Hardware Podcasts

Episode Thirty: Ana Cavalcanti

Robot Talk · 16 Dec 2022 · 24m · Episode 30

What this episode covers

Professor Ana Cavalcanti from the University of York and the RoboStar Centre of Excellence discusses the challenges of verifying software for autonomous mobile robots. She critiques the industry's reliance on code-centric development and simulation, arguing for model-based engineering to ensure safety and trustworthiness in service robots. Cavalcanti explains how her team uses mathematical foundations to verify system properties before coding begins, addressing the multidisciplinary nature of robotics.

Most relevant to

established companies

Topics

Mentioned

Companies
University of York, Dyson
Also
RoboStar, Royal Academy of Engineering, Trustworthy Autonomous Systems Hub

Related episodes from other shows

Other shows on the same ground, closest first.

  • The Robot Report Podcast · 1h 09m

    AI's Role in the Future of Robotics: Insights from 3Laws

    Andrew Singletary, CEO, and Amir Sharif, COO of 3Laws, discuss their company's evolution from academic research at Caltech to commercializing safety software for robotics. They explain how their core technology, control barrier functions, provides a deterministic safety layer that safeguards AI-driven physical systems, contrasting this with traditional heuristic-based safety. The conversation covers their product lines, Supervisor Pro and Supervisor for ROS, and how these solutions help companies achieve certification and reduce costs by enabling the use of non-safety-rated sensors.

    Embedded firmwareAI & machine learning
    7 chapters

  • Embedded.fm · 1h 19m

    460: I Don't Care What Your Math Says

    Dr. Greg Wilson, founder of Software Carpentry and author of 'Software Design by Example', argues that the software industry lacks the empirical rigor and professional standards found in medicine or mechanical engineering. He discusses the 'It Will Never Work in Theory' project, which synthesizes scientific literature on software development to debunk myths like the efficacy of Test-Driven Development (TDD). Wilson advocates for teaching computer science students empirical methods and establishing liability standards for software to ensure safety and accountability.

    Careers & educationDebugging & failure analysis
    9 chapters

  • Embedded.fm · 1h 10m

    473: Math Is Not the Answer

    Philip Koopman, a professor at Carnegie Mellon University and author of 'Understanding Checksums and Cyclic Redundancy Checks,' explains why empirical analysis and engineering trade-offs are superior to pure mathematics for selecting error detection algorithms. He details the critical differences between checksums and CRCs, highlighting how modulus choices (like mod 255 vs 256) drastically impact fault detection and how CRCs rely on XOR-based long division over Galois fields. Koopman also discusses the state of autonomous vehicle safety, arguing that statistical machine learning approaches struggle to meet the ultra-high reliability standards required for safety-critical systems.

    Embedded firmwareDebugging & failure analysis
    7 chapters

  • Embedded.fm · 1h 23m

    99: You Can Say a Boat (Repeat)

    Andrei Chichak, a systems developer at CBF Systems, discusses the MISRA-C coding standard, its origins in the British automotive industry, and its application in high-reliability embedded systems. He explains how static analysis tools like PC Lint help enforce rules to prevent bugs, citing the Apple 'goto fail' incident as a cautionary tale. The conversation also covers the ethical responsibilities of engineers, contrasting the regulated professional engineering title in Canada with the lack of formal ethics training in many US computer science programs.

    Embedded firmwareDebugging & failure analysis
    8 chapters

  • The Next Byte Podcast · 1h 02m

    192. Karsten Roscher & The Need For Safe AI (Interview)

    Karsten Roscher, Head of the Certifiable AI department at the Fraunhofer Institute for Cognitive Systems (IKS), discusses the critical need for safety in AI development, particularly in autonomous driving. He explains the challenges of out-of-distribution detection, where models fail on unseen data like e-scooters or horse carriages, and introduces Robuscope, a free tool for evaluating model robustness without sharing proprietary data. Roscher also critiques the industry's over-reliance on benchmarks and highlights a new SAE International course designed to bridge the gap between machine learning engineers and safety experts.

    AI & machine learningDebugging & failure analysis
    6 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.