New AI method helps robots learn from the past without being trapped by it
Peer-Reviewed Publication
Updates every hour. Last Updated: 21-Oct-2026 03:16 ET (21-Oct-2026 07:16 GMT/UTC)
Robots often learn from historical experience collected under older controllers and physical conditions. When tasks, hardware or environments change, that data can become misleading. Researchers developed Occupancy-Matching Policy Optimization (OMPO), which anchors large replay buffers to a small set of recent interactions. Across locomotion, transfer, manipulation and a real-robot test, OMPO improved stability, sample efficiency and adaptation.
A new study in Nature Health suggests that delays are no longer necessary, offering the opportunity to assess whose health is most at risk, take protective measures, and test whether interventions are working while the heat season is still underway.
Bridging the gap between public health professionals, policymakers, and disease modelers was the focus of the Workshop on Public Health Policy, Risk Perception, and Infectious Disease, held Aug. 10-12, 2026, at the Essex Resort & Spa in Essex, VT.