The Conference on Robot Learning (CoRL) is where cutting edge research on robotics and machine learning intersect. This year, CoRL 2026 makes its way to Austin, Texas, running from November 9-11.

Since its debut in 2017, CoRL has championed breakthroughs that redefine intelligent robotic control and perception. Celebrating its milestone 10th year, the conference will bring together leading global researchers, engineers, and visionaries to present pioneering studies, discuss technical challenges, and chart the next frontier of autonomous systems.

Research teams at the RAI Institute had several papers and workshops accepted, covering topics such as handheld data collection, loco-manipulation, robot mapping and perception, sim-to-real policy transfer and zero-shot long-horizon planning.

Publications

HLO-AMP: Hierarchical Loco-Manipulation from Object-Aware Motion Priors

Hierarchical pipelines for humanoid loco-manipulation typically pair a task-agnostic whole-body controller (WBC) with a downstream task policy. Such WBCs are often trained through manual reward shaping, which yields visibly machine-like motions; more consequentially, they are trained without the manipulated object in the scene. Object dynamics are instead approximated through randomized end-effector force disturbances, a proxy that fails to reproduce the distributed multi-contact wrench a real payload applies across the arms and torso, with the gap becoming critical near the platform’s payload limit.

In this work, we train the low-level (LL) controller itself as an object-aware motion prior: a single policy trained via motion imitation on retargeted motion-capture (MoCap) references that span the core motion primitives of the downstream task, with the manipulated object physically present in the scene. This single stage inherits natural human-like coordination from the references and absorbs the realistic contact wrench at its source, while exposing the same compact task-space command interface used by command-tracking WBCs. A reference-free high-level (HL) policy is then trained on top, with its action channel anchored to the empirical envelope of the low-level training data so that it generalizes the priors to a long-horizon multi-stage task while structurally preventing queries that would push the low level off-distribution. We instantiate the framework on a Unitree G1 for bimanual box pick-and-place: the same deployed policy generalizes across object masses and dimensions, initial and goal poses, and reactively recovers from intentional drops mid-carry.

Learning Loco-Manipulation From SMPC Demonstrations With Sparse Offline-to-Online RL

Integrating locomotion and manipulation is essential for robot autonomy, but scaling standard Reinforcement Learning (RL) to complex tasks is severely bottlenecked by the slow, manual process of dense reward shaping. To bypass this limitation, we leverage Sample-based Model Predictive Control (SMPC) entirely in simulation as an automated, rapidly tunable expert to generate massive offline datasets. Because this data solves the fundamental exploration problem, we can train an off-policy RL agent using purely sparse task rewards, drastically reducing the time required to learn new skills and eliminating the need for manual tuning. Integrating this high-level agent with a low-level dynamic stability controller yields more optimal behaviors that strictly align with true task objectives, ultimately allowing the learned policies to surpass the original optimal control teacher. We validate the robustness of this sim-to-real framework by successfully deploying complex loco-manipulation skills across different morphologies, including an arm-equipped Spot quadruped and a G1 humanoid.

RoMo-GS: Efficient Online 3D Gaussian Splatting for Robot Mapping via Exposure-Integrated Motion Blur

Online 3D Gaussian Splatting (3DGS) is emerging as a practical way to build a representation for robotics applications. However, deploying it on real-world platforms exposes multiple challenges. Rapid maneuvers of a robot induce motion-blurred images, which corrupt the reconstructed geometry and cause camera pose drift. Furthermore, the number of Gaussians grows without bound during continuous online mapping.

In this study, we present RoMo-GS, an online 3DGS system designed to address these challenges. Instead of averaging N discrete sub-frames as in prior motion blur-aware 3DGS, we analytically integrate each projected Gaussian over the exposure interval, yielding an O(1) complexity per Gaussian. Building on this, we bound the Gaussians’ growth with a Hessian-based pruning score, while jointly refining pose, velocity, and Gaussians with Adam and an online LM step over the Gaussians. On multiple public datasets and a self-collected dataset, RoMo-GS matches or exceeds offline deblurring 3DGS methods in rendering quality while running 15–150×faster, and operates online at competitive quality with about 20% fewer Gaussians than the online 3DGS baseline.

ExpertGen: Scalable Sim-to-Real Expert Policy Learning from Imperfect Behavior Priors

Learning generalizable and robust behavior cloning policies requires large volumes of high-quality robotics data. While human demonstrations (e.g., through teleoperation) serve as the standard source for expert behaviors, acquiring such data at scale in the real world is prohibitively expensive. This paper introduces EXPERTGEN, a framework that automates expert policy learning in simulation without any human demonstrations, while preserving strong sim-to-real transfer performance. EXPERTGEN first initializes a behavior prior in using a diffusion policy trained on synthetic data (e.g., Code as Policy). Reinforcement learning is then used to steer this prior in massively parallel simulation toward high task success by optimizing the diffusion model’s initial noise while keeping the denosing process frozen. In this way, EXPERTGEN regularizes exploration to remain within safe behavior manifolds, enabling stable learning from sparse rewards without additional reward engineering.

Extensive experiments on challenging manipulation benchmarks demonstrate that EXPERTGEN reliably acquires high-quality policies with strong robustness and generalization. On long-horizon manipulation tasks, EXPERTGEN achieves success rates of up to 85%, while on industrial assembly benchmarks it achieves a 91.1% overall success rate, outperforming all baseline methods. To validate sim-to-real transfer, the learned state-based expert policies are further distilled into visuomotor policies via DAgger and successfully deployed on real hardware.

NovaPlan: Zero-Shot Long-Horizon Manipulation via Closed-Loop Video Language Planning

Solving complex long-horizon robotic tasks requires joint reasoning over abstract task structure and low-level physical interaction. While combining Vision-Language Models (VLMs) and video generation models offers a promising path for zero-shot planning, their individual tendencies to hallucinate physics or violate geometric consistency often compound over time, preventing reliable real-world execution. We introduce NovaPlan, a hierarchical framework that enables robust, zero-shot long-horizon manipulation by systematically proposing, verifying, and repairing visual plans. At the high level, a VLM planner decomposes tasks and filters out dynamically inconsistent futures by verifying multiple candidate video rollouts. To translate these imagined futures into reliable physical actions, NovaPlan utilizes a hybrid geometric representation that adaptively switches between object-centric flow and human hand flow. 

Finally, NovaPlan closes the loop by continuously monitoring execution to verify outcomes and synthesize local, non-prehensile corrective behaviors, such as fingertip poking, when failures occur. Across diverse multi-stage tasks, NovaPlan substantially outperforms prior zero-shot systems, achieving complex assembly and dexterous error recovery entirely without task-specific training or demonstrations.

Workshops

Eucalyptus: A High-Fidelity Multimodal UMI

Portable data-collection interfaces enable the acquisition of real-world manipulation demonstrations without requiring a robot during collection. We present Eucalyptus, an extension of UMI designed for sustained, force-aware data collection. Eucalyptus combines a robustified and ergonomics-oriented mechanical design with a high-fidelity multimodal sensing suite comprising RGB vision, gripper width, interaction forces, and 6-DoF pose. The mechanical redesign eliminates measurable backlash, expands the maximum gripper width, improves ergonomics for a wide range of hand sizes, and supports an expanded sensing stack. Together, these changes improve the experience of data collection for the operator and provide richer measurements of the demonstrator’s interaction with the environment. We characterize the system and validate its utility through downstream visuomotor policy learning of several contact-rich tasks.

Koala Gripper: Co-designing Robotic Grippers and Data-Capture Devices for Scaling Dexterous Manipulation Learning

As the demand for larger manipulation datasets grows, handheld robotic gripper data collection and the associated gripper designs become more vital. Current data collection device designs trend towards matching the morphologies of existing robotic grippers, sacrificing ergonomics and manipulation performance. In this paper, we propose a co-design framework that guides the simultaneous development of both data collection and robotic execution devices by weaving both platform constraints into the design process. Through this workflow, we present the Koala Gripper system, a data capture device and robotic gripper platform that improves dexterity and grasp capability compared to parallel jaw grippers while preserving scalability and ease-of-use. The design introduces a novel force-optimized finger/trigger linkage mechanism with directional reflected mass characteristics, a unique monolithic dual-thumb, and user-centered ergonomic design. The design’s actuated robotic fingers are backdrivable, with effective mass on the order of tens of grams. We show that these grippers are capable of secure grasps over a wide range of objects, forceful tool use, and precise singulation. We further validate the platform by deploying it with an end-to-end data collection and policy execution pipeline that highlights its capabilities through learning from demonstration.

An Empirical Analysis of Controller Selection for Safe Execution of UMI-Style Policies

Handheld devices such as the Universal Manipulation Interface (UMI) enable scalable, embodiment-agnostic data collection for behavior cloning. The resulting policies predict Cartesian end-effector trajectories and delegate execution to a robot-specific controller, which trades off tracking accuracy with compliance. However, current policy evaluation protocols treat the controller as a fixed implementation detail, conflating policy and execution errors. We disentangle the two by replaying real handheld demonstrations on a simulated Franka FR3 while varying only the controller across speeds, payloads, and contact-directed reference errors. We find that controller choice alone substantially changes task success, even for kinematically feasible references; soft impedance closes the drawer in only 6% of trials at nominal speed, whereas tuned impedance always succeeds. Under a 20 mm reference error, stiff position control drops to 0% success on bottle placement and uprighting, while compliant controllers risk applying insufficient force. Force- aware residual control mitigates this trade-off, yet no single controller performs best under learned flow-matching policies. We argue that the community should carefully consider the controller as an explicit “tunable knob” during inference.

Sceniris: A Fast Procedural Scene Generation Framework

Synthetic 3D scenes are essential for developing Physical AI and generative models. Existing procedural generation methods often have low output throughput, creating a significant bottleneck in scaling up dataset creation. In this work, we introduce Sceniris, a highly efficient procedural scene generation framework for rapidly generating large-scale, collision-free scene variations. Sceniris also provides an optional robot reachability check, providing manipulation-feasible scenes for robot tasks. Sceniris is designed for maximum efficiency by addressing the primary performance limitations of the prior method, Scene Synthesizer. Leveraging batch sampling and faster collision checking in cuRobo, Sceniris achieves at least 234x speed-up over Scene Synthesizer. Sceniris also expands the object-wise spatial relationships available in prior work to support diverse scene requirements.

Bridging Handheld and Teleoperated Supervision for Contact-Rich Manipulation via State-Gated Experts

Handheld data collection systems, such as the Universal Manipulation Interface (UMI), enable scalable data collection across diverse environments but only capture observed actions rather than the desired actions executed by a robot controller. In contrast, teleoperation captures desired actions directly, but is prohibitively time-consuming to collect. We revisit this trade-off through the lens of label validity across task phases. We observe that handheld trajectories provide valid supervision in tolerant, free-space phases, but lack dynamic feasibility in contact-sensitive phases, where tracking observed trajectories at high stiffness produces large, unsafe contact forces. We study the interaction between these two supervision types for contact-rich manipulation and find that training policies that combine handheld data with a small number of targeted teleoperated demonstrations provides an efficient hybrid strategy. Specifically, rather than teleoperating the entire task, we only collect partial teleoperated demonstrations for task segments where base handheld policies fail. However, naively mixing handheld and teleoperated phase-specific data yields worse performance than training on handheld data alone. To address this mismatch between observed and desired supervision, we propose Bi-modal Routing for Imitation Data via Gated Experts (BRIDGE), a mixture of diffusion policy experts that routes between specialist task phase heads conditioned on the current robot state. Notably, our approach produces measurably lower end-effector forces under contact and improves success rates over handheld-only baselines by up to 36.7% across three contact-rich manipulation tasks.