Development kits for robotics and intelligent machines are helping designers incorporate artificial intelligence, machine learning (ML) and new connectivity technologies like 5G to not only speed up development time but also to develop smarter and safer machines. These development kits and platforms bundle everything needed — hardware, software, security, connectivity and AI — to fully develop, demonstrate and test new designs.
Addressing design complexity, particularly for projects that require expertise across engineering disciplines, these development kits and platforms simplify AI development and prototyping for a range of robotics applications. Here is a sampling of development kits and platforms.
Qualcomm Technologies Inc. recently unveiled its plans for 5G and edge AI robotics solutions with the launch of the Qualcomm Robotics RB6 platform and the Qualcomm RB5 AMR Reference Design. The new solution combines the Qualcomm AI Engine and 5G capabilities along with support for global sub-6-GHz and mmWave bands in mainstream, enterprise and private networks. It also provides a flexible architecture with expansion cards to support evolving connectivity features. This will enable the platform to support 3GPP Release 15 and Releases 16, 17 and 18 features as cards become available in the future.
Targeting a variety of applications like autonomous mobile robots (AMRs), delivery robots, automated manufacturing robots, collaborative robots, urban-air–mobility aircrafts, industrial drone infrastructure and autonomous defense solutions, the platform delivers edge AI and video-processing capabilities through the enhanced Qualcomm AI Engine, with support for 70–200 trillion operations per second (TOPS) (INT8). It supports 12 cameras via D-PHY and 18 cameras via C-PHY (seven concurrent), as well as AI processing with up to 24 simultaneous 1080p video streams.
Other key features include a Qualcomm Kryo 585 CPU and Qualcomm Spectra 480 image signal processor (ISP) that can process 2 gigapixels per second with high-performance capture of 200-MP photos, 8K video recording, and 4K HDR video capture. It also includes the Qualcomm Secure Processing Unit (SPU), which supports hardware root of trust, Qualcomm Trusted Execution Environment, secure boot and camera security.
The robotics platform provides an extensive set of hardware and software development tools that are customizable. The Qualcomm Intelligent Multimedia software development kit (SDK) combines multimedia, AI/ML, computer vision and networking building blocks to support end-to-end deployment of robotic applications, the company said.
The RB6 platform, together with the RB5 AMR reference design, will help robotics designers integrate robots into industrial applications. These applications range across sectors from warehousing and logistics to health care and utilities.
Built on Qualcomm’s robotics platforms, the Robotics RB6 platform provides customizable hardware and software development tools. This includes Qualcomm’s AI SDK, the Qualcomm Intelligent Multimedia SDK. This SDK combines multimedia, AI/ML, computer vision and networking building blocks to support end-to-end deployment of robotic applications. These tools are in combination with the RB5 AMR reference design, claimed as the first AMR reference design to offer integrated enhanced AI and 5G capabilities.
Qualcomm’s Robotics RB6 platform (Source: Qualcomm Technologies Inc.)
Nvidia also offers a robotics platform, and it is extensive. The Nvidia Isaac Robotics Platform is a complete end-to-end solution for robotics design. Tools include the Isaac Sim simulation platform, which the company recently upgraded with more capabilities, cuOpt for route optimization, Isaac ROS (hardware-accelerated software modules that run on the robots), Replicator (a synthetic data-generation tool) and pre-trained models. Here, we’re covering some of Nvidia’s most recent robotics development kits, designed to advance robotics and edge AI.
First up is Nvidia’s Jetson AGX Orin developer kit, an energy-efficient AI supercomputer for advanced robotics, autonomous machines, and next-generation embedded and edge computing. The company boosted the processing power over 8× compared with the predecessor, the Jetson AGX Xavier, with 275 TOPS in the same form factor and pin compatibility. Key features include Nvidia’s Ampere architecture GPU, Arm Cortex-A78AE CPUs, next-generation deep-learning and vision accelerators, high-speed interfaces, faster memory bandwidth and multimodal sensor support to feed multiple, concurrent AI application pipelines.
Nvidia’s Jetson AGX Orin developer kit (Source: Nvidia)
Along with the development kit, designers also can access the full Nvidia CUDA-X accelerated computing stack, Nvidia JetPack SDK, pre-trained models from the Nvidia NGC catalog and the latest frameworks and tools for application development and optimization, such as Nvidia Isaac on Omniverse, Nvidia Metropolis and Nvidia TAO Toolkit. All of these tools are designed to solve some of the biggest robotics and edge AI challenges, including 3D perception, natural-language understanding and multi-sensor fusion, the company said.
If you’re looking to get started with edge AI and robotics, the Nvidia Jetson Nano 4-GB Developer Kit is also available. Nvidia relaunched its Jetson Nano developer kit at the end of 2022. The Jetson Nano, launched three years ago, is an entry-level AI computer for embedded and edge AI applications that needs as little as 5 W to run, Nvidia said. It supports multiple neural networks in parallel for applications like image classification, object detection, segmentation and speech processing.
The key features of the Jetson Nano include a 128-core Nvidia Maxwell GPU, a quad-core Arm A57 processing system, a video encoder and decoder, 4-GB LPDDR4 and 16-GB eMMC memory and a variety of interfaces and I/Os. Nvidia said the same software stack supports all Jetson modules and provides access to a host of technologies, including Jetson Linux, developer tools and CUDA-X accelerated libraries.
For rapid AI vision prototyping, Adlink Technology Inc. recently launched its AI Camera Developer Kit, based on the Jetson Nano. The vision development kit integrates the Jetson Nano module and 8-MP image sensor with lens and vertical I/O as well as a variety of peripherals.
The AI Camera Dev Kit enables AI vision developers to quickly go from concept validation to viability testing. It comes with two AI vision applications — pose detection and object classification — and sample code pre-installed and Adlink’s edge vision analytics software, EVA, for labeling, training and inference to quickly develop a proof of concept (POC) or application.
In addition to the MIPI camera module, Jetson Nano SoC and EVA software, features include an industrial DI/O interface, USB Type-C hub for easy powering, a display, keyboard/mouse connection and data transmission, SD memory card with Linux OS, Nvidia’s JetPack SDK and OpenCV. Adlink said the kit enables users to start their AI vision POC in 5 minutes and convert the kit to Adlink edge AI vision devices.
Also targeting AI vision applications is eYs3D’s XINK computer vision development platform for vision-equipped robots. Introduced at CES 2023, XINK is described as both a platform-as-a-service and a hardware and development kit, targeting the development of next-generation autonomous robotic applications like the artificial intelligence of things, smart cities, indoor cleaning robots and outdoor agricultural robots for industrial and retail sectors equipped with field analysis, object recognition, obstacle detection, object tracking and following and route-planning functions.
The development platform provides all the necessary elements for product development, including high-performance compute power, AI accelerator, I/O controls, Flexi-bus communication peripherals, smart power management and machine-vision subsystems. The edge AI processing is powered by eYs3D’s new eCV1 AI chip that incorporates four multi-core ARM 64-bit CPUs and a 4.6 TOPS neural processor unit. An additional low-power ARM Cortex M4 processor can be used as an MCU, according to eYs3D.
The modular platform handles low-level programming, which frees developers to use cut-and-paste coding for application-specific design while reducing design cycles for quicker commercialization, the company said.
The platform provides H.264 compression for video streaming as well as ISP support features. Image data can come from either an external ISP, such as eYs3D’s separate eSP87x series stereo video and depth processor, or from the ISP soft code inside the XINK CPU.
In addition, XINK’s Crypto Engine includes ARM’s TrustZone for security as well as a pseudorandom-number generator and other encryption to protect user data through hardware-based isolation. The platform supports a variety of AI inference tools, including TensorFlow, TensorFlow Lite, PyTorch, Caffe and TVM.
Advertisement
Learn more about NVIDIAQualcomm
