Share story
Robotics Science

Biped control study reports higher obstacle-navigation success in simulation

Researchers report that a two-level learning system improved a simulated biped's ability to reach goals while avoiding obstacles, according to an arXiv preprint. Across 100 evaluation trials per method, it succeeded in 98.0% of static-obstacle trials and 88.0% of dynamic-obstacle trials, compared with at most 78.0% and 68.0% for three planner-based alternatives. A higher-level controller uses robot position and obstacle measurements to issue movement commands; a lower-level controller translates those commands into joint targets. Both learn together. The robot has four actuated joints per leg and no hip or ankle roll. Evaluation took place in randomized PyBullet simulation environments, with path lengths within 4% of the A* reference.
Sources
arXiv Query: search_query=cat:cs.RO&id_list=&start=0&max_results=30
Published by Tech & Business, a media brand covering technology and business. This story was sourced from arXiv Query: search_query=cat:cs.RO&id_list=&start=0&max_results=30 and reviewed by the T&B editorial agent team.
Back to Newswire
Keep reading
Full wire
Science Robotics
Science Robotics

Researchers improve robot task success with navigable video demonstrations

Researchers report a training-free method that improves robot task success by letting an AI agent consult selected parts of a demonstration video. In their preprint, Recursive Video In-Context Learning raises success from 92.6% to...

arXiv Query: search_query=cat:cs.RO&id_list=&start=0&max_results=30
Science Robotics
Science Robotics

Researchers demonstrate transferable visual attack on robot control models

Researchers report that a fixed visual patch can defeat multiple robot control models without querying the target policy. Their TAPDreamer preprint describes an attack built with a public visual encoder, which converts camera imag...

arXiv Query: search_query=cat:cs.RO&id_list=&start=0&max_results=30
Science Robotics
Science Robotics

InterMimicGen researchers expand humanoid motion training through simulation

Researchers report in a preprint that InterMimicGen expands executable humanoid motion data through repeated simulation tests and demonstrates transfer to real robots. A single tracking policy handles varied whole-body movements a...

arXiv Query: search_query=cat:cs.RO&id_list=&start=0&max_results=30
Science Robotics
Science Robotics

SimForcing researchers report gains in robot video prediction and policy learning

Researchers report in a preprint that SimForcing, a system for predicting robot videos from robot actions, outperformed the compared methods on four video-quality measures on the Bridge dataset without external embodied pretrainin...

arXiv Query: search_query=cat:cs.RO&id_list=&start=0&max_results=30