NVIDIA launched a new NVIDIA Omniverse experience for the Nvidia Isaac SIM to support the need for accurate, reliable and simple simulation tools in robotics. With many research laboratories and universities closed for an indefinite period of time, robotics around the world are separated from materials and physical environments to develop, test and implement independent machines. Now, more than ever the best simulation tools can help maintain vital stresses in robotics and automation.
This robotic experience is NVIDIA ISAAC simulation of SIM simulation tools included in Omniverse. The ISAAC SIM building on Omniverse platform brings several useful improvements on existing robot flows:
Take advantage of very precise physics Omniverse, which directs direct integration with leading physics executives, such as Nvidia PhysX SDK for the solid body dynamics and the flow of Nvidia for fluid dynamics.
It has a renewed access of interoperability, with deep integration with NVIDIA ISAAC SDK and ROS Extensions.
Omniverse Experience for ISAAC SIM can easily expand. You can now use Omniverse Kit Python Python interface and the Python Expansion system to customize ISAAC SIM on your own cases of use.
It is designed to be abandoned, with an architecture that supports the course of work in its local workstation on the edge of the hardware Nvidia Jetson and through the cloud with NGC NVIDIA.
Here is looking closest to reviews and examples on these versions.
Bridge workflow
The simulation environment of SIM Omniverse Isaac Sias recognizes two important robotic frameworks: Nvidia Isaac SDK and ROS.
ISAAC SDK is a professional quality framework for application development, navigation and manipulation and robotics. ISAAC SIM communicates for ISAAC SDK using the Robot motor bridge extension to send simulated sensor data, units, terrible truths and TCP / IP scenarios. ISAAC SDK Documents Highlighted a series of simulation scenarios navigation between the autonomic warehouse of several complex robots with simple handling with Jupyer gears. In addition, ISAAC SDK with virtual and physical robots of the ISAAC SIM bridge for multiple benchmarks based on Jetson such as Carter and Kaya robots.
Omniverse Isaac Sim also supports Ross, popular robotist settings. The Izaac Sim Ros Bridge Extension permits robot researchers to boom their present operating approaches with Omniverse Multi-GPU simulation and collaboration, simulation and real-time cooperation. This version provides an ovid pattern that passes through publication and subscribe to themes, including a common condition, common control, TF, camera and repeatedly.
Agile automation
Another advantage of ISAAC SIM in Omnivere is that the dynamics of a complex solid body is now easy to simulate for robotic and automation scenario. UR10 applications illustrate containers of container platforms to contain container platforms using the universal robot robot and suction base clamps. In the container filling, the user falls from different parts in a container maintained by the robot until the trash cannot be so difficult to tackle the suction clip does not succeed. The parties fall everywhere. A robotic arm can detect a fire tank, pick it up and position it to receive new parts. Although the filling and stacking containers are standard procedures, these examples show how intelligently increase, so robotic automation is agile and more adaptable to dynamic environments.
UR10 sample applications come with a very detailed source and documentation. Mix several central components of simulation:
- Loading the robot model using URDF importer
- Creating a state automation machine.
- Combining traditional road planning with advanced movement generation using the Riiman Movement Policy (RMP)
- Simulating sensors inside and outside the robot.
Custom Robots
In addition to Robot UR10, Omniverse Isaac SIM has a library of the property, effects and environment ready to use. However, you can also work with your own custom robots using the ISAAC SIM URDF importer extension. This expansion of accounts and converts the URDF file to USD Omnic representation, which makes it more useful and persistent format for advanced simulation. The URDF importer extension is also supplied to multiple imports of samples and Python API for loading and configuring their own robots.
In this version, the documentation has Kaya pattern that leads you through the process of importing and configuring a custom robot for Omniverse Isaac SIM. He has Nvidia Kaya as a reference material. Kaya is a printed 3D robot launched Nvidia Jetson Nano. It is designed to make robotics available amateurs and education. However, it is a particularly stimulating reference platform that is precisely simulated. It has several CAD files, a complex holonomic unit and a number of sensors that must be configured to simulate your task, location, object recognition and the Neometry Options.
Feel the world
In this experience of robotics, Omniverse you can view the benefits of real-time reality for robotic simulation. In addition to real visuals, Real-Time Omniverse Police simulates physical behavior of light. This allows the ISAAC SIM to correctly model the physical properties of multiple light sensors, including RGB cameras, depth rooms and lido. The Lidar sample shows how to use the ISAAC SIM LIDAR API to create and activate these sensors in the form of independent robots. Define resolution, display field and rotation speed so that the hardware properties of the room used for a physical robot.
Omniverse and Isaac
Omniverse is the basic base of the NVIDIA simulator, including the ISAAC platform, which now includes several new features.
Built on the NVIDIA Omniverse platform, ISAAC SIM is a robotic simulation application and a synthetic data generation tool. Allows robotics to train and more efficiently test their robots by providing a realistic robot simulation that interacts with convincing environments that can increase coverage outside of what is possible in the real world.
This version of ISAAC SIM also adds support options and multi-camera sensors and importer PTC OneShec CAD to facilitate the introduction of 3D assets. These new features will expand the amplitude of the robot and environment that can be modeled and successfully distributed in all aspects: from the design and development of the physical robot to form a robot and the implementation of “Twin Digital” and tests.
The developers have been using the benefits of a strong simulation environment for testing and exercising the robot. Too often, failures restrict the adoption of the simulator, but Isaac Sim addresses these shortcomings, “Andrews said.
Real simulation
In order to offer realistic robotic simulations, ISAAC SIM uses powerful Omniverse Platform, including an advanced graphic processing (GPU), physical simulation and physical simulation with PHSX 5, Real Time Tracking Photo and Language Material Definition ( MDL) that supports the display physically.
ISAAC SIM is designed to address many robotic use cases, including handling, autonomous navigation and synthetic data for training data. Its modular design users allow users to easily adjust and expand the set of tools to customize many applications and environments.
In addition to Omniverse, ISAAC SIM enjoys the core connectors and omniere connectors, which enable the cooperation, exchange and import of environmental and robot models in the standard description of the Pixar Universal Scene (USD). Engineers can easily connect the brain from the robot to the virtual world through the ISAAC SDK and ROS / ROS2 interface, with all Python scripts and accessories for importing robotic models and environments.
The synthetic generation of data is an important tool that is increasingly used to train perception models in today’s robots. Get properly marked information about the real world is expensive and expensive effort. But in the case of robotics, some of the necessary training data can be too heavy or dangerous to collect in the real world. This is especially true in robots that have to work near people.
ISSAC SIM has integrated assistance to different important types of sensors in forming perception models. These sensors include images and segmentation of RGB, depth, demarcation and segmentation, andrews said.
So how much is that supposed to be?
“You just want, for a reason, close the SIM breach,” Andrews said. “If you have a little mistake, you can accumulate in your simulation, you can recover over time to model where it does not make something correct with wheel descriptions [functions], your robot can well go well. But this The error is accumulated and the robot can be completely out of the real world course. “”
He added, “The more you can enter reality, there is only a better experience that you will have when engineers try to use. In the world of simulation, always facing the idea” now that I have a real material, what is the value to continue using the simulator? “”
Get better data
In the open beta, Nvidia has the ability to publish synthetic data in kitts format. These data can be used directly with NVIDIA transmission tool to improve model performance with certain case data, andrews said.
The randomization of the domain distinguishes parameters that define a simulated scene, such as lighting, color and texture of materials on the scene. One of the main goals of the domain randomization is to improve the formation of automatic learning templates (ML) by exposing the neural network with a wide range of domain simulation parameters. This will help the model generalize when you find real world scenarios. Indeed, this technique helps learn a model for ignoring. Isaac Sim recognizes the randomization of many different attributes that help define a particular scene. With these capabilities, ML engineers can be sure that the synthetic data set contains enough diversity to promote the performance of a robust model.
Simulations can save time and other resources.
In real life, 50 engineers can work on the project, but they can have one prototype hardware. With something like Isaac, 50 software engineers could act at the same time, Andrews said. And engineers should not be in the same physical space because they can work on the parts of the project remotely.
On top of that, it could be dangerous to try the real world robot if your controls are not correct. You can find yourself in human or worse. But if you try in Omniverse, the simulation will only be born.
Over time, Nvidia has added objects such as multi-camera support, Fish-eyed camera lens and other sensors that improve robot functions and their ability to feel the environment. The more improved components in the real world, the more Isaac simulation can be updated in Omniverse, “Andrews said.
Import data CAD and Robot.
In addition, Isaac SIM enjoys Omnius and Omniere core connections, which enable the cooperation, exchange and import of the construction of robotic environments and models in the description of the universal scene (USD), Nvidia said. Isaac SIM includes importer PTC-Onghe Nave CAD to facilitate the import of 3D property.
“For example, factory planners can establish a factory plan, but the robot team must take data to build a simulation,” Andrews said. “Omniverse drivers earn time in data transfer between formats. It also includes a robot physics such as Dofbot, UR10 and Frank Emika as well as Kaya Reference Robot.”
Support for multiple cameras, sensors.
The open beta version of NVIDIA supports reduced cameras with synthetic data. The ISAAC SIM also supports ultrasonic sensors, power sensors, inertial measurement units (IMU) and custom modeling models.
“Isaac Sim sends sensors with multiple cameras to the RVI-La Ros display,” Andrews said. “3D flows are important for robotics and autonomous vehicles in construction and other applications.”
Synthetic data generation.
“Getting the real world, the marked data is expensive enough and expensive effort,” Nvidia said. This can be especially difficult in cases where autonomous robots or vehicles for safe action around people must be trained. In addition, the data note implies a lot of human work.
The ISAAC SIM has integrated support for the types of simulated sensors, including RGB, depth, boxes and segmentation, said Nvidia. Open Beta generates data in Kitts format and can be used with NVIDIA transmission tool to improve model performance with certain data case.
In combination with some real-world data, synthetic data can help solve long-tailed problems, “Andrews said.” There are many software engineers, but you will never have thousands of physical testing. There is a complete ecosystem of companies that come and succeed with synthetic data. “
4 comments