[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"branding":3,"analytics":7,"article-a-cheaper-way-to-teach-robots-to-walk-using-cameras":10,"sections":35},{"siteName":4,"siteTagline":5,"publisherName":4,"contactEmail":6},"The Revision","Tech news, decoded.","editor@therevision.news",{"gaMeasurementId":8,"adsenseClientId":9},"G-ZW2MV82GYR","ca-pub-8533917693782264",{"article":11},{"id":12,"slug":13,"title":14,"dek":15,"body_md":16,"tags_json":17,"published_at":18,"created_at":19,"updated_at":20,"status":21,"review_note":22,"review_notes":23,"image_url":22,"persona_id":22,"persona_name":22,"section":24,"tags":25,"sources":30,"feedback":34,"feedback_at":22,"cost_usd":34,"total_tokens":34},7017,"a-cheaper-way-to-teach-robots-to-walk-using-cameras","A Cheaper Way to Teach Robots to Walk Using Cameras","A single-GPU method combines simulated trial-and-error with gradient learning to teach a robot vision-based locomotion, then runs on real hardware.","A new robot-training method lets a single GPU teach a legged robot to see obstacles through a camera and adjust its gait on the fly.\n\nResearchers describe a system called Sampling-Guided Policy Search, which pairs two techniques that usually work against each other. One is sampling-based model predictive control: running many trial actions in simulation and picking the best one. The other is first-order policy gradients, a faster but sometimes brittle way of tuning a control policy with calculus shortcuts that can settle into odd, unintended movements. Training starts by copying the sampled actions, then alternates between resampling and short bursts of gradient tuning under randomized starting positions and physics, and it skips the rendering step in the math so the policy can train directly on simulated depth-camera images instead of relying on a separate policy that cheats by seeing the underlying simulator state. The team tested it on simulated Unitree Go2 and G1 robots across walking, obstacle crossing, crate-pushing, and two-handed carrying tasks, then moved a distilled version onto a real Go2, which used its onboard depth camera to trot, crawl, clear hurdles, and switch between those behaviors without further retraining.\n\nVision-based robot training is usually where budgets blow up: turning raw camera pixels into reliable movement typically demands racks of GPUs and elaborate simulation-to-reality tricks. Doing the whole pipeline, from simulated training through real-world deployment, on one GPU chips away at a resource barrier that has largely kept this kind of legged robotics inside well-funded labs.\n\nThe demo is still a single Go2 unit in a controlled setting, so this doesn't close the simulation-to-reality gap for every task. But trimming a normally GPU-hungry training run down to one card is the kind of unglamorous efficiency gain that tends to matter more, over time, than another flashy robot video.","[\"robotics\",\"reinforcement-learning\",\"computer-vision\",\"simulation\"]","2026-09-18T04:00:00.000Z","2026-09-19T07:26:35.644Z","2026-09-19T07:26:47.550Z","published",null,[],"ai",[26,27,28,29],"robotics","reinforcement-learning","computer-vision","simulation",[31],{"name":32,"url":33},"arXiv cs.AI","https:\u002F\u002Farxiv.org\u002Fabs\u002F2609.20575",0,{"sections":36},[37,41,46,51,56,61,66,71,75,80,85,90,95,100],{"name":38,"slug":24,"count":39,"latest_published_at":40},"AI",4114,"2026-09-19T18:33:46.000Z",{"name":42,"slug":43,"count":44,"latest_published_at":45},"Security","security",673,"2026-09-19T17:30:00.000Z",{"name":47,"slug":48,"count":49,"latest_published_at":50},"Policy","policy",345,"2026-09-19T19:22:20.000Z",{"name":52,"slug":53,"count":54,"latest_published_at":55},"Deals","deals",179,"2026-06-29T20:02:07.000Z",{"name":57,"slug":58,"count":59,"latest_published_at":60},"Hardware","hardware",156,"2026-09-19T11:00:00.000Z",{"name":62,"slug":63,"count":64,"latest_published_at":65},"Science","science",129,"2026-09-19T19:45:00.000Z",{"name":67,"slug":68,"count":69,"latest_published_at":70},"Consumer Tech","consumer-tech",99,"2026-09-09T17:27:33.000Z",{"name":72,"slug":73,"count":74,"latest_published_at":18},"Dev Tools","dev-tools",78,{"name":76,"slug":77,"count":78,"latest_published_at":79},"Software","software",75,"2026-09-10T20:41:21.000Z",{"name":81,"slug":82,"count":83,"latest_published_at":84},"Startups","startups",55,"2026-09-09T23:14:29.000Z",{"name":86,"slug":87,"count":88,"latest_published_at":89},"Gaming","gaming",43,"2026-09-10T12:18:06.000Z",{"name":91,"slug":92,"count":93,"latest_published_at":94},"General","general",42,"2026-09-18T22:35:10.000Z",{"name":96,"slug":97,"count":98,"latest_published_at":99},"Reviews","reviews",20,"2026-06-24T12:00:01.000Z",{"name":101,"slug":102,"count":103,"latest_published_at":104},"How-To","how-to",6,"2026-06-16T09:00:00.000Z"]