Robots Atlas>ROBOTS ATLAS

Tag: Imitation learning

From 0% to 40%: adding touch to ACT wasn't enough, predicting it was
Aktualności27 października 2026

From 0% to 40%: adding touch to ACT wasn't enough, predicting it was

ROBOTIS compared three manipulation policies on the AI Worker robot. Vision scored 0%, touch as input 10%, and predicting future touch 40%.

Sharpa's trio: the D01 humanoid, W02 hand and AE01 data glove
Aktualności19 października 2026

Sharpa's trio: the D01 humanoid, W02 hand and AE01 data glove

Sharpa brought the D01 humanoid, the 21-DoF W02 hand and the AE01 data glove to IROS 2026. Together they are not a parts catalogue but a ready-made chain for collecting the contact data that trains policies.

LIFT teaches VLA models force feedback only in post-training
Aktualności2 października 2026

LIFT teaches VLA models force feedback only in post-training

Shanghai Jiao Tong University researchers presented LIFT, a method that gives a pretrained VLA policy contact-force reactivity only in post-training, with no force labels in pretraining.

Toyota Plans 400,000 Factory Robots, Trained by Veteran Workers
Aktualności30 września 2026

Toyota Plans 400,000 Factory Robots, Trained by Veteran Workers

Toyota estimates that automating its factories will need 400,000 robots, with spending of a trillion yen a year from 2028. The wheeled ELEY is being taught by 18,000 experienced workers at 60 factories.

Psi-R2.5 Teaches a Robot From One Demo Instead of Fifty
Aktualności29 września 2026

Psi-R2.5 Teaches a Robot From One Demo Instead of Fifty

PsiBot unveiled Psi-R2.5 — a model shown a manipulation task once rather than dozens of times. The company reversed its pipeline, generating human data from robot data rather than the other way round.

Skild AI Unveils S1: A Robot Learns a Task From One Video
Aktualności3 września 2026

Skild AI Unveils S1: A Robot Learns a Task From One Video

Skild AI has released S1, a robot foundation model that picks up a new task from a single video, with no post-training. It hits 66% success on unseen tasks versus 9% for language-prompted policies.

GEN-1.5: a robot learns a new task from one demo
Aktualności26 sierpnia 2026

GEN-1.5: a robot learns a new task from one demo

Generalist AI unveiled GEN-1.5 — a robot foundation model that learns a new task from a single 3–12 s demo, with no fine-tuning. One-shot yields 59% average success, few-shot reaches 83%.

ACT-2 folds laundry at 99.1% success with zero per-home data
Aktualności10 sierpnia 2026

ACT-2 folds laundry at 99.1% success with zero per-home data

Sunday Robotics unveiled ACT-2, a manipulation model that folded laundry across 25 unseen homes at 99.1% success, with no per-home data or fine-tuning.

Reimagine Robotics: robots that learn on the factory floor
Aktualności4 sierpnia 2026

Reimagine Robotics: robots that learn on the factory floor

A startup founded by four ex-Google DeepMind engineers left stealth. Its robots learn tasks directly from workers, no programmers required — cutting prototype build time from a day to 10 minutes.

Tacta Systems debuts a touch-sensing dexterous robot hand
Aktualności30 lipca 2026

Tacta Systems debuts a touch-sensing dexterous robot hand

Palo Alto startup Tacta Systems came out of stealth on July 28, 2026 with TactaBot — a 15-joint robotic hand with dense tactile sensing, a data-collection glove and an AI model. The target is precision manufacturing assembly, shipping in early 2027.