News

What does ChatGPT mean for robotics?

Source: The Robot Report, June 5, 2023 [1]
ChatGPT, a conversational AI tool, has the potential to enhance robotics by improving communication and decision-making capabilities. It uses Natural Language Processing (NLP) and generates more human-like text, impacting robotics. Corobotics, or cobots, are designed to work safely in the same environment as humans, using techniques like machine vision and reinforcement learning. However, ChatGPT faces potential issues both technically and commercially, highlighting the potential challenges in achieving natural interaction with robots.

AMD and MassRobotics announce robotics startup challenge to accelerate robotics innovation

Source: AMD, May 25, 2023 [2]
AMD and MassRobotics launched the AMD Robotics Innovation Challenge, focusing on adaptive computing and integrating AMD Kria SOMs in robotics projects. Early-stage startups with digital and analog design expertise are eligible, with evaluations starting mid-June and ending October 2023.

Harvard dropouts raise $5 million for LLM accelerator

Source: EETimes, June 5, 2023 [3]
Harvard students Gavin Uberti and Chris Zhu have raised $5.36 million in a seed round for Etched.ai, an AI accelerator chip for large language model (LLM) acceleration. The seed round valued the company at $34 million. Uberti, CEO, explains the need to improve 8-bit MAC SIMD operations and the changing world of language models.

NVIDIA unveils updates for metropolis for factories, Isaac AMR & more

Source: The Robot Report, May 30, 2023 [4]
NVIDIA CEO Jensen Huang unveiled new projects, focusing on generative AI advancements, embracing accelerated computing and AI adoption by 40,000 companies and 15,000 startups.

The case for running AI on CPUs isn't dead yet

Source: IEEE Spectrum, June 3, 2023 [5]
A group of AI researchers, led by Julien Simon, have shown the potential of the CPU in AI development. They demonstrated Intel's Q8-Chat, a 32-core Intel Xeon processor LLM, with blazing-fast queries. GPU hardware outperformed CPUs and dedicated AI accelerators, thanks to the massively parallel architecture of GPUs and their integration into open-source frameworks like TensorFlow and PyTorch.

Software-defined hardware architectures

Source: Semiconductor Engineering, June 2, 2023 [6]
Hardware-software co-design is established, but the migration to multiprocessor systems adds many new challenges.

How many sensors for autonomous driving?

Source: Semiconductor Engineering, June 1, 2023 [7]
Sensor technologies are still evolving, and capabilities are being debated.

Take AI learning to the edge with NVIDIA Jetson

Source: NVIDIA Technical Blog, June 5, 2023 [8]
The NVIDIA Jetson Orin Nano and Jetson AGX Orin Developer Kits are now available at a discount for qualified students, educators, and researchers. These high-performance, low-power modules offer a compact yet powerful platform for developing robotics and edge AI vision. With 80x performance, they enable prototyping advanced AI-powered robots and other edge AI devices.

Papers

HuNavSim: A ROS 2 human navigation simulator for benchmarking human-aware robot navigation

Source: arXiv, May 17, 2023 [9]
HuNavSim is an open-source tool for simulating human-agent navigation behaviors in mobile robot scenarios, using ROS 2 framework and realistic human behaviors.

Accelerated Deep-Learning inference on FPGAs in the space domain

Source: Technical Unviersity of Munich, May 11, 2023 [10]
Artificial intelligence has found its way into space, and similar to the situation on ground demands powerful hardware to unfold its full potential. With the heterogeneous compute platform that is offered by the space-grade variant of the Versal, AMD Xilinx presents a system that is particularly targeted at accelerating AI inference in space. This paper investigates the design flow and the achievable performance of this novel device. We present benchmark results in terms of concrete figures and measurements, i.e., throughput, latency, and power consumption, achieved by a predesigned hardware accelerator realized on the system, and compare them to a previous generation platform.

Previous Hardware Acceleration in Robotics Newsletters

Past ROS 2 Hardware Acceleration Working Group meetings

  1. Crowe, S. (2023, June 5). What does ChatGPT mean for robotics? The Robot Report. https://www.therobotreport.com/what-does-chatgpt-mean-for-robotics/ ↩︎

  2. AMD and MassRobotics announce robotics startup challenge to accelerate robotics innovation. (2023, May 25). AMD.com. https://community.amd.com/t5/adaptive-computing/amd-and-massrobotics-announce-robotics-startup-challenge-to/ba-p/608928 ↩︎

  3. Ward-Foxton, S. (2023, June 5). Harvard dropouts raise $5 million for LLM accelerator. EETimes. https://www.eetimes.com/harvard-dropouts-raise-5-million-for-llm-accelerator/ ↩︎

  4. Wessling, B. (2023, May 30). NVIDIA unveils updates for metropolis for factories, Isaac AMR & more. The Robot Report. https://www.therobotreport.com/nvidia-unveils-updates-for-metropolis-for-factories-isaac-amr-more/?spMailingID=90836&puid=2502618&E=2502618&utm_source=newsletter&utm_medium=email&utm_campaign=90836 ↩︎

  5. Smith, M. (2023, June 3). The case for running AI on CPUs isn't dead yet. IEEE Spectrum. https://spectrum.ieee.org/ai-cpu ↩︎

  6. Bailey, B. (2023, June 2). Software-defined hardware architectures. Semiconductor Engineering. https://semiengineering.com/software-defined-hardware-architectures/?cmid=3eccebfc-4c47-4228-8caf-43ed20dfd131 ↩︎

  7. Koon, J. (2023, June 1). How many sensors for autonomous driving? Semiconductor Engineering. https://semiengineering.com/how-many-sensors-for-autonomous-driving/?cmid=e405ffc8-7484-4bcf-bb2e-7e2e6dcd6aca ↩︎

  8. Black, J. (2023, June 5). Take AI learning to the edge with NVIDIA Jetson. NVIDIA Technical Blog. https://developer.nvidia.com/blog/take-ai-learning-to-the-edge-with-jetson/ ↩︎

  9. Perez-Higueras, N., Otero, R., Caballero, F., & Merino, L. (2023, May 17). HuNavSim: A ROS 2 human navigation simulator for benchmarking human-aware robot navigation. arXiv.org. https://arxiv.org/abs/2305.01303 ↩︎

  10. Petry, M., Gest, P., Koch, A., Ghiglione, M., & Werner, M. (2023, May 11). Accelerated Deep-Learning inference on FPGAs in the space domain. Technical Unviersity of Munich - Big Geospatial Data Management - Willkommen!. https://www.bgd.ed.tum.de/pdf/2023_Computing_Frontiers_Versal_Michael.pdf ↩︎