Cloth folding stands as an intricate subject in robot manipulation, requiring robots to fold diverse fabrics into different configurations according to human intentions. Previous work in this area falls into three primary categories: imitation learning, reinforcement learning, and geometric model-based planning methods. While each paradigm has its merits, they generally lack inherent multi-step reasoning ability and struggle to generalize to novel cloth appearances and tasks. To tackle these problems, our key insight is incorporating the common sense reasoning and generalization abilities of Large Language Models (LLMs) into cloth manipulation, while addressing the limitations of LLMs in manipulating deformable objects, which involves an effective grounding module and rational planning hierarchy. To this end, we present PolyFold, a novel language-conditioned bimanual cloth folding framework that leverages the parameterized polygon model as an effective abstraction and grounding module for cloth representation. Moreover, PolyFold enables LLMs to infer an intermediate-level action—specifically, the symmetrical fold line, while delegating the pick-and-place calculations to a fold-line-guided downstream policy, which is learned through self-supervision using random data. Experiments on 70 cloth folding tasks and 4 cloth types show that PolyFold excels in zero-shot generalization and inherent multi-step reasoning capability, while also operating in a sample-efficient expert-demonstration-free manner, surpassing previous SOTA vision-conditioned and language-conditioned methods. Our method can also be directly deployed in real-world scenarios. Videos, code and appendix are available at our project webpage: https://sites.google.com/view/polyfold.

Grounding 3D object affordance is a task that locates objects in 3D space where they can be manipulated, which links perception and action for embodied intelligence. For example, for an intelligent robot, it is necessary to accurately ground the affordance of an object and grasp it according to human instructions. In this paper, we introduce a novel task that grounds 3D object affordance based on language instructions, visual observations and interactions, which is inspired by cognitive science. We collect an Affordance Grounding dataset with Points, Images and Language instructions (AGPIL) to support the proposed task. In the 3D physical world, due to observation orientation, object rotation, or spatial occlusion, we can only get a partial observation of the object. So this dataset includes affordance estimations of objects from full-view, partial-view, and rotation-view perspectives. To accomplish this task, we propose LMAffordance3D, the first multi-modal, language-guided 3D affordance grounding network, which applies a vision-language model to fuse 2D and 3D spatial features with semantic features. Comprehensive experiments on AGPIL demonstrate the effectiveness and superiority of our method on this task, even in unseen experimental settings. Our project is available at https://sites.google.com/view/lmaffordance3d.

We study the task of language-conditioned pick and place in clutter, where a robot should grasp a target object in open clutter and move it to a specified place. Some approaches learn end-to-end policies with features from vision foundation models, requiring large datasets. Others combine foundation models in a zero-shot setting, suffering from cascading errors. In addition, they primarily leverage vision and language foundation models, focusing less on action priors. In this paper, we aim to develop an effective policy by integrating foundation priors from vision, language, and action. We propose A2, an action prior alignment method that aligns unconditioned action priors with 3D vision-language priors by learning one attention layer. The alignment formulation enables our policy to train with less data and preserve zero-shot generalization capabilities. We show that a shared policy for both pick and place actions enhances the performance for each task, and introduce a policy adaptation scheme to accommodate the multi-modal nature of actions. Extensive experiments in simulation and the real-world show that our policy achieves higher task success rates with fewer steps for both pick and place tasks in clutter, effectively generalizing to unseen objects and language instructions. Videos and codes are available at https://xukechun.github.io/papers/A2.

We focus on the task of unknown object rearrangement, where a robot is supposed to reconfigure the objects into a desired goal configuration specified by an RGB-D image. Recent works explore unknown object rearrangement systems by incorporating learning-based perception modules. However, they are sensitive to perception error, and pay less attention to task-level performance. In this article, we aim to develop an effective system for unknown object rearrangement amidst perception noise. We theoretically reveal that the noisy perception impacts grasp and place in a decoupled way, and show such a decoupled structure is valuable to improve task optimality. We propose grasp, see, and place (GSP), a dual-loop system with the decoupled structure as prior. For the inner loop, we learn a see policy for self-confident in-hand object matching. For the outer loop, we learn a grasp policy aware of object matching and grasp capability guided by task-level rewards. We leverage the foundation model CLIP for object matching, policy learning, and self-termination. A series of experiments indicate that GSP can conduct unknown object rearrangement with higher completion rates and fewer steps.

Visual servoing (VS) is a widely used technique in industries where there are hundreds of robots, but it requires accurate camera calibration including camera intrinsic and extrinsic parameters. However, it is labour-intensive to calibrate robots one-by-one in practical use. In this paper, we propose a neural uncalibrated VS policy (NUVS) that can adapt to calibration disturbances with an adaption mechanism and a control-oriented guidance. It bridges the disturbance adaption of classical VS methods and the large convergence of learning-based VS methods. NUVS estimates the calibration embedding from past observations and servos to the desired pose under the supervision of a PBVS that can access the ground truth in simulation. With this adaption mechanism, NUVS outperforms the classical IBUVS algorithm when facing large initial camera pose offsets under the calibration disturbance. Supplementary material in: https://sites.google.com/view/neural-uncalibrated-vs

Self-assessment rules play an essential role in safe and effective real-world robotic applications, which verify the feasibility of the selected action before actual execution. But how to utilize the self-assessment results to re-choose actions remains a challenge. Previous methods eliminate the selected action evaluated as failed by the self-assessment rules, and re-choose one with the next-highest affordance (i.e. process-of-elimination strategy [1]), which ignores the dependency between the self-assessment results and the remaining untried actions. However, this dependency is important since the previous failures might help trim the remaining over-estimated actions. In this paper, we set to investigate this dependency by learning a failure-aware policy. We propose two architectures for the failure-aware policy by representing the self-assessment results of previous failures as the variable state, and leveraging recurrent neural networks to implicitly memorize the previous failures. Experiments conducted on three tasks demonstrate that our method can achieve better performances with higher task success rates by less trials. Moreover, when the actions are correlated, learning a failure-aware policy can achieve better performance than the process-of-elimination strategy.

We focus on the task of language-conditioned grasping in clutter, in which a robot is supposed to grasp the target object based on a language instruction. Previous works separately conduct visual grounding to localize the target object, and generate a grasp for that object. However, these works require object labels or visual attributes for grounding, which calls for handcrafted rules in planner and restricts the range of language instructions. In this paper, we propose to jointly model vision, language and action with object-centric representation. Our method is applicable under more flexible language instructions, and not limited by visual grounding error. Besides, by utilizing the powerful priors from the pre-trained multi-modal model and grasp model, sample efficiency is effectively improved and the sim2real problem is relived without additional data for transfer. A series of experiments carried out in simulation and real world indicate that our method can achieve better task success rate by less times of motion under more flexible language instructions. Moreover, our method is capable of generalizing better to scenarios with unseen objects and language instructions.

We focus on the task of object manipulation to an arbitrary goal pose, in which a robot is supposed to pick an assigned object to place at the goal position with a specific orientation. However, limited by the execution space of the manipulator with gripper, one-step picking, moving and releasing might be failed, where a reorientation object pose is required as a transition. In this paper, we propose a learning-driven anytime prioritized search-based solver to find a feasible solution with low path cost in a short time. In our work, the problem is formulated as a hierarchical learning problem, with the high level finding a reorientation object pose, and the low level planning paths between adjacent grasps. We learn an offline-training path cost estimator to predict approximate path planning costs, which serve as pseudo rewards to allow for pre-training the high-level planner without interacting with the simulator. To deal with the problem of distribution mismatch of the cost net and the actual execution cost space, a refined training stage is conducted with simulation interaction. A series of experiments carried out in simulation and real world indicate that our system can achieve better performances in the object manipulation task with less time and less cost.

In-flight objects capture is extremely challenging. The robot is required to complete trajectory prediction, interception position calculation and motion planning within tens of milliseconds. As in-flight uneven objects are affected by various kinds of forces, which leads to the time-varying acceleration, motion prediction for them is difficult. In order to compensate the system’s non-linearity, we propose using a recurrent neural network model, which we call the Neural Acceleration Estimator (NAE), to estimate the varying acceleration by observing a small fragment of previous deflected trajectory without any prior information. Moreover, end-to-end training with Differantiable Filter (NAE-DF) gives a supervision for measurement uncertainty and further improves the prediction accuracy. Experimental results show that motion prediction with NAE and NAE-DF is superior to other methods and has a good generalization performance on unseen objects. We test our methods on a robot, performing velocity control in real world and respectively achieve 83.3% and 86.7% success rate on a ploy urethane banana and a gourd. We also release an object in-flight dataset containing 1,500 trajectorys for uneven objects, which can be found on the project website:https://sites.google.com/view/neural-motion-prediction.