Embedded.fm · 14 Feb 2018 · 50m · Episode 234
Dusty Franklin, a developer evangelist at NVIDIA, discusses the Jetson TX2, a low-power embedded module designed to bring AI and computer vision to edge devices like robots and drones. He explains how the platform's GPU enables parallel processing for tasks such as autonomous navigation, object detection, and reinforcement learning, often using simulation environments like ROS Gazebo to train agents safely. Franklin also details the 'Two Days to a Demo' tutorial, which guides developers through deploying and retraining neural networks using NVIDIA's TensorRT and DIGITS tools, and mentions the Redtail Project as an example of autonomous drone navigation.
early-stage companies
Other shows on the same ground, closest first.
Hackaday editors Mike Szczys and Elliot Williams discuss new AI-on-the-edge hardware from BeagleBone and Google, the reverse engineering of a keytar firmware, and a critical analysis of Tesla's self-driving safety data. They also cover a custom smart lamp build, flexible PCB contests, and Elliot provides an audio tour of the Make Munich maker fair.
Alex Nesic, Co-Founder and Chief Business Officer of Drover AI, explains how the company uses edge-based computer vision and AI to solve GPS inaccuracies in shared micromobility fleets. He details the PathPilot module, which provides contextual location awareness to ensure regulatory compliance and safety for e-scooters and bikes. Nesic also discusses the strategic decision to launch as a retrofit module before integrating directly into vehicle designs to reduce costs and accelerate time-to-market.
Mike Weitzman and Gene Demaitre of The Robot Report recap their attendance at NVIDIA GTC24 in San Jose and the R24 robotics conference in Odense, Denmark. They highlight NVIDIA's new Blackwell architecture and Jetson Thor system-on-chip designed for humanoid robots, alongside the industry's shift toward generative AI and foundation models for robotics. The episode also details the unique collaborative ecosystem in Denmark, featuring companies like Universal Robots and Mobile Industrial Robots, and discusses the rapid development of physical humanoid robots such as Agility Digit, Unitree H1, and Boston Dynamics' robots.
Justin Schneck, co-founder and CPO of Peridio, discusses the evolution of AI from cloud to edge, introducing Peridio's upcoming 'Avocado' OS designed to simplify embedded product development. He and host Ryan Prosser of Very explore the spectrum from cloud AI to embedded AI, noting how specialized silicon and miniaturized models are enabling practical applications like Garmin's radar-enabled bike lights. Schneck argues that while cloud AI handles massive compute, the real value lies in robust, self-initializing embedded systems that reduce hardware complexity for developers.
Thom Kirwan-Evans, co-founder of Origami Labs, discusses applying AI and computer vision to industrial robotics, drawing on his background at DSTL and Phasor Nash Consultancy. He details defense applications like the DASA autonomous last-mile resupply program, focusing on scale, speed, and separation, while highlighting the critical gap in current 'smart' systems regarding true autonomy and trust. Kirwan-Evans also explains technical challenges in low-power drone vision, specifically using Google Coral TPUs to optimize neural networks for high-speed, low-false-alarm object detection. Finally, he contrasts the near-term impact of soft robotics with the complex, emergent nature of swarm robotics and the associated safety case difficulties.
The summary, chapters and topics are generated from the episode’s audio by this directory, not written by Embedded.fm. The episode itself belongs to its publisher.