Hardware Podcasts

Episode Forty-Seven: Helmut Hauser

Robot Talk · 5 May 2023 · 33m · Episode 47

AI in practice: one real use of AI described

What this episode covers

Helmut Hauser, an Associate Professor in Robotics at the University of Bristol, explains morphological computation, where robot bodies exploit physical properties for sensing and control. He discusses bio-inspired examples like swimming dead trout and erodium seeds, and details a simulation project where a frog robot grows from a tadpole to optimize reinforcement learning. Hauser also describes using 3D-printed spider webs for flow sensing and the potential of growing robots using biological cells.

AI in practice

Where this conversation describes AI being used to build or deliver a product. Part of the AI in practice tracker.

Engineering & simulationreinforcement learningmuch quicker than if you learn from scratch

The team used reinforcement learning to optimize control parameters for a virtual frog robot that grew from a tadpole stage, resulting in faster learning times and more robust controllers compared to learning from scratch.

“And we did it in simulation, so we had discrete steps between different growth stages. We use reinforcement learning and always use the previous kind of solution for the next kind of morphological structure. And what we were able to show was that if you sum up all the learning time that you need in order to learn with reinforcement learning, actually it's much quicker than if you learn from scratch in the full robots.”

Helmut Hauser · Hear it at 10:21

Most relevant to

early-stage companies

Topics

Mentioned

Companies
University of Bristol, Bristol Robotics Laboratory, Lauder Lab, Oxford

Related episodes from other shows

Other shows on the same ground, closest first.

  • EE Times Current · 39m

    Understanding Real Brains with Robotic Rats, and Vice Versa

    Dr. Mitra Hartmann, a professor at Northwestern University, explains how modeling rat whisker mechanics and neural processing informs both neuroscience and neuromorphic engineering. She details the development of a 3D physics-based simulation and soft robot to study how rats localize objects through whisker bending and skin stretch, addressing the challenge of closed-loop sensorimotor control. Hosts Julia D’Angelo and Ralph Etienne-Cummings discuss how these embodied systems validate circuit hypotheses and highlight the reciprocal benefits of neuromorphic hardware, such as Reed Harrison’s neural recording chips, for advancing biological research.

    Mechanical engineeringAI & machine learning
    7 chapters

  • The Next Byte Podcast · 22m

    35. The Feisty Crustacean That Inspired A New Generation of Super Robots

    Hosts Daniel Scott Mitchell and Farbod Moghaddam discuss biomimicry in robotics, starting with a Harvard team's creation of a 1.5-gram robotic actuator that mimics the mantis shrimp's two-latch mechanism to achieve 60 mph acceleration. They then cover research from Carnegie Mellon University's Biohybrid Organic Robotics Group (BORG), led by Professor Victoria Webster Wood, where PhD student Wen Huan developed tunable collagen tendons using electrocompaction and AI to create biodegradable, biohybrid robots for environmental safety.

    Robotics & autonomyMaterials & manufacturing processes
    6 chapters

  • EE Times Current · 44m

    Neurons Close the Loop from Insect Perception to Action

    Professor Barbara Webb of the University of Edinburgh explains how she uses physical robots to validate neural mechanisms in crickets, ants, and bees, demonstrating that biological cognition requires complete sensory-motor loops. She details her work on cricket sound localization using analog electronics for precise timing, bee path integration via the central complex, and ant grasping mechanics using passive hairs. The discussion highlights the value of embodied robotics in testing biological hypotheses and inspiring efficient, low-power robotic control strategies.

    Mechanical engineeringEmbedded firmware
    8 chapters

  • The Next Byte Podcast · 23m

    75. Size Matters: Tiny Biomimicry Leads to Big Breakthroughs

    Hosts Daniel and Forbode discuss biomimicry breakthroughs from the University of Washington and Harvard, focusing on dandelion-inspired sensors and microscale robotic cilia. The University of Washington team created 24-milligram solar-powered sensors using Microchip's A20 Tiny processor and Kapton parachutes to deploy environmental monitoring networks via drones. Meanwhile, Harvard researchers developed self-propelled cilia from photo-responsive liquid crystal elastomers, enabling complex, programmable motion for applications in drug delivery and physical encryption.

    Prototyping & iterationSensors & measurement in products
    8 chapters

  • The Next Byte Podcast · 17m

    233. Reading a Fish’s Mind to Build Smarter Robots

    Researchers from EPFL and Duke University reverse-engineered zebrafish navigation by using transparent larvae to map live neural activity, creating a simulation called SimZFish that matched real fish behavior. They validated this model by building an 80-centimeter robotic fish that uses only vision inputs to maintain position in a current, proving that complex sensory systems are unnecessary for this specific task. The team made their simulation tools and findings freely available to the scientific community to encourage further cross-disciplinary collaboration in bio-inspired robotics.

    Robotics & autonomySensors & measurement in products
    8 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.