[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"branding":3,"analytics":7,"article-new-physics-model-simulates-collisions-through-local-contacts":10,"sections":34},{"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":29,"feedback":33,"feedback_at":22,"cost_usd":33,"total_tokens":33},11060,"new-physics-model-simulates-collisions-through-local-contacts","New physics model simulates collisions through local contacts","RiCo, a new neural method, models rigid body contacts locally instead of whole scenes, slashing collision prediction errors by about a third.","A new simulator predicts how rigid objects bounce, slide and collide by tracking only the patches of surface that are actually touching, not the whole scene.\n\nResearchers built the system, called RiCo (Rigid-body Contact Reasoning), to fix a weakness in existing physics world models: most of them try to reason about every object in a scene at once, even though contact forces only ever pass between surfaces that are physically touching. RiCo instead looks at small neighborhoods of contact points, factoring in each point's position, motion and material properties, then lets information flow across an object to figure out how those local touches add up to overall movement. On the MOVi benchmark, that approach cut position and orientation errors over 100 simulated frames by 31 to 35 percent and about 38 percent, respectively, compared with baseline models. It also matched real contact behavior closely, differing from ground truth by just 11 percent on how long objects stay in contact and 2.22 millimeters on how deep they penetrate. The model trained on small scenes but generalized, without retraining, to simulations containing 270 objects, and the team ran a preliminary real-world multi-ball collision test to check whether the simulated physics hold up outside a computer.\n\nRobotics and self-driving systems lean on physics predictions to plan movements safely, and sloppy contact modeling is a known soft spot. Treating contact as local, rather than scene-wide, is a reasonable response to that, the same way local attention mechanisms helped language models scale without comparing every token against every other token.\n\nStill, this is one arXiv paper with one small real-world test and a single academic benchmark. A 270-object synthetic scene is not a warehouse floor or a highway, and sim-to-real transfer claims deserve the skepticism they always get until someone reproduces them outside the lab that wrote the paper.","[\"ai\",\"robotics\",\"physics-simulation\",\"research\"]","2026-10-09T04:00:00.000Z","2026-10-10T04:17:25.486Z","2026-10-10T04:17:31.181Z","published",null,[],"ai",[24,26,27,28],"robotics","physics-simulation","research",[30],{"name":31,"url":32},"arXiv cs.AI","https:\u002F\u002Farxiv.org\u002Fabs\u002F2610.12333",0,{"sections":35},[36,39,43,48,53,57,61,66,71,76,80,85,90,95],{"name":37,"slug":24,"count":38,"latest_published_at":18},"AI",6783,{"name":40,"slug":41,"count":42,"latest_published_at":18},"Security","security",932,{"name":44,"slug":45,"count":46,"latest_published_at":47},"Policy","policy",486,"2026-10-08T22:40:11.000Z",{"name":49,"slug":50,"count":51,"latest_published_at":52},"Deals","deals",474,"2026-10-08T22:00:00.000Z",{"name":54,"slug":55,"count":56,"latest_published_at":18},"Hardware","hardware",232,{"name":58,"slug":59,"count":60,"latest_published_at":18},"Science","science",193,{"name":62,"slug":63,"count":64,"latest_published_at":65},"Consumer Tech","consumer-tech",181,"2026-10-08T23:26:35.000Z",{"name":67,"slug":68,"count":69,"latest_published_at":70},"Startups","startups",117,"2026-10-08T16:45:00.000Z",{"name":72,"slug":73,"count":74,"latest_published_at":75},"Software","software",114,"2026-10-08T17:57:01.000Z",{"name":77,"slug":78,"count":79,"latest_published_at":18},"Dev Tools","dev-tools",106,{"name":81,"slug":82,"count":83,"latest_published_at":84},"General","general",66,"2026-10-09T04:46:11.000Z",{"name":86,"slug":87,"count":88,"latest_published_at":89},"Gaming","gaming",58,"2026-10-08T20:08:45.000Z",{"name":91,"slug":92,"count":93,"latest_published_at":94},"Reviews","reviews",34,"2026-10-08T14:00:22.000Z",{"name":96,"slug":97,"count":98,"latest_published_at":99},"How-To","how-to",8,"2026-10-05T09:00:00.000Z"]