The android robot Andrea was deployed autonomously for six days in a German museum, engaging visitors in multilingual conversations. Researchers tested three conditions: no emotion simulation, ChatGPT 4.1-driven emotions, and WASABI emotion simulation. Statistical analysis of visitor feedback found that neither form of emotion simulation significantly improved visitor ratings or was consciously detected by participants.
Why it matters: The findings indicate that current emotion simulation techniques in humanoid robots may not meaningfully enhance user acceptance or experience in real-world public settings.
Retriever is a new framework for building long-horizon robot agents, spanning an asynchronous decision model, programming model, runtime, and a sample closed-loop agent pipeline. It represents agents as graphs of stateful causal stream functions executed on explicit run clocks, which enables systematic debugging and deterministic replay. The system is evaluated through a real-robot case study and controlled studies of runtime overhead and replay behavior.
Why it matters: Retriever offers a unified solution for composing closed-loop robot systems with components running at different clocks and variable latency, improving reproducibility, debugging, and reuse.
VIDAR is a visual-inertial dense reconstruction framework that integrates SVO+IMU odometry with the Depth Anything 3 foundation model. The system uses visual-inertial odometry to provide camera poses, scale, and a consistent world frame, enabling accurate alignment of dense geometry predictions from the foundation model. On the EuRoC dataset, pose injection reduces scale error to about 1% and achieves a mean F@0.10 of 0.463, while a decoupled hybrid approach improves this to 0.676 without relying on ground-truth poses.
Why it matters: VIDAR offers a practical solution for achieving metric dense monocular reconstruction by combining visual-inertial odometry with a geometric foundation model.
Researchers introduce SAGE, a generative engine designed for socially-aware navigation among heterogeneous multi-agent teams. SAGE uses a Heterogeneous Graph Transformer to model asymmetric interactions and a diffusion-based generative module for joint trajectory prediction and planning. A training-free safety-social energy guidance mechanism refines robot trajectories to enhance safety and social compliance. Experiments on real-world and synthetic datasets show that SAGE reduces collision and social-violation rates and scales to teams of up to 20 robots.
Why it matters: This work presents a scalable approach to safe and socially compliant robot navigation in complex, multi-agent environments, addressing key challenges in real-world deployment.
A new method called Value-Aware MARO is proposed to enhance multi-agent coordination when communication between agents is unreliable. By weighting prediction loss with advantage estimates, the approach focuses learning on high-return behaviors, helping agents maintain performance even as communication failures increase. Experiments in simulated environments show that Value-Aware MARO prevents the performance collapse seen in standard methods under high communication dropout, achieving over 20% higher mean returns and a 64.7% reduction in variance compared to baselines.
Why it matters: This work offers a practical advance for multi-agent systems operating in real-world conditions where communication is often unreliable, improving robustness without extra hardware.
Researchers introduce SPIN, a framework that models multi-agent communication topologies as compressed tensor networks to address the combinatorial scaling problem in decentralized swarm control. By factorizing joint policy tensors into matrix product state chains, SPIN achieves linear scaling in clique length and removes the need for online training or scenario-specific optimization. Simulation experiments show that SPIN functions as a reusable decentralized coordination layer, with notable gains in multi-goal coordination and dense local interaction regimes.
Why it matters: SPIN could enable scalable, reusable swarm control for applications like drone swarms and multi-robot systems without retraining or scenario-specific optimization.
Researchers introduce PRIME, a method to address neuron dormancy in multi-agent reinforcement learning under non-stationary conditions. PRIME extends the bidirectional Silent Neuron framework to cooperative settings, reinitializing only neurons that are both activation-dormant and gradient-silent. In a UAV emergency communication simulator, PRIME improves interquartile mean return by 24.9% over MAPPO and reduces dormant neuron fractions to 10–20% compared to 40–45%.
Why it matters: This work offers a principled solution to preserving learning capacity in multi-agent systems facing changing objectives, with implications for robotics and autonomous networks.
Researchers introduce SaaF, a novel 3D language field based on Gaussian Splatting, designed to improve interactive object retrieval in real-world scenes using natural language. SaaF addresses limitations of prior methods by employing metric learning to enhance instance discrimination and by training on multiple text labels, including ambiguous descriptions, to better handle ambiguous queries. Experiments show that SaaF achieves higher retrieval accuracy and can robustly detect and manage ambiguity in user queries.
Why it matters: This work represents a meaningful advance in enabling service robots to more accurately and interactively retrieve objects in complex environments using natural language, even when queries are ambiguous.
Researchers have introduced xperception, a zero-shot 6D pose estimation system for robotic grasping that leverages CAD models and foundation models such as DINOv2 and GeDi. The system achieves millimeter-accurate pose estimation without object-specific fine-tuning or data annotation, and demonstrates robustness to occlusions in industrial tasks like bin picking. xperception is validated at TRL 6 and is engineered for deployment on industrial edge hardware, including NVIDIA Jetson Thor.
Why it matters: This approach could streamline robotic automation in flexible manufacturing by removing the need for retraining or data collection when new objects are introduced.
Researchers propose incorporating depth as a geometric prior during training of JEPA world models, alongside an isotropy-inducing latent regularizer (SIGReg), to improve learning from complex real-world robot video data. Their approach achieves a 33% reduction in visual odometry probe error, enhances surprise detection, and improves rollout fidelity on outdoor datasets, all without increasing inference time. The method demonstrates improved generalization and transferability of compact world models in challenging outdoor environments.
Why it matters: This work demonstrates that introducing a lightweight geometric prior during training can significantly enhance the generalization and practical utility of world models for real-world outdoor robotics without added inference cost.
A new preprint introduces RT-SHCUA, an architecture designed to enable real-time, secure use of self-hosted computer-use agents (SHCUAs) for controlling unmanned aerial vehicles (UAVs). The system addresses the mismatch between language agent reasoning and the strict timing, safety, and security requirements of UAV operation by separating semantic reasoning from onboard execution. It enforces contract-bound skill invocations with explicit checks for timing, state, and authority, and supports degraded handling and auditable evidence preservation. Prototype evaluation demonstrates bounded responsiveness and trusted enforcement without embedding the full agent in critical control loops.
Why it matters: This work proposes a novel architecture to safely and securely bridge language agents with real-time UAV control, addressing key challenges in timing, authorization, and accountability.
Researchers have developed a Generalist Controller that leverages attention mechanisms and a mixture-of-experts neural architecture to control a wide variety of single-input single-output (SISO) dynamical systems. Trained on over 314,000 demonstrations from 25 different systems—including stable, unstable, linear, and nonlinear cases—the controller matches the performance of system-specific LQI controllers and generalizes to new, unseen operating conditions. This approach enables a single neural network to adaptively control systems with different orders and dynamics without architectural changes or system-specific tuning.
Why it matters: This work demonstrates a significant advance toward universal AI controllers that could streamline and unify control system design across multiple engineering domains.
BayesContact is a Simulation-Based Inference framework that combines depth and force/torque observations to estimate object pose during peg-in-hole insertion tasks. The method maintains a particle belief over pose, using a renderer and physics simulator to score hypotheses against real observations. In both simulated and real-robot experiments, BayesContact improves pose observability and increases insertion success by 30% compared to vision-only inference.
Why it matters: This work offers a practical approach to fusing vision and tactile sensing for precise pose estimation in contact-rich robotic manipulation, potentially reducing the need for retraining across different environments and geometries.
Researchers have introduced Handroid, a 27-degree-of-freedom (DoF) reconfigurable robot capable of switching between a dexterous hand and a desktop-scale humanoid form (0.33 m tall, 2.05 kg). The platform supports teleoperation, dexterous grasping, in-hand manipulation, humanoid locomotion, and long-horizon tasks that involve changing its embodiment. Handroid is validated through real-world manipulation, reinforcement-learning-based locomotion, and tasks requiring reconfiguration and coordinated actions.
Why it matters: Handroid enables unified research on dexterous manipulation and humanoid mobility within a single, compact, and reconfigurable platform, advancing the field of morphology-reconfigurable robotics and cross-embodiment learning.
Researchers introduce Orbis 2, a hierarchical world model for autonomous driving that separates future prediction into two levels: a high-level predictor for coarse scene structure over long time horizons and a low-level generator for detailed predictions. The model is trained in two stages, first with diffusion forcing pretraining to enhance internal representations, followed by teacher forcing fine-tuning for stable rollouts. Orbis 2 achieves state-of-the-art results on standard driving world model benchmarks, including long-horizon generation fidelity and steering responsiveness.
Why it matters: This work demonstrates a significant advance in autonomous driving world models by combining long-horizon spatial reasoning with high perceptual fidelity, leading to improved performance and internal representations.
A new preprint introduces DPNeXt, a lightweight multi-scale feature fusion decoder designed to replace the standard Dense Prediction Transformer (DPT) for multi-task dense prediction in robotics. DPNeXt employs dual depthwise separable inverted bottlenecks and a Multi-Task Boundary Guidance strategy to improve efficiency and mitigate negative transfer between tasks. The method reduces trainable parameters by 78.6% compared to DPT and achieves state-of-the-art results on Cityscapes and NYUv2 benchmarks, with faster inference on resource-constrained hardware.
Why it matters: This work advances efficient multi-task perception for robotics, enabling real-time scene understanding on devices with limited computational resources.
SLAC is a reinforcement learning method that uses a low-fidelity simulator to pretrain a task-agnostic latent action space through unsupervised skill discovery, enabling efficient and safe real-world learning for high-degree-of-freedom robots. The approach achieves state-of-the-art performance on bimanual mobile manipulation tasks, learning contact-rich, whole-body behaviors in under an hour of real-world interaction without demonstrations or hand-crafted priors.
Why it matters: SLAC advances the feasibility of real-world reinforcement learning for complex robots by combining simulation-based pretraining with safe, sample-efficient real-world learning.
Researchers introduce a method that focuses dense data collection on the critical, contact-rich segments of manipulation tasks, while using traditional planning for simpler free-space motions. By combining automated data collection with offline deep reinforcement learning, their approach achieves a 96% average success rate across four real-world tasks using only 2–2.5 hours of autonomous data, significantly outperforming the strongest baseline at 55%. The method also maintains high performance in out-of-distribution scenarios where end-to-end approaches typically struggle.
Why it matters: This work shows that targeted data collection can greatly improve the efficiency and generalization of robotic manipulation, reducing reliance on large datasets and teleoperation.
This paper presents an embodied active learning method that adapts object detectors to unknown environments while operating under constraints on both robot navigation time and annotation budget. The approach leverages spatial inconsistency to select informative trajectories and images, targeting failure cases to improve model performance. Experiments in both simulated (AI2-THOR) and real-world (Boston Dynamics Spot robot with YOLOv5) settings demonstrate that the method achieves higher detection accuracy than baseline approaches under the same resource constraints.
Why it matters: The work offers a practical solution for efficiently adapting robotic vision systems to new environments with limited resources, potentially improving real-world deployment of autonomous robots.
Researchers have introduced IMBench, a new benchmark designed to evaluate intuitive manipulation in robots by integrating perception, physical reasoning, action generation, and iterative execution. IMBench features 35 tasks and 14,000 filtered trajectories, requiring models to infer physical structure and generate feasible action sequences under explicit constraints. Initial experiments show that current vision-language and vision-language-action models struggle to combine reasoning with executable plans and to generalize across diverse manipulation scenarios.
Why it matters: IMBench highlights a significant gap in current AI and robotics systems' ability to integrate reasoning with action execution, which is crucial for developing more adaptive and capable robotic manipulation.