This paper presents a hierarchical control framework using model predictive control (MPC) and reinforcement learning (RL) for active roll control to manage lateral load transfer during autonomous racing of a wheeled quadruped. The framework integrates offline time-optimal raceline generation, an online MPC planner that actively minimizes the lateral Load Transfer Ratio (LTR), and a low-level, whole-body RL policy deployed directly onto the robot's 16 actuators. Physical experiments show that active roll control reduces mean LTR by up to 44%, improves fastest lap time by 8.7%, and boosts peak lateral acceleration by 21.3% to 1.98 m/s², maintaining robust high-speed stability.

NavIsaacLab, a framework built on Isaac Lab, enables physics-based and photo-realistic simulations of pedestrians and scenes for benchmarking human-aware robot navigation. It leverages GPU parallel simulation and data-driven pedestrian models (trajectory diffusion + adversarial motion learning) to overcome the scarcity of diverse, high-quality scenario data, providing a robust benchmark for navigation algorithms.

This paper introduces TaskNPoint, a training protocol that enables humanoid robots to learn dynamic skills from a single human demonstration and under an hour of GPU training. By focusing on a critical interaction window, the protocol successfully taught a Unitree G1 humanoid to perform tennis strokes, soccer kicks, and pick-and-place tasks without per-task reward tuning.

RoboTales is a low-cost robotic storytelling system that animates narratives using expressive sock puppetry. Implemented autonomously on a Baxter robot as a test case, RoboTales synchronizes narration, gestures, and mouth movements to perform character-driven stories. In a pilot study, puppet-based storytelling outperformed a gesture-only mode, producing higher HRIES ratings and improved story recall, suggesting that embodied puppetry enhances engagement and narrative comprehension. Designed to be modular and platform-agnostic, RoboTales can be adapted to other manipulators and offers a screen-free alternative to passive media, supporting future deployment in child-centered learning environments.

OmniContact is a hierarchical framework using contact flow (CF) representation to chain meta-skills for long-horizon humanoid loco-manipulation. Low-level CF-Track learns a unified skill library, while high-level CF-Gen synthesizes future contact flow sequences. Experiments achieve 98.7% on Carry Box and 76.5% on Push-Stack Boxes, outperforming baselines by 40.9% (meta-skill) and 66.5% (chaining). The framework integrates with VLMs for semantic task decomposition.

This paper presents the first morphology-specific closed-loop task-space control framework for logarithmic-spiral continuum arms. Using a segmented tendon-driven model and online Jacobian error compensation (Broyden update and Kalman filter), it achieves accurate robust control, outperforming piecewise-constant-curvature methods in simulations, and enables manipulations like grasping and obstacle-assisted motions.

This paper introduces LiMoDE, a two-stage learning scheme using Mixture of Dynamic Experts for lifelong robot manipulation. It first learns prior knowledge via multi-task pre-training with dynamic MoE, then adapts to new tasks with a lifelong MoE mechanism. Experiments show superior performance on simulation and real-world tasks.

This paper proposes RMTL (Reinforced Micro-task Learning), which decomposes long-horizon manipulation tasks into language-described micro-tasks and trains an agent to switch between them. Using multi-view VLM rewards, reverse curriculum, and a hierarchical policy, RMTL provides more informative reward signals than single-prompt VLM rewards, enabling faster learning. Experiments on the Fetch manipulation environment validate its effectiveness.

Researchers developed a physically grounded simulation of a blood capillary network, training deep RL agents to navigate via chemotaxis. They systematically mapped the physical limits of navigation, discovered a forbidden regime, and observed agents independently discovering multiple universal strategies. Without retraining, agents perform targeted blocking and unblocking of capillary flow, restoring throughput to healthy baseline levels.

A new fully unsupervised method, VMTAD, uses transformer architecture and a memory module to detect obstacles in dynamic agricultural scenes in real-time. It achieves state-of-the-art performance on a rapeseed dataset with 0.973 detection and 0.997 segmentation AUC, and a lightweight variant runs in 14 ms.

A study audits video, image, and audio deepfake benchmarks using linear probes on frozen self-supervised representations, finding that general-purpose representations can approach bespoke detector performance, suggesting benchmarks may reward general modality understanding rather than forensic skills.

This paper proposes a differentiable architecture search method to automatically discover the optimal fusion scheme for combining image and prompt tokens in visual prompt tuning. The approach jointly optimizes learnable prompts and their fusion mechanisms, introducing affine transformation and cross-attention as new fusion schemes. Experiments on 34 datasets demonstrate consistent improvements over baselines, revealing that hybrid fusion better leverages layer semantics in Vision Transformers.

Researchers introduce the Turbid Underwater Baseline (TUB) dataset and a new metric, PCD, to quantify information loss in underwater scenes with extreme turbidity. PCD correlates strongly with instance segmentation performance, outperforming common metrics.

GeMoE models token routing as a Minimum Description Length problem, using gating entropy to adaptively select experts, achieving 99.5% performance retention while increasing expert activation sparsity by 36.5%.