S10E223: Robots on Site - Why Humanoids Won't Build Your Project with Eliot Dixon

Every robotics demo looks the same: a humanoid doing kung fu, a humanoid dancing, a humanoid folding laundry. Eliot Dixon has spent his career in the parts of robotics that never make the showreel, and his view is blunt. The humanoid form factor exists because human environments are a compromise, not because it is the best way to get work done. Eliot's route into robotics started on a farm and a deep dislike of tractors. Physics, then a robotics degree, then years at Dstl working on sensing and UAVs, then Tata Motors building an autonomous vehicle for UK roads that was ultimately headed for the chaos of Indian ones. From there into ag tech, leading robotics and AI at Agri-EPI Centre, and now into infrastructure and construction. That path gives him an uncomfortable message for our sector. Agriculture has already run the experiment. Brilliantly engineered robotic systems, well funded, technically successful, and commercially dead, because they demanded that the entire farm change around them. As Eliot puts it, you are not designing a robot. You are designing a farm. Swap in construction and the warning lands the same way. So what actually works on a site today? Inspection and monitoring. Quadrupeds running nightly LIDAR scans, compared against the BIM, feeding tomorrow's work plan instead of a six-monthly architect's report. And the economics are shifting fast: platforms that cost well over £100,000 are now available at a few thousand for smaller sites. We also get into why drones are effectively grounded by UK airspace rules, why robotics almost never reduces headcount but changes the job mix instead, where the real return on investment hides, what data poisoning and teleoperated "autonomous" systems mean for trust, and why a dynamic construction site is a genuinely harder problem than a road in Delhi. **In this episode:** - Why the humanoid is a compromise, not an endpoint - The agricultural robotics failures construction is about to repeat - Inspection and monitoring: the only mature site use case right now - Nightly LIDAR against BIM, and the decision loop it creates - The collapse in quadruped pricing and what it unlocks - Why UK airspace, not technology, is the drone blocker - Where robotics ROI actually comes from when nobody loses a job - Autonomy 2.0, visual models, and the behaviour problem - Strawberries, dexterity, and why touch feedback is so hard - Eliot's closing advice for anyone briefing a roboticist **Guest:** Eliot Dixon, Roboverse Reply UK**Hosts:** Dale Foong and Val Matthews The Project Chatter Podcast is free to access and always will be. If you get value from it, pay it forward and share this episode with someone who needs it. **Connect:**Eliot Dixon: https://www.linkedin.com/in/eliot-dixon/Project Chatter: https://www.linkedin.com/company/projectchatterpodcast/?viewAsMember=trueWebsite: https://projectchatter.com/