RSS 2026
PulseAugur coverage of RSS 2026 — every cluster mentioning RSS 2026 across labs, papers, and developer communities, ranked by signal.
6 day(s) with sentiment data
Advancements in minimal data training for robots are a key theme at RSS 2026 and related research
Multiple clusters highlight research into training robots with significantly less data, including techniques like leveraging human video, synthesizing data, and aligning tactile information without paired examples. This trend, also seen in the development of model Ψ₀, suggests a broader industry push to overcome data limitations in robotics.
Chinese firms to showcase embodied AI hardware ecosystems at RSS 2026
Evidence suggests Chinese firms are moving beyond sponsorship at RSS 2026 to actively present their own embodied AI hardware ecosystems. This indicates a growing competitive landscape and potential for new integrated solutions emerging from China in the embodied AI space.
RSS 2026 to feature dedicated sessions on VLA Models, Humanoids, and World Models
The RSS 2026 conference agenda indicates a strong focus on VLA Models, Humanoids, and World Models, suggesting these areas are converging and becoming central to embodied AI research. This convergence may accelerate the development of more integrated and capable embodied AI systems.
Chinese firms to announce new embodied AI hardware ecosystems within 6 months
The RSS 2026 evidence highlights a growing prominence of Chinese firms in robotics, moving beyond sponsorship to hardware ecosystems. This suggests a strategic push towards controlling key components of embodied AI development. We hypothesize that these firms will announce new integrated hardware solutions or platforms in the near future to solidify their position.
Humanoid robot foundation models to prioritize minimal real-world data training
The development of the Ψ₀ model, which trains effectively on limited real-world data by leveraging human video and smaller robot datasets, indicates a significant trend. This approach addresses a key bottleneck in robotics. We hypothesize that future humanoid robot foundation models will increasingly adopt similar data-efficient training methodologies.
-
Robotic planning lags video generation, says Georgia Tech researcher
Danfei Xu, an assistant professor at Georgia Tech and NVIDIA researcher, argues that current high-fidelity video generation models do not equate to true robotic planning capabilities. He highlights the "Video-Action Gap…
-
Robotic hand's 'morphology as computation' wins RSS Test of Time Award
At the RSS 2026 conference, the prestigious Test of Time Award was presented to a 2014 paper detailing a low-cost, compliant robotic hand known as the "RBO Hand." This award recognizes the paper's profound and lasting i…
-
Chinese researchers dominate RSS 2026 robotics conference sessions
The 22nd Conference on Robot Learning (RSS 2026) kicked off in Sydney, Australia, with a strong showing from Chinese researchers. The first day featured oral presentations on manipulation and world models, highlighting …
-
OpenDriveLab's Li Chen proposes compositional world models for safer AI policies
OpenDriveLab's Li Chen presented a compositional world model approach at RSS 2026, separating prediction and evaluation components for embodied AI policies. This decoupling aims to improve safety and inspectability by a…
-
Northwestern University unveils 'invisible' spinning drone
Researchers at Northwestern University have developed a drone named Phantom Twist that is nearly invisible to the human eye. By spinning at a high frequency of 15-25 Hz, the drone exploits the persistence of vision, cau…
-
Humanoid robot model Ψ₀ trains effectively on limited real-world data
Researchers at the University of Southern California's PSI Lab have developed a new foundation model for humanoid robots called Ψ₀, which focuses on optimizing the use of limited real-world robot data. Instead of relyin…
-
Robotics Scholar Li Hongyang Wins RSS Award, Eyes Intelligent Humanoids
Hong Kong University Assistant Professor Li Hongyang has become the first Chinese scholar to win the RSS Early Career Spotlight award, a prestigious honor in robotics. His research spans from autonomous driving, with no…
-
Embodied AI takes center stage at RSS 2026, with Chinese firms gaining prominence
The 22nd Robotics: Science and Systems (RSS 2026) conference in Sydney will feature a significant focus on embodied AI, with independent sessions dedicated to VLA Models, Humanoids, and World Models, indicating a conver…
-
Robots learn complex tasks with minimal data via new AI techniques · 9 sources tracked
Researchers are developing new methods for robots to learn complex manipulation tasks with significantly less data. Innovations include synthesizing diverse training data from single human demonstrations, aligning tacti…
-
World Models Fail Long-Horizon Tasks Due to Kinematic Imagination Flaws
Researchers have identified a key reason for long-horizon failures in world models: they tend to imagine kinematically rather than dynamically. This distinction is crucial because while kinematic imagination might remai…
-
Qianxun Intelligent's Legato enables fluid robot movements · 1 source tracked
Researchers from Qianxun Intelligent's Gaoyang Team have developed Legato, a novel training method for Vision-Language-Action (VLA) models that enables robots to perform actions with natural, smooth transitions. Unlike …
-
Tsinghua University releases GS-Playground for efficient embodied AI simulation
Researchers from Tsinghua University's AIR DISCOVER Lab have developed and open-sourced GS-Playground, a novel simulation framework designed to overcome bottlenecks in visual-centric embodied AI training. The framework …
-
Galbot releases LDA-1B world-action model and open-sources framework
Galbot, a prominent Chinese embodied AI company valued at over $2.8 billion, has introduced LDA-1B, a 1.6 billion-parameter model designed for world-action learning. This model exhibits scaling behavior with increased t…