Hardware Podcasts

Episode 164: Accelerating robot learning - Michelle Lu

Robot Talk · 2 Oct 2026 · 29m · Episode 164

What this episode covers

Michelle Lu, co-founder and CEO of Vsim Technology, explains how her company uses GPU-accelerated physics simulation to accelerate robot learning and close the sim-to-real gap. Previously at NVIDIA creating Isaac Gym, Lu details how Vsim overcomes the scalability and performance limitations of traditional CPU-based reinforcement learning frameworks. The discussion highlights specific challenges in simulating tactile sensors, deformable objects, and tendon systems for the ARIA Robot Dexterity program, aiming to enable complex dexterous manipulation tasks.

Most relevant to

early-stage companies

Topics

Mentioned

Companies
Vsim Technology, NVIDIA, ARIA, Open AI, Bristol Robotics Laboratory
Products
Isaac Gym, Shadow Hand, OptaTactile

Related episodes from other shows

Other shows on the same ground, closest first.

  • The Robot Report Podcast · 1h 35m

    The inside scoop on food manufacturing with Chef Robotics

    Rajat Bhageria, founder and CEO of Chef Robotics, explains how his company automates food assembly in high-volume manufacturing plants using AI-enabled robots with interchangeable titanium grippers. He details the challenges of manipulating deformable, sticky food items where simulation fails, requiring real-world data collection to train models for consistent portioning and handling. The episode also features Joyce Sidopoulos, co-founder and COO of MassRobotics, discussing their no-equity accelerator program for robotics startups and initiatives to support women in STEM.

    Robotics & autonomyAI & machine learning
    7 chapters

  • The Robot Report Podcast · 1h 17m

    Ken Goldberg on robotic grasping; Carbon Robotics’ new weeding robot

    Ken Goldberg and Jim Liefer of Ambi Robotics discuss their Sim2Real AI approach to robotic grasping, explaining how they trained neural networks on simulation data to handle diverse e-commerce objects without human intervention. Paul Mikesell, founder of Carbon Robotics, details the development of their third-generation autonomous weeding robot, which uses high-powered lasers and computer vision to identify and destroy weeds in row crops. The episode also covers WTWH Media's acquisition of Mobile Robot Guide and updates on the Mars 2020 Ingenuity helicopter.

    Robotics & autonomyAI & machine learning
    7 chapters

  • The Amp Hour · 58m

    Hands-on Physical AI with Kevin Cloutier

    Kevin Cloutier, North American Lead for Physical AI at a global engineering consulting firm, explains the transition from legacy, statically programmed industrial robots to modern systems using imitation learning and 'Physical AI.' He details how to build and calibrate the affordable $200 3D-printed SO101 robot arm using the LeRobot open-source framework or Intel’s Physical AI Studio, orchestrating complex robotic systems entirely on local edge hardware. Cloutier contrasts the high cost and complexity of humanoids with the practical utility of specialized arms, while discussing the democratization of AI through small language models and local inference on SOCs with integrated GPUs and NPUs.

    Robotics & autonomyAI & machine learning
    9 chapters

  • The Robot Report Podcast · 1h 05m

    Recap: Promat and GTC 2025 show highlights

    Host Steve Crowe and editors Eugene and Mike recap their experiences at ProMat and NVIDIA GTC 2025, highlighting the shift toward handling chaotic real-world environments in logistics automation. They discuss specific product demonstrations from Agility Robotics, Locust Robotics, and Boston Dynamics, while also covering NVIDIA's announcements on AI simulation and humanoid robotics at GTC.

    Robotics & autonomyAI & machine learning
    7 chapters

  • The Robot Report Podcast · 1h 11m

    Deep Dive into Arm’s Physical AI and Robotics Strategies with Drew Henry

    Drew Henry, Executive Vice President of Arm’s Physical AI business unit, explains how Arm is shifting from component-level design to system-level architecture for robotics and autonomous vehicles. He details the critical engineering constraints of physical AI, including the need for sub-millisecond latency, power efficiency measured in 'tokens per joule per kilogram,' and the integration of perception-driven and interactive-driven compute planes. Henry also discusses Arm's 'Seamless Fabric' software strategy, the industry's reliance on simulation to bridge the data gap for humanoid robots, and the move toward distributed computing architectures to handle complex sensor streams.

    Microcontrollers & dev platformsAI & machine learning
    10 chapters

The summary, chapters and topics are generated from the episode’s audio by this directory, not written by Robot Talk. The episode itself belongs to its publisher.