Real-world robotic assembly at sub-millimeter tolerances demands spatial precision, compliant interaction, and robustness to contact failures. We present Facet-0, a robotic foundation model that predicts and values the contact consequences of its actions. Facet-0 unifies multimodal representation learning and reinforcement learning (RL) post-training around a joint action-wrench proposal: a causal wrench history is aligned with vision-language semantics and kinematic state, and flow matching generates each action chunk together with the future wrist-wrench profile it is expected to induce. Deployment rollouts train a distributional Action-Wrench Critic to distinguish motions with similar task progress but different contact outcomes, while phase-aware rewards and contact-selective credit concentrate policy improvement on decisive interactions. To accommodate part-specific dynamics, a lightweight bounded actor reuses the frozen representation for on-robot adaptation; RL remains defined over executable Cartesian actions, while an auxiliary wrench head preserves predictive, non-commanded action-contact coupling. Trained on ManuFacet-1K, a 1,000-hour force-synchronized corpus spanning three embodiments and multiple manufacturing cells, the bounded task-adapted system reaches 82% mean success on five sub-millimeter computer-assembly tasks, compared with 15% for the strongest baseline, with 0.5 mm placement accuracy and 50 ms command latency.
Embodied navigation requires agents to translate heterogeneous goals and visual observations into actions across tasks, environments, and robot embodiments. Modern vision-language models (VLMs) already encode spatial priors for visual grounding, spatial reasoning, and pointing, but these capabilities are rarely elicited directly for robot control. Existing navigation systems instead rely on task- or embodiment-specific components, fragmenting perception, reasoning, and action while offering limited generalization. Here we present LightNav-0, a compact generalist embodied navigation model that elicits the spatial intelligence of a pretrained VLM and aligns it with navigation, without task-specific prediction heads. LightNav-0 represents diverse navigation tasks through a unified token interface: dual-channel pointing expresses task-, scene-, and embodiment-agnostic spatial intent, while a residual vector-quantized action tokenizer maps this intent to precise, embodiment-specific trajectories. Together with temporally aware visual history compression, ER mid-training, supervised fine-tuning, and reinforcement learning, this formulation supports instruction following, open-vocabulary object navigation, and visual tracking within a single model. The navigation training corpus spans 2K+ scenes and 4K+ hours of embodied navigation data. LightNav-ER, the embodied-reasoning checkpoint used to initialize LightNav-0, attains the highest complete-set average across 8 embodied-reasoning benchmarks, while LightNav-0 achieves state-of-the-art monocular success rates across all 10 public navigation simulation settings. Real-world evaluations further demonstrate zero-shot generalization across robot embodiments, diverse scenes, and static and dynamic targets. These results establish compact VLMs as a unified and transferable backbone for generalist embodied navigation.
Cheng Gu, Qiusheng Zhao, Anbang Liu +2cs.LG cs.AI math.OC
Large-scale industrial robot fleets share constrained physical infrastructure, making vehicle travel times dependent on safety separation, intersection access, downstream blocking, and station contention. We study this problem in overhead hoist transport (OHT) systems, a representative ceiling-mounted material-handling system used in semiconductor fabs. Static shortest-path routing cannot account for these time-varying traffic costs, whereas tabular Q-routing adapts online but learns each destination--node--action value independently, limiting information sharing across sparsely visited routing contexts and making startup behavior sensitive to inaccurate value estimates. We propose Neural Double Q-routing, which replaces destination-indexed tables with a shared state--action value network. The network is warm-started through return-to-go regression on mixed simulator-generated routing trajectories and then refined online using Double-Q updates, local congestion correction, and event-stratified structured replay. Across nine matched fleet-size--arrival-rate settings with 100, 150, and 200 OHTs, the proposed framework reduces mean completion time relative to tabular Double Q-routing by $0.8\%$--$8.8\%$. It achieves the lowest mean completion time among all compared methods in the six 150- and 200-OHT settings, whereas Dijkstra remains best in the three 100-OHT settings. Completed-task counts remain within $1\%$ of tabular Double Q-routing in eight of nine settings, and 95th-percentile completion time decreases in eight settings. In two matched startup scenarios, offline initialization increases the number of completed tasks by up to $23\%$ and reduces tail completion time by up to $15\%$.
Robot crowd navigation requires safe and efficient decision-making under dense, dynamic, and multimodal human--robot interactions. Existing reinforcement-learning methods typically output a single reactive action at each timestep, which limits their ability to represent diverse short-term avoidance strategies. We propose Planning Diffusion Policy Optimization (PDPO), an offline-to-online reinforcement-learning framework that uses a diffusion policy to generate short-horizon action chunks for crowd navigation. PDPO is first pretrained on collision-avoidance demonstrations and then fine-tuned online with PPO by treating the denoising process as an internal decision process. During execution, the policy generates a five-step action chunk and applies it in a receding-horizon manner. Furthermore, we observe an evaluation artifact in common crowd-navigation benchmarks: without explicit boundary constraints, learned agents may leave the valid domain and bypass dense crowds. To address this, we introduce a setting in which boundary violations are treated as collisions. Experiments show that PDPO obtains an improved success rate over strong baselines, and ablations demonstrate that action chunks are especially important for the modified bounded benchmark.
Lehong Wu, Yuxiao Qu, Zheyuan Hu +4cs.RO cs.AI cs.CL cs.LG
Reasoning in language allows foundation models to spend more test-time compute on hard problems, such as those requiring decomposition, constraint tracking, and prediction of future consequences. Whether this mechanism can improve robotic manipulation remains unclear, where long-horizon tasks require tracking partial progress, reasoning about object relations, recovering from mistakes, and steering noisy low-level policies. In this paper, we study whether VLMs can be trained to reason directly in natural language to guide low-level manipulation policies. We introduce $R^3$, a simple post-training recipe that turns off-the-shelf VLMs into robotic reasoners: it first mid-trains a VLM on expert-generated reasoning traces to initialize the desired reasoning style, then improves the reasoner with single-step rubric-based RL from offline action data. Unlike prior robotic reasoning methods that mostly use structured traces as auxiliary supervision, $R^3$ trains free-form language reasoning to produce test-time guidance for action. We instantiate $R^3$ on Language Table and simulated bimanual grocery packing, two controlled testbeds for studying robotic reasoning and long-horizon manipulation. $R^3$ improves exploration and generalization across unseen tasks and significantly outperforms instruction-only imitation learning baselines on both benchmarks. Our analyses suggest that free-form language reasoning can function as a test-time compute mechanism for steering low-level policies. Our project page is available at https://robotic-reasoner.github.io/.
World models enable agents to reason about future outcomes and learn policies from their knowledge of state transition, but existing approaches primarily focus on reconstructing future observations or features, which introduces unnecessary complexity and limits their effectiveness for decision making. In this work, we propose a compatibility prediction Latent World Model (LWM) for robot navigation that predicts action-conditioned latent feature compatibility rather than reconstructing observations. Our key insight is that spatial proximity correlates with latent feature similarity, enabling action consequences to be evaluated directly in latent space. To support counterfactual training, our model leverages action sequences sampled across trajectories and learns to predict which sequences lead closer to the goal. Furthermore, we demonstrate how the learned world model can supervise policy learning from unlabeled video data and further improve policies through reinforcement learning entirely within the world model. This imagination-driven framework eliminates the need for action annotations and additional environment interaction. Extensive experiments on multiple real-world robot navigation datasets show that our approach significantly outperforms prior world model and imitation learning methods in prediction accuracy, policy learning, and real-world navigation performance. The code, pretrained models, and additional materials are available at https://wzm206.github.io/latent-world-model-nav.
Safe and efficient shape-aware navigation in heterogeneous crowds and robot fleets remains challenging. Traditional approaches often assume homogeneous robots, sparse workspaces, simplified geometry, offline computation, or handcrafted parameters to make the problem tractable, which limits their deployment in dense crowd scenarios. Toward this end, we propose Shape-Aware Reinforcement Learned Model Predictive Control (SRL-MPC), a method for safe, efficient, and adaptive navigation in crowds with heterogeneous shapes without geometry simplification. To encode shape-aware safety, we formulate high-order control barrier function (HOCBF) constraints from geometric separation features (GSFs) based on support function transformation. A reinforcement learning (RL) framework then learns a neural policy that reads GSFs and outputs real-time MPC parameter updates, enabling the MPC solver to adapt to neighboring crowd geometries. The key advantage of SRL-MPC is that it preserves the safety structure and generalizability of MPC while integrating the adaptability and intelligence of RL. Experiments in randomized crowd scenarios with arbitrary shaped robot fleets demonstrate the effectiveness, scalability, and robustness of SRL-MPC. The results show that SRL-MPC substantially outperforms representative baselines in safety and adaptability. Project website: https://hanruihua.github.io/srl_mpc_project/
Bhavya Sukhija, Oliver Groth, Mohit Shridhar +5cs.AI
How to efficiently finetune robot policies to learn new tasks on the fly? State of the art robotic manipulation policies are based on behaviour cloning of large vision-language-action (VLA) models with billions of parameters on huge teleoperation datasets. While this simple approach has enabled significant advances for robotic manipulation, finetuning of VLA policies for learning new tasks still remains an open problem. In particular, collecting teleoperation datasets requires hundreds of hours of expensive human labour and the alternative, reinforcement learning (RL), can be notoriously sample-inefficient especially for long-horizon tasks. In addition, RL with VLAs imposes several challenges due to the model's size and architectural design. In this work, we propose EXIMO, an efficient algorithm for finetuning of VLA policies. EXIMO operates in three stages: explore, imitate, and optimize. During the explore phase, EXIMO equips the VLA with a vision language model (VLM) that acts as a planner. The VLM thinks and breaks down challenging long-horizon problems into shorter ones for the VLA. The VLM, together with the VLA, is used to collect an orchestrated dataset on new tasks. During the imitate phase, the VLA is finetuned with the orchestrated data. Finally, during the optimize stage, we use residual off-policy RL to further finetune the policy. In our experiments, we ablate all three stages of EXIMO and show that it outperforms existing approaches significantly in terms of sample-efficiency and final performance.
Jun Hyuk Lee, Chihyeong Lee, Jooeun Ahncs.RO cs.GR cs.LG
The massive overactuation in the human musculoskeletal system makes it challenging to train musculoskeletal models to generate human-like motion via reinforcement learning, primarily because exploration in the resulting high-dimensional and redundant action space is extremely inefficient. To address this problem, we propose the $λ$-hold controller, inspired by the equilibrium-point (EP) hypothesis, which has been widely supported by extensive evidence from human motor control studies. The policy's control variable is the per-muscle EP threshold length $λ$, from which a stretch-reflex recruitment law computes the muscle excitations automatically. Holding each $λ$ over an interval of the gait phase also sharply reduces the frequency at which the policy must be queried. Consequently, the controller, to our knowledge for the first time, enables a muscle-actuated skeletal model to learn human-like sprinting using only a minimal reward within an hour of training. The efficient exploration through the proposed $λ$-hold controller is not merely an engineering trick but an approach grounded in physiology, bringing together the EP hypothesis, intermittent control, and optimal feedback control. Beyond encapsulating human-like behavior in predictive simulation, this achievement contributes to developing a learnable model of the human motor controller.
Humanoid robots hold great promise as general-purpose agents in human-centered environments, yet generalist vision-language-action (VLA) foundation models are not readily applicable to humanoid whole-body loco-manipulation. The high dimensionality and interdependence of humanoid motions make it challenging for conventional single-stage VLA architectures to coordinate locomotion, waist posture, and dual-arm manipulation effectively. Moreover, policies trained through offline behavior cloning can remain suboptimal during real-world deployment. Although online reinforcement learning can refine policies through real-world interaction, directly tuning large VLA backbones demands excessive computation and may introduce safety risks during real-robot exploration. To address these bottlenecks, we introduce HAF (Humanoid Adaptation Framework), a two-part framework consisting of HAF-VLA and HAF-Steer that transfers off-the-shelf generalist VLA foundation models to humanoid whole-body loco-manipulation. HAF-VLA is a hierarchical action-flow generator built on a pretrained flow-matching VLA. It splits full-body action denoising into three sequential stages with stage embeddings and cross-stage KV caches that retain kinematic dependencies, avoiding incoherent whole-body actions from one-shot generation. On top of the frozen HAF-VLA, HAF-Steer is a latent offline-to-online RL pipeline that leverages flow-matching invertibility and DCT-based dimensionality reduction to restrict RL optimization to a compact noise subspace and train a regularized SAC policy. This avoids updating the large VLA backbone and enables efficient real-world policy refinement. Evaluated on seven real-world humanoid loco-manipulation tasks, HAF surpasses vanilla single-stage VLA baselines and improves whole-body coordination and task performance. Project website: https://grange007.github.io/HAF .
Training locomotion policies for complex unstructured terrain requires a curriculum to avoid early exploration failures. However, since unstructured terrain lacks explicit difficulty ordering for curriculum design, existing methods resort to heuristic curricula over parameterized terrains. This abstraction limits generalization, as policies can overadapt to near-fixed perceptual patterns. To address this, we propose \textbf{\ourname{}}, an \textbf{T}rajectory-level \textbf{A}utomatic \textbf{C}urriculum \textbf{L}earning framework that generates training tasks directly from unstructured terrain maps. At each curriculum update, the evaluator learns a difficulty function for the current policy that maps a given trajectory task to a difficulty score. The sampler then proposes new trajectories guided by the learned evaluator as the curriculum for the next policy update. This forms a closed loop in which the curriculum is iteratively matched to the evolving policy. Quantitative and qualitative experiments show that \ourname{} continuously provides effective curricula on unstructured terrain, improving trajectory success rate by \(56.3\%\) over direct training without curriculum. Compared with handcrafted curriculum learning, our method improves success rate by \(18.5\%\) on the hardest terrain tasks and by up to \(39.74\%\) when evaluating traversal from diverse approach directions on the same obstacle type.
Vision-language-action (VLA) models improve robotic manipulation but remain vulnerable to compounding errors, scene changes, and off-trajectory states. Reinforcement learning can refine pretrained VLA policies, yet sparse success signals hinder exploration, while engineered dense rewards are costly and task-specific. Existing learned visual reward models often rely on static before-after observations, causing temporal ambiguity and weak discrimination between robustness-preserving variations and task-invalid failures under out-of-distribution (OOD) execution. We introduce Robo-Dopamine 2.0, a history- and OOD-aware process reward model with a pairwise prediction interface. It combines (1) history-conditioned pairwise rewards that use source-aligned reference panels for synthetic OOD queries and observed rollout history for online queries, while preserving the queried endpoints, and (2) an OOD-aware signed progress space that represents valid progress, robustness, failure, and recovery. A Signed-Hop Curriculum with transition-aware replay learns coarse execution ordering before fine-grained progress calibration. We also construct an OOD trajectory dataset and a five-family benchmark. Reference panels improve mean visual order consistency (VOC) from 0.967 to 0.986 and OOD-robust VOC from 0.906 to 0.958. With the same 400K pairwise-reward budget, Signed-Hop training with 25% replay reaches 0.9872 mean VOC, compared with 0.9858 for a matched-pool shuffled control. In downstream reinforcement learning, the full model achieves 86.8% mean RoboTwin success and 71/80 successful real-world insertions.
Human-in-the-loop (HIL) online reinforcement learning for real robots must absorb human interventions quickly while continuing to improve beyond the human prior. We present a training method for this setting based on two components. First, an \emph{MC Q-chunk} critic regresses chunk-level action values onto Monte Carlo returns from the replay buffer, performing sample-average (behavior) policy evaluation so that intervention trajectories are credited directly rather than diluted by current-policy TD backups. Second, \emph{max-Q selective imitation} updates the actor by imitating, at each state, the higher-$Q$ action between the current policy action and a buffer sample under a hard winner-take-all rule. This rule automatically switches between learning from interventions and on-policy self-improvement: when the autonomous policy is stronger, targets align with the policy distribution, reducing the policy--target-sample gap that otherwise induces execution-time distribution shift. In practice we score candidates with a standard critic ensemble mean to reduce comparison noise, without softening targets or introducing score-gap thresholds. On a real USB pick-and-insertion task with 20 demonstrations, ACT QChunk-MCBC attains 99\% success within 30 minutes of HIL training, whereas HIL-SERL requires about 5 hours to converge. In simulation on Peg Insertion and Square, ACT/Flow Q-chunk variants similarly reach $\ge$96\% success within roughly half an hour of effective training, outperforming HIL-SERL, EXPO, and E2HiL on the success--time frontier.
Unmanned aerial vehicle vision-language navigation (UAV-VLN) requires agents to translate visual observations and language instructions into reliable flight actions in complex environments. Although recent end-to-end UAV vision-language-action (UAV-VLA) policies reduce reliance on separately designed perception, planning, and control modules, their behavior-cloning objectives provide limited corrective supervision for interactive closed-loop execution. Reinforcement learning (RL) offers a promising solution, while its effectiveness is constrained by inefficient use of samples, long-tailed scene distributions, and policy distribution shift during optimization. To this end, we propose RecoverFly, a failure-aware RL post-training framework for end-to-end UAV-VLA policies. Specifically, RecoverFly adapts token-level RL for stable optimization of grammar-constrained autoregressive UAV actions, revisits unresolved failure cases to strengthen corrective learning and sample utilization, and combines a two-stage long-tail scene curriculum with reference-policy regularization to improve scene adaptation while preserving acquired capabilities. Experiments on the TravelUAV benchmark demonstrate that RecoverFly achieves the best performance on the seen, unseen-map, and unseen-object splits. Moreover, compared to the AerialVLA initialization, RecoverFly improves success rate by 3.12 to 8.37 percentage points under a total rollout budget of about 30\% of the training-set size, validating its effectiveness, robustness, and generalization capabilities.
Digital-twin calibration requires interaction data that is expensive to collect. We study two acquisition decisions: which trajectories to generate, and when to spend a limited budget on privileged parameter measurements. Our framework couples an excitation-oriented reinforcement learning controller, a recurrent parameter estimator with predictive uncertainty, and a budgeted query policy. In Pendulum, a Random Forest diagnostic recovers gravity only weakly from task-oriented trajectories and does not recover mass or length, while a GRU trained on excitation-oriented trajectories reaches a mean absolute error of 0.0066 with no queries. We then withdraw continuous oracle access partway through an episode, so that the twin must run on the estimator's output for the remainder. The estimator-plus-policy pipeline achieves a terminal error of 0.0092 under a three-query budget, against 0.2031 for an uncalibrated twin. In partially observable Waterworld, five controllers produce different observed error profiles across three hidden parameters, and an estimator trained on a five-controller mixture reaches online normalized errors of roughly 4-5%. These exploratory case studies are not controlled ablations, but they motivate treating trajectory design and query allocation as explicit design variables in data-scarce calibration.
World-Action Models (WAMs) improve end-to-end autonomous driving by transferring video dynamics priors to action prediction, but existing methods incur costly test-time future imagination. We present SimWAM, a simple yet effective WAM that leverages future-video prediction as a training-time supervision signal. It co-trains a pretrained video expert and a lightweight action expert with joint flow matching. An isolated attention mask keeps action prediction independent of future frames, allowing trajectory prediction without explicit future-frame generation at inference. Since the two experts share no parameters and interact only through a unified attention interface, the video backbone could be replaced and the action expert scaled independently without modifying the learning objective or inference pipeline. We further apply reinforcement learning to optimize a compositional driving reward beyond trajectory imitation. Our SimWAM achieves 91.5 PDMS on NAVSIM, surpasses state-of-the-art WAM-based planners with substantially lower latency, and transfers zero-shot to nuScenes. These results position SimWAM as a simple yet solid baseline that could readily benefit from advances in video generation for efficient autonomous driving. The code and model weights are available at https://github.com/H-EmbodVis/SimWAM/.
Vision-language-action (VLA) models are commonly adapted to downstream manipulation tasks via supervised fine-tuning (SFT) or online reinforcement learning (RL) post-training. SFT is prone to distribution mismatch, and existing RL approaches typically apply a single, uniform update strategy to all model components, ignoring their distinct functional roles. We propose TEMPO, a semantic-action decoupled, two-timescale RL post-training framework for VLA models. TEMPO freezes the pretrained vision-language backbone to preserve general semantic representations, and restricts adaptation to two components with dedicated RL optimization loops: the semantic projection layer and the low-level action expert. We update them at different rates--the semantic projection layer infrequently, to keep the latent action stable, and the action expert frequently, to rapidly incorporate control feedback from online interaction. This decoupling RL fine-tuning strategy prevents fast policy updates from destabilizing high-level semantic representations while still allowing the action expert to learn efficiently from online feedback. Experiments on the CALVIN benchmark and real-world manipulation tasks demonstrate that TEMPO consistently outperforms both pretrained state-of-the-art VLA models and the RL post-training baseline, while reaching and maintaining higher evaluation rewards on two real-world tasks.
High-quality demonstrations for dexterous robot manipulation are costly and difficult to collect, whereas monocular human videos provide a scalable source of diverse manipulation behaviors. However, transferring such demonstrations to dexterous robots remains challenging: monocular hand-object interaction (HOI) reconstruction often produces temporally unstable contacts and physically implausible interactions, while conventional retargeting methods struggle to preserve task-relevant contacts and local interaction geometry across different hand embodiments. We present C2Dex, a video-to-dexterous-manipulation framework built around a shared interaction representation: stable object-side contacts recovered by aggregating noisy frame-wise observations in the canonical object space. These stable contacts serve a dual role: as trajectory-level constraints that guide reconstruction toward temporally coherent and physically plausible human HOI trajectories, and as explicit transfer targets for the dexterous hand, where Laplacian interaction optimization preserves the local hand-object geometry across embodiments and residual reinforcement learning refines the trajectory in simulation. Experiments on DexYCB and TACO show that C2Dex achieves end-to-end trajectory success rates of 57.78% and 26.67%, respectively, substantially outperforming the strongest baselines (17.78% and 10.00%) under identical evaluation criteria. Real-robot replay experiments further demonstrate physical feasibility across diverse contact-rich manipulation tasks. Project page: https://k-jie.github.io/C2Dex/
Vision-language models are increasingly serving as the reasoning core of embodied agents. Robot execution is inherently iterative: each action reshapes the scene and physical state, continually renewing what must be perceived, reasoned about, and verified. Meeting these demands requires complementary capabilities that differ in supervision signals, prediction formats, and verification criteria. Existing approaches typically develop these capabilities against isolated, task-specific objectives, leaving open how they should be organized and integrated around execution as a whole. We present Capek 0.5, an embodied vision-language model built around an execution-centric capability taxonomy. Rather than organizing training by datasets or tasks, the taxonomy groups embodied capabilities according to their functional roles throughout execution and comprises four capability families: Spatial Reasoning, Temporal Understanding, Action Guidance, and State Verification. Each capability is first acquired by a dedicated specialist through reinforcement learning with verifiable rewards from a shared backbone, and the specialists are then consolidated into a single inference-time model through weight-space merging followed by routed policy-space distillation. We instantiate Capek 0.5 at the 2B and 35B-A3B scales and evaluate it from three complementary perspectives: comprehensive benchmark suites including Capek-StateBench, a new benchmark for state verification; a controlled study of capability retention from specialists to the unified model; and closed-loop evaluation in simulated embodied environments. Capek 0.5 improves the large majority of matched benchmark rows over its initialization, retains all four specialized capabilities in one checkpoint with quantified losses, and transfers to closed-loop embodied task execution.
Embodied intelligence demands both long-horizon reasoning and real-time closed-loop responsiveness. Recent dual-system Vision-Language-Action (VLA) architectures combine fast reactive control with slow deliberative reasoning to balance inference speed and task success rate. However, existing dual-process VLAs tightly couple the fast module to intermediate representations of the slow module, necessitating end-to-end joint training and limiting modularity, extensibility and flexible system switching. In this paper, we propose Environment-aware Model Selection (EMS), an adaptive VLA inference framework that switches between two fully decoupled systems of different scales through environment-aware model selection. The large-scale deliberative system provides globally consistent trajectory planning to ensure task success, while a lightweight reactive system enables high-frequency closed-loop control. A reinforcement-learning-based switching policy dynamically selects which system to invoke based on real-time feedback, enabling sparse use of the slow system and thereby balancing pretrained knowledge utilisation with runtime efficiency. Our design offers three key advantages over prior hierarchical VLA frameworks: (1) a fully decoupled and modular dual-system architecture that supports plug-and-play model replacement; (2) an adaptive, environment-aware switching strategy; (3) high-frequency inference for responsive closed-loop control. We extensively evaluate EMS in both simulation and real-world environments. On the LIBERO benchmark, EMS achieves success rates comparable to the large-scale baseline while increasing the effective action frequency to 93.4 Hz. The framework further demonstrates strong extensibility in real-world dual-arm manipulation tasks, where it accelerates task completion while maintaining robust performance.
He Jiang, Jingtian Yan, Yulun Zhang +5cs.RO cs.AI cs.MA
Lifelong Multi-Agent Path Finding (LMAPF) requires repeatedly planning collision-free paths for agents that continuously receive new goals upon reaching their current ones. While many learning-based planners have been proposed for LMAPF, most rely on oversimplified kinematic assumptions that may overlook motion constraints critical to real-world performance. In this work, we study a more realistic LMAPF model derived from many real-world automated warehouse systems, termed LMAPF-R2, which incorporates robust safety constraints and in-place rotation constraints. These constraints substantially increase coordination difficulty, particularly in highly constrained spaces. To address these challenges, we propose Search-Aided Joint Reinforcement Learning (SJRL). We first augment neural policies with Causal PIBT, a single-step search-based planner that resolves agents' collisions and propagates their intentions. We then introduce a unified RL formulation that jointly optimizes agent and environment policies, where the environment policy learns graph edge costs to provide global movement guidance via backward Dijkstra search. Experiments demonstrate that SJRL achieves significant improvements over the strong search-based planner, Causal-PIBT, across multiple high-density maps. We further validate SJRL in a challenging mixed-reality warehouse environment with 8 physical robots and 248 virtual robots.
Xuyang Ma, Chaewoon Kim, Haonan Zhang +3cs.AI cs.GR
An interlocking assembly is an assembly in which component parts are connected purely through their geometric arrangement, without relying on external connectors such as glue and nails. Such assemblies have been widely used in a variety of real-world applications due to their structural stability. The problem of generating interlocking assemblies is generally formulated as a shape decomposition problem, where a target 3D object represented as a voxel grid is partitioned into a prescribed number of interlocking pieces. We observe that generating interlocking assemblies is inherently a sequential decision-making problem, where an agent repeatedly decides which piece each voxel should be assigned to. Inspired by the observation, we propose the first reinforcement learning framework RL-Lock for generating interlocking assemblies, without relying on handcrafted search heuristics as existing works did. RL-Lock combines structured action chunking with MCTS-guided policy-value learning to efficiently navigate the large combinatorial search space for interlocking assembly generation. We demonstrate through experiments that RL-Lock allows effective generation of interlocking assemblies, especially for challenging cases in which existing approaches take too long or even fail to find a valid solution.
Senyu Fei, Xiaopeng Yu, Siyin Wang +3cs.RO cs.CL cs.CV
Reinforcement learning (RL) post-training of Vision-Language-Action (VLA) models has shown strong promise for robotic manipulation. Among RL methods, critic-based approaches rely on a value estimator that predominantly operates on single-frame observations or single-frame VLM backbone latents, which is a fundamental mismatch with the partially observable nature of robot control. A naive approach to incorporate observation history into the critic incurs exponential complexity with high-dimensional visual space, and still fails because pure scalar-return regression provides insufficient supervision for learning cross-temporal dynamics. We identify the root cause as a state approximation problem: without an explicit world modeling objective, the critic's representation cannot capture the temporal structure needed for accurate value estimation. To address this, we propose the World Critic Model (WCM), built on a lightweight LeJEPA architecture; WCM jointly predicts future latent state and estimates values, such that the critic's representation is explicitly trained to capture temporal dynamics rather than merely regress scalar returns. WCM integrates seamlessly into both on-policy and off-policy training pipelines and is compatible with state-of-the-art VLA backbones including Pi0, Pi0.5, and OpenVLA-OFT. Extensive experiments on 149 tasks across four benchmarks demonstrate that WCM consistently achieves state-of-the-art performance in both in-distribution and out-of-distribution settings, with particularly strong generalization gains. We further validate WCM on seven real-world manipulation tasks using OpenVLA-OFT and Pi0.5 with off-policy RL, confirming stable deployment across diverse settings.
Reinforcement learning can produce robust humanoid controllers, but each new task is typically trained as a separate policy with its own reward design and training process. Motion imitation provides an alternative source of motor competence by training policies to track retargeted human motions, yet the resulting controllers remain reference trackers and are not directly usable as task policies. We propose a three-stage pipeline that turns motion-imitation skills into a reusable hybrid motion prior (HMP) for humanoid locomotion. First, an expert policy is trained to imitate retargeted human motion-capture clips. Second, the expert is distilled into a frozen architecture composed of a proprioceptive encoder, a residual vector-quantized (RVQ) codebook, and an action decoder. Third, task-level policies are trained to solve locomotion tasks by selecting discrete codebook entries while the HMP remains frozen. We evaluate the method on velocity tracking, point-goal navigation, and fall-recovery velocity tracking in simulation, and deploy the velocity-tracking policy on a real Unitree G1 robot. The distillation process preserves the tracking behavior of the expert, while the resulting HMP can be reused without retraining as the action interface for different downstream locomotion policies. The learned HMP reveals an interpretable codebook structure in which the number of active RVQ stages modulates the available gait patterns. We further show that training the codebook with the rotation trick improves latent organization and reduces downstream falls compared with a standard straight-through estimator.
Deploying navigation systems at scale requires a recipe that minimizes sensor assumptions, generalizes across robot embodiments, and trains efficiently. Yet, today's best systems depend on depth sensors, multi-camera rigs, or pre-built maps, limiting the hardware they support and increasing deployment cost. We introduce Robostral Navigate, an 8B vision-language model built around this scalability objective. The model consumes only a stream of monocular RGB images - the most ubiquitous sensor across robotic platforms and predicts waypoints by pointing to the next target location in the current camera view. Operating purely in image space, rather than robot-specific coordinates, makes the policy naturally robust to changes in camera intrinsics and scene scale, enabling deployment across wheeled, legged, and aerial robots without recalibration. We generate 2.4 million trajectories across 350k simulated scenes to reduce the reliance on real-world data collection and scale easily. We further introduce a prefix-caching training recipe that packs entire episodes into single training sequences, reducing training tokens by 22x and cutting training time from months to days. A tree-based attention mask prevents conditioning on previous ground-truth actions, encouraging visually grounded action prediction, and reinforcement learning is used to further improve exploration and recovery capabilities. On the Room-to-Room and Room-Across-Room in Continuous Environments (R2R-CE and RxR-CE) benchmarks, Robostral Navigate sets a new state of the art. On R2R-CE, it achieves a 77.4% success rate, surpassing the best monocular method by 10.5 points and the strongest depth- or multi-camera system by 5.3 points despite using only a single RGB camera. On RxR-CE, it reaches 75.1% success rate, outperforming all monocular baselines.
Elena Sorina Lupu, Patrick Spieler, Khurram Javed +4cs.RO cs.AI eess.SY
Reinforcement learning (RL) research has demonstrated success in both physical and simulated domains; however, the predominant methodology remains rooted in simulations. The predominance of simulations makes translating research to physical reality uncertain for both algorithms and researchers. We propose a physical platform that is designed to simplify the transition. In this paper, we present the Open Ant: a physical variant of the commonly used Gymnasium Ant environment, along with a simulation. We demonstrate that competent walking policies can be learned from scratch in approximately one hour directly from the physical robot's experience for two substantially different RL algorithms: SARSA($λ$) and Soft Actor-Critic (SAC). Separately, we show policies that were learned in simulation transfer to reality. We also examine how well the platform supports a nimble experimental ecosystem. Specifically, we observe the speed with which new users from diverse backgrounds achieve their first success with the platform, and how easily the platform can be repaired and updated when hardware issues arise. Both the hardware design and software are available as open-source on GitHub for ease of customization. In summary, we advocate for the use of the Open Ant for RL researchers who frequently use simulated environments, so they can more easily include robot experiments in their evaluations.
Action-conditioned world models are a key component of embodied AI, serving as scalable policy evaluators that reduce reliance on expensive real-world rollouts. To accurately capture diverse action-induced dynamics, such models should satisfy three key objectives-Physical Plausibility (P), Action Adherence (A), and Visual Fidelity (V), collectively referred to as PAV-while remaining robust to both in-distribution (ID) expert demonstrations and out-of-distribution (OOD) actions. However, existing methods primarily rely on ID action-video pairs and pixel-level reconstruction losses, which do not explicitly optimize PAV objectives and generalize poorly beyond expert data. To address this, we propose PAVXploreRL, a reinforcement learning framework built on a pretrained latent world model that explicitly optimizes PAV objectives through reward-driven training. To improve action generalization, our method jointly leverages ID trajectories and noise-driven OOD action exploration, without paired video supervision. Experiments show that PAVXploreRL consistently outperforms pretrained baselines, achieving a 5.6% average gain across benchmarks and producing higher-quality PAV properties. As a policy evaluator, it also yields more reliable performance estimates and reduces the overestimation bias of prior expert-only world models such as Ctrl-World. Code: https://github.com/Social-AI-Studio/PAVXploreRL
Autonomous driving vision-language models (VLMs) struggle in roadwork zones, where familiar visual cues such as lane markings and permanent signs are altered or absent, and temporary devices such as cones and barriers redefine the drivable corridor. VLMs can detect these objects, but without explicit guidance they anchor their reasoning on familiar elements from pre-training and fail to connect work-zone observations to correct planning decisions. We propose WorkDrive, a framework that constructs perception-grounded causal reasoning for work zones and aligns it with trajectory prediction. An automated multitask perception pipeline extracts structured scene facts and injects them into a Chain-of-Causation (CoC) annotation pipeline, redirecting the annotator's attention to domain-specific elements. The resulting reasoning labels are used for supervised fine-tuning, followed by reinforcement learning with a single reward: consistency between lateral meta-actions and the predicted trajectory. On ROADWork, the largest public work-zone dataset, the proposed roadwork CoC reduces trajectory average displacement error (ADE) by 9.0\%, and consistency-based GRPO yields a further 3.0\%, achieving progressive improvement over the trajectory-only baseline. Code and data will be publicly released.
Jun-Gill Kang, Jaehyun Park, Tae-Gyu Song +3cs.RO cs.AI cs.LG
Enabling quadrupedal robots to traverse complex terrains-from rugged outdoor environments to urban landscapes-requires seamless integration of multiple motor skills, smooth transitions between gaits, and high-speed perceptive locomotion using only onboard sensors. We present APT-RL (Action Pretrained Transformer-based Reinforcement Learning), a unified framework that enables multi-skill locomotion to achieve high-speed traversal in complex environments through autonomous skill transitions utilizing only onboard perception and computation. Our approach generates large-scale, feature-rich 2D motion datasets through trajectory optimization with simplified dynamics. These datasets enable training of diverse, reusable locomotion skills that transfer effectively to a real quadruped robot operating on complex uneven terrains. The resulting high-quality skills serve as strong priors for efficient learning of complex downstream tasks and extend naturally to 3D environments, enabling smooth, high-speed multi-skill locomotion in deployed policy. Real-world experiments demonstrate the framework's capabilities: the robot performs agile maneuvers through complex indoor obstacles and outdoor wild environments, including dynamic drop-down maneuvers that reach instantaneous peak speeds of up to 6 meters per second. A single onboard policy enables robust traversal of diverse obstacles, including stairs, hurdles, stepping stones, gaps, and fallen branches, demonstrating the versatility and effectiveness of our approach.
Zhouchonghao Wu, Akshay Rangesh, Weixin Li +4cs.LG cs.AI cs.RO
Training robust autonomous driving agents requires a simulator that is fast enough for reinforcement learning at scale, realistic enough to ground behavior in real-world map structure, and diverse enough to cover the safety-critical long tail that logged data rarely contains. We present TerraZero, a procedural driving simulator and self-play training stack. A configurable C engine runs simulation on the CPU and policy inference on the GPU over a zero-copy path, sustaining 1.3M agent-steps per second on a single server-grade GPU, far faster than existing object-level simulators, while keeping fidelity lighter single-agent systems omit: heterogeneous agents, multiple dynamics models, and full traffic-rule enforcement. TerraZero treats logged data only as a source of real-world map geometry, populating each map with randomized rule-based road users and signal controllers and randomizing agent dynamics, rewards, and sizes per episode, so a map yields an unbounded set of scenarios. Every reported policy trains from scratch by reinforcement learning alone on a compute-efficient self-play recipe across GPUs, with zero human demonstrations and no fallback planner at inference. Policies generalize zero-shot across cities and datasets, including emergent left-hand-traffic driving without explicit supervision. As an ego policy, TerraZero is the first fully learned policy to top the InterPlan long-tail benchmark, ahead of larger learned planners; on routine-driving val14 it ranks among the best approaches and is the safest, posting the best collision and time-to-collision scores. On Waymo Open Sim Agents realism the same recipe outperforms other demonstration-free methods and is competitive with the strongest reference-anchored self-play method. One stack serves both roles: driving policies across dynamics for cars and trucks, and sim agents that jointly control vehicles, pedestrians, and cyclists.