Humanoid soccer requires a robot to seamlessly combine agile locomotion, ball manipulation, and impact-rich contact dynamics across a continuous, dynamic behavioral spectrum. Yet existing approaches either cover only a limited set of skills, or forego motion priors entirely and produce unnatural behaviors, leaving the combination of a unified skill set and human-like motion quality unaddressed. We present soccerbot, a hierarchical reinforcement learning framework that resolves both limitations through a structured latent motion prior distilled from large-scale human soccer motion capture data. By constraining policy learning within this dynamically plausible motion manifold, a single skill-conditioned policy naturally learns to perform moving, dribbling, trapping, and kicking in a unified formulation, while inherently producing human-like motion and fluid transitions during commanded skill switching. Augmented by robustness-oriented simulation design, soccerbot transfers zero-shot to the Unitree G1 humanoid robot, demonstrating agile, natural, and robust execution of the full soccer skill repertoire under real-world conditions.
https://arxiv.org/abs/2611.06564
Coordinated whole-body control of humanoid robots remains challenging due to high-dimensional coupled dynamics and substantial physical heterogeneity across joint degrees of freedom (JDoFs). Existing multi-agent reinforcement learning methods either employ coarse agent partitions or directly apply generic heterogeneous-agent algorithms, while hierarchical controllers commonly optimize high-level planning and low-level execution independently, leading to uncoordinated motion. In this paper, we propose H^2MARL, a hierarchical heterogeneous multi-agent reinforcement learning framework with two key mechanisms. We first formulate humanoid whole-body control using a JDoF-level physical-heterogeneity-aware policy factorization, where each torque-actuated JDoF is conditioned on an explicit descriptor of its actuation, kinematic, and dynamic properties. Moreover, we develop a closed-loop policy-executor co-optimization mechanism that jointly adapts the high-level pose planner and the low-level joint executor using shared interaction trajectories, thereby reducing the distribution mismatch between planning and execution. Extensive experiments in MuJoCo and Isaac Gym demonstrate that H^2MARL consistently improves locomotion stability, gait coordination, and motion quality over homogeneous, coarse-grained, and independently optimized alternatives.
https://arxiv.org/abs/2611.06523
In the real world, mobile manipulation robots must perform complex object interactions and make multi-step decisions, often relying on task-relevant information that is no longer observable. However, existing benchmarks provide limited evaluation of such memory-dependent decision-making in mobile manipulation. To address this gap, we introduce RoboCasa-Mem, a new benchmark comprising nine mobile manipulation tasks that require agents to use interaction history at key decision points. Specifically, the tasks are grouped into two categories of memory: One is environment memory, which captures spatial locations, object attributes, and object relationships. The other is process memory, which captures single-task progress and multi-task scheduling. Particularly, we introduce Behavior-Conditioned Evaluation. For tasks with multiple valid execution paths, instead of fixing a single target or trajectory, the evaluator derives later task constraints from the robot's earlier actions. This allows us to evaluate whether later behavior is consistent with the robot's own interaction history while reducing shortcuts based on fixed target assignments. We provide 1,800 successful demonstrations for policy training. We also develop a baseline on our benchmark, namely Hierarchical Memory-Augmented Policy(HMAP). Its high-level planner uses structured entity and process memories to generate subtasks. Its low-level VLA controller uses subgoal-conditioned queries to incorporate VGGT-Omega geometry into action generation. Together, RoboCasa-Mem and HMAP baseline support systematic evaluation of memory-dependent behavior and the development of memory-aware mobile manipulation policies.
https://arxiv.org/abs/2611.06475
Vision-Language-Action Models (VLAs) enable robots to perform manipulation tasks by leveraging Vision-Language Models (VLMs) to process visual observations and language instructions. However, the action prediction objective itself lacks fine-grained task-specific perceptual supervision, making it prone to relying on task-irrelevant visual shortcuts, leading attention away from target objects and key interaction areas, thereby limiting its operational performance and generalization ability. To address this, we propose Automatic Perception-Guided Attention Alignment for Robot Manipulation ($A^3$-VLA), a framework that derives task-specific visual guidance from the manipulation process itself. Specifically, $A^3$-VLA first generates heuristic candidate regions for target objects and the robot from interaction-induced visual changes, and refines them with a visual segmentation model offline. These regions serve as explicit perceptual guidance through two complementary objectives: an attention allocation loss that encourages sufficient focus on task-critical regions when language or robot state tokens query the visual input, and a contrastive learning loss that aligns the representations of candidate target regions with language instructions while separating them from task-irrelevant regions. Extensive experiments on simulated and real-world manipulation tasks demonstrate that $A^3$-VLA consistently improves multiple VLA backbones, with an average success rate gain of 6.5% - 25.2% across benchmarks.
https://arxiv.org/abs/2611.06447
Online reinforcement learning offers an effective approach to precise laboratory manipulation, yet costly real-world demonstrations and limited access to dense process rewards hinder its practical application. We present TwinBridge, a digital twin-driven framework that supplies policy initialization and dense visual rewards for real-world online learning. We construct digital twins by combining 3D Gaussian splatting with physics simulation. Here, simulator states provide task events and geometric progress under physical prerequisites and ordering constraints, yielding rewards for policy pretraining with simulated demonstrations and reinforcement learning. We distill this supervision into a visual reward model. During real-world interaction, its predictions are combined with execution history to compute dense rewards for human-in-the-loop policy refinement. Across five real-world laboratory tasks, TwinBridge achieves a final mean autonomous success rate of 98% with at most 25 minutes of real-world interaction per task. Experiments show that dense rewards improve interaction efficiency over sparse rewards, and simulation-based policy initialization further accelerates learning.
https://arxiv.org/abs/2611.06442
Myoelectric Control Systems (MCSs) enable voluntary control of wearable robots through muscle activity. However, intention decoding, assistance shaping, and low-level actuator control are often not explicitly distinguished, making their contributions difficult to evaluate and compare. This review introduces a functional decoder-shaper-controller framework for continuous myoelectric control of upper-limb exoskeletons. The decoder estimates an intention-related variable from surface electromyography (sEMG), the shaper maps it to a control reference according to an assistance policy, and the controller realizes that reference through actuator control. This decomposition separates intention estimation from assistance-policy design and enables comparison of heterogeneous architectures. Representative approaches are critically examined in terms of module design, calibration, and evaluation. The analysis highlights that the shaper, despite determining the magnitude and shape of assistance, lacks established design and tuning methodologies and is often embedded within the decoder or controller. Key open issues include inconsistent reporting of assistance levels, limited use of interaction-aware stability and passivity tools, and insufficient validation across degrees of freedom, users, and tasks. Building on these findings, this review provides design guidelines for systematic module-wise tuning and identifies future research priorities to bridge the gap between accurate intention decoding and safe, effective, and generalizable assistance under real-world operating conditions.
https://arxiv.org/abs/2611.03418
Vortex grippers are commonly integrated into automated systems for handling objects with complex geometries in packaging, sorting, and assembling. However, their interaction with compliant objects remains difficult to predict, particularly with respect to the vortex-induced torque transmitted during grasping. This study investigates vortex-induced torque during the grasping of compliant objects and quantifies the coupled effects of supply pressure, tangential nozzle diameter, surface geometry, and curvature. The results show that increasing the tangential nozzle diameter primarily enhances the sensitivity of the generated torque to supply pressure rather than producing a uniformly higher torque. Surface geometry also strongly influences the rotational interaction, with dome-concave surfaces generally producing substantially greater torque than corresponding dome-convex surfaces, whereas the difference between cylindrical concave and convex surfaces is comparatively small. The strongest and most repeatable response was observed for a dome-concave surface exhibiting favorable object-gripper geometric matching. Among the evaluated machine-learning regression methods, Gaussian Process Regression provided the best overall predictive performance under repeated cross-validation, demonstrating that vortex-induced torque can be estimated from combined pneumatic and geometric parameters. These findings provide a basis for predicting and utilizing vortex-induced torque in the dexterous manipulation of soft objects.
https://arxiv.org/abs/2611.06359
World Action Models (WAMs) use the predictive knowledge of pretrained video diffusion models for robot control through future video prediction. However, iterative video denoising is costly, and inaccurate visual predictions can compromise action prediction. Our key idea is to make this predictive knowledge available through visual representations computed from the current observation, enabling action prediction without generating future videos. We introduce DreamFree, a WAM that repurposes a pretrained video diffusion model as a predictive visual encoder through representation distillation. The visual encoder learns to predict a frozen video diffusion model's representation using only the current observation. An action module is then trained to predict actions by extracting action-relevant information from the visual encoder. Across real-world and simulated tasks, DreamFree achieves higher success rates than state-of-the-art VLAs and WAMs, while performing real-time closed-loop control at 11Hz with a 14B video diffusion backbone. Experiments show that DreamFree can translate stronger video priors and additional video data into better robot control and enable learning new robot tasks from video demonstrations, including videos from a different embodiment, without action annotation for the new tasks.
https://arxiv.org/abs/2611.04527
Feed-forward geometric foundation models enable dense monocular reconstruction without requiring camera calibration, but independently reconstructed submaps can accumulate scale drift, and even a valid revisit can yield an inconsistent cross-submap measurement. We extend VGGT-SLAM 2.0 with fixed-root scale anchoring and short-sequence-assisted direct loop connections. SALAD selects an image pair between the initial submap and each eligible new submap; joint VGGT inference then uses learned implicit semantic-geometric cues to supply a scale-only anchor while preserving the original overlap chain. For loop closure, neighboring image pairs support a consistency check and translation-scale alignment. Each accepted candidate contributes one robust Sim(3) factor between existing center-frame nodes, without an auxiliary loop submap. A 17-sequence comparison lowers mean Sim(3) ATE on nine TUM sequences from 0.1018 to 0.0451 m relative to the VGGT-SLAM 2.0 baseline. Anchoring improves all seven no-loop module ablations. On three Clio scenes, a fixed-loop comparison favors the direct connection pipeline over the tested auxiliary-submap pipeline. A separate candidate audit exposes additional rejections of visually overlapping pairs by the short-sequence check. These comparisons distinguish the gains from long-range scale anchoring and loop connection choices from the limitations of candidate verification.
https://arxiv.org/abs/2611.06308
This paper addresses emergency-braking decisions for a redundant safety channel in automated vehicles. The channel is responsible for identifying hazards from sensor data and determining whether braking is necessary, given the constrained computational resources of embedded hardware. We present an emergency-braking approach that uses a predicted motion corridor to filter \ac{lidar} measurements and then decide whether the ego vehicle must brake. \ac{pmcm} constructs a curvature-aware boundary from the planned path and vehicle geometry so that the monitored region contains the full swept body on both straight and curved paths. \ac{gdeta} then decides whether the ego vehicle must command an emergency brake for objects inside this boundary, scoring geometric occupancy and temporal urgency independently. Hard gates permit braking when either source is conclusive; soft-OR arbitration combines partial evidence when neither gate is active. Consecutive-scan confirmation reduces sensitivity to transient measurement noise. Corridor coverage is evaluated in controlled collision simulations and recorded campus driving. Braking decisions are evaluated in Monte Carlo simulations of lead-vehicle braking and transverse-crossing encounters.
https://arxiv.org/abs/2611.06287
Scaling real-world dexterous-grasp data collection is limited by the sustained human effort required to operate the robot, verify grasp outcomes, and reset objects between trials. Simulation can generate grasp candidates at scale, but their success on hardware remains uncertain. We present AutoDex, a system that collects real-world dexterous-grasp data without continuous human operation. Given generated grasp candidates, AutoDex plans and executes grasps, labels their outcomes, and resets the object to continue collection across stable poses. Multi-view object tracking supports labeling under hand-object occlusion, while residual-torque collision monitoring and pose-directed reset enable the loop to operate without continuous human supervision. Using AutoDex, we collect over 3,500 grasp trials across 100 objects with Allegro and Inspire hands, recording both successes and failures together with synchronized visual observations and robot states. AutoDex reduces human effort by more than an order of magnitude, requiring an estimated 0.2 human-hours to collect 500 trajectories compared with 9.3 for teleoperation. As AutoDex collects trials, it builds a library of physically validated grasps for subsequent execution. In new object placements and obstacle configurations, retrieval from this library achieves 76% success, compared with 34% for candidates screened only in simulation. Code and data are available at this https URL.
https://arxiv.org/abs/2606.23689
Inspired by the coordination of global motion and local shape adaptation in snakes, elephant trunks, and octopus arms, we propose a unified planning and control framework for mobile continuum robots in cluttered and confined environments. Existing approaches often address high-level path planning and local motion control separately, limiting coordinated base-arm motion, kinematic consistency, and collision avoidance. Our framework integrates long-horizon trajectory planning with real-time geometric refinement under kinematic and environmental constraints. The global layer combines A* search with minimum-jerk planning to generate smooth reference trajectories. The local layer uses a geometry-consistent inverse kinematics solver and Riemannian optimization to coordinate mobile-base motion and continuum deformation while avoiding obstacles. We evaluate the framework through simulated pipe-traversal tasks with multiple baseline planners and real-world experiments involving navigation among dense obstacles and manipulation in confined spaces. Results demonstrate improved motion consistency, feasibility, and adaptability, supporting coordinated whole-body motion of mobile continuum robots in constrained environments.
https://arxiv.org/abs/2611.06214
Ensuring safety in advanced robot control methods such as Model Predictive Control (MPC) and Reinforcement Learning remains an open challenge. A common approach relies on control-invariant sets, yet their exact computation for nonlinear systems is generally intractable, and numerical approximations compromise theoretical guarantees. In this work, we show that safety does not need to rely exclusively on control invariance. We introduce novel classes of safe sets, each broader than classical control-invariant sets, and analyze corresponding MPC schemes that guarantee recursive feasibility by constraining different nodes of the prediction horizon rather than only the terminal state. Furthermore, when the safe set is an inner approximation of a suitable control-invariant set, our methods can ensure safety through a safe-abort strategy, even in the absence of recursive feasibility. We validate our approaches in extensive simulations on a 6-joint robot manipulator, showing fewer constraint violations than state-of-the-art methods while maintaining comparable tracking performance and computational efficiency. Hardware experiments further demonstrate the practicality of our schemes for real-world robotic applications.
https://arxiv.org/abs/2611.06196
Robotic manipulation often requires remembering information that is no longer visible, such as an object's location before occlusion or the order of a demonstrated action sequence. A large vision-language planner can read this history, but calling it before every action chunk is slow and costly. We present a three-tier system for the RoboMME benchmark. GPT-6 Astra (System 2) selects a grounded subtask from stored visual history, a $\pi_{0.5}$ vision-language-action policy (System 1) executes it, and a fine-tuned Qwen3-VL-4B monitor (System 1.5) checks after each action chunk whether the subtask is finished. The instruction is kept until the monitor reports completion or a task-specific event occurs, while the policy keeps acting on new observations. On RoboMME, the system succeeds in 79.13% of episodes with an average of 3.63 Astra calls and 80.84 seconds per episode. This is 34.62 points above the best published non-oracle baseline and 4.95 points below GroundSG+Oracle, which gives the low-level policy ground-truth subgoals.
https://arxiv.org/abs/2611.06184
This extended abstract discusses a specific and, in our experience, practically relevant role for large language models (LLMs) in underwater robotics. Our main premise is that, in underwater fielding missions, the immediate difficulty is often not the synthesis of new low-level control laws, but the operation of complex robotic systems under limited communications, partial observability, and expensive field time. In particular, a significant fraction of the effort during sea trials is devoted to fault diagnosis, telemetry interpretation, mission adjustment, and coordination across heterogeneous assets and operators. We argue that LLMs may be useful at this higher operational level, especially when grounded in structured vehicle state, alarms, logs, mission context, and existing planning or validation tools. At the same time, current limitations in grounding, consistency, and causal reasoning make these systems lack the predictability required to guarantee direct control, especially in real-world critical applications, although we do not exclude a future evolution in this direction. This viewpoint suggests a layered architecture in which conventional autonomy remains responsible for control, perception, estimation, and validated mission execution, whereas LLM-based modules serve as an additional support layer for debugging, operator interaction, constrained replanning support, orchestration, and post-mission analysis.
https://arxiv.org/abs/2611.06162
Projector-based augmented reality enables natural interaction between users and devices by overlaying digital information onto the physical world through projected imagery. A key technical requirement of such systems is accurate positional alignment between the projector and the devices. Pixel-level Visible Light Communication (PVLC) achieves this alignment without external sensing infrastructure by embedding invisible spatial codes into each projected pixel through high-speed brightness modulation that preserves image quality. However, the information that PVLC transmits is inherently planar. PVLC has therefore been limited to devices moving on a flat surface and cannot estimate the 3D pose of devices in three-dimensional space, such as aerial robots. To extend PVLC-based pose estimation to three dimensions, we formulate it as a Perspective-n-Point (PnP) problem. The 2D pixel positions that four coplanar photosensors on the device receive via PVLC serve as the image points, and their known 3D layout provides the object points. Experiments show that the proposed system estimates 3D pose with a latency of approximately 8~ms and a 3D position RMSE of 19.2~mm after Kalman filtering. These results confirm the first realization of real-time 3D pose estimation coordinated with high-quality projected image content without external sensing infrastructure.
https://arxiv.org/abs/2611.06121
Manipulating visually identical objects with different physical properties requires adapting actions based on information acquired through interaction. We address the problem of learning from demonstrations how to adapt manipulation actions, such as regrasping and changing contact locations, based on physical-property information acquired during interaction. To this end, we propose a learning-from-demonstration method that learns latent representations of physical properties needed to predict object responses and conditions a policy on online estimates of these representations. A dynamics model and object-specific latent representations are jointly learned from demonstrations that include behaviors adapting to the objects' physical properties, and the learned representations are then used to condition an imitation policy. During execution, the dynamics model and policy weights remain fixed, while the latent representation is updated based on the discrepancy between predicted and actual observations, allowing interaction with the object to inform action generation. Real-world and simulation experiments with objects whose centers of mass differ suggest that online adaptation through latent representations of physical properties can improve task success rates. However, in simulation, success rates on center-of-mass locations absent from the demonstrations remained around 30%, indicating limited adaptation to unseen physical properties. Project page: this https URL
https://arxiv.org/abs/2611.06118
Swarm robots have attracted growing interest as a physical interface for human--computer interaction, since a swarm can reshape its formation to fit the objects and surroundings it works with. For a human operator, however, adapting the swarm's behavior remains difficult. Most existing methods require the user to specify where the swarm should go, and they offer no intuitive way to change the collective's behavior as each task demands. We present Fluidic Body, a system that lets a person operate a swarm of fluid-imitating robots as an extension of the hand. The system places a virtual fluid between the hand and the swarm, and the user adapts the swarm's behavior to each task by changing the fluid's properties through simple hand movements. When the user opens the hand, the swarm becomes soft enough to deform and flow around obstacles. Closing the hand turns it into a hard, cohesive body that can push objects. We evaluated the controller's performance and its operability in two simulation experiments and one demonstration with physical robots. First, our controller handled both obstacle avoidance and object transportation better than an existing method that behaves the same way in every task. Second, in a VR user study, 17 participants completed the combined avoidance-and-transportation task faster and more reliably than with that method. They also reported a lower workload and a stronger sense of agency. Finally, this experience carried over to physical robots in a real-world demonstration (88 participants). These results suggest that treating a swarm as a fluid, a material people already understand, allows a robot swarm to work as a deformable extension of the human body.
https://arxiv.org/abs/2611.06068
Boundary-layer Sliding Mode Control (SMC) has been applied to vision-based stabilization, but identifying its activation condition against simpler causal baselines (LPF, PID) has been elusive. The operational condition is plant binding: the disturbance velocity approaches or exceeds the actuator slew limit, so the bounded per-step actuation budget binds. We validate this in two independent vision domains: handheld translation stabilization (12 1080p clips; LPF, PID, SMC, L1) and angular re-framing against a virtual slew-rate-limited pan-tilt-zoom (PTZ) plant. Outside binding, leave-one-out cross-validation makes SMC and LPF on translation statistically indistinguishable, and PID beats SMC on the angular 99th-percentile pointing offset by 5.85 degrees (95% CI clear of zero); inside binding -- reached only by sweeping the simulated slew limit, as both captured datasets land outside it -- the translation gap stays a tie and SMC beats PID on the angular worst case (Cohen's d up to +0.83). The PTZ plant's 100 degrees/s default sits on the cross-over, which explains a small-margin PID result previously reported in the angular setting. A complementary ablation shows raising the SMC reaching gain k costs mean tracking outside a narrow window.
https://arxiv.org/abs/2611.06028
Manipulating diverse objects with multi-fingered dexterous hands is challenging due to the high dimensionality and complex dynamics. Human-Object Interaction (HOI) datasets provide rich knowledge about task information and embodied interactions. Instead of solely imitating the human demonstrations, our method learns to holistically predict future hand-object states by leveraging these datasets. The predicted future states of the object can serve as a general-purpose reward term for reinforcement learning, reducing reliance on task-specific reward engineering and enhancing generalization across tasks. We conduct extensive experiments on three manipulation tasks in simulation and the real world. Our approach outperforms existing SOTA methods in both success rate and generalizability on novel objects. Furthermore, we validate the cross-embodiment compatibility of our methods by successfully deploying the skills on different robot hands.
https://arxiv.org/abs/2611.05979