brought up a solution in just a few months.
Seamless transition to GPU-accelerated compute with intensive computer vision workloads.
A production-ready platform operating in challenging environments.
Foundation for quick adoption and deployment of advanced AI capabilities like 3D vision perception, semantic understanding and object detection.
As wheel.me continues to push the boundaries of what's possible in autonomous mobile robots, their work with NVIDIA and RGo Robotics positions them at the forefront of innovation in the field.
Ready to bring human-level perception and intelligent decision-making to your robotics solutions? Contact us to learn how RGo Robotics can empower your AMRs with cutting-edge perception and AI technologies.
Kommentare