IMPACT: Intelligent Motion Planning with Acceptable Contact Trajectories via Vision-Language Models

Motion planning involves determining a sequence of robot configurations to reach a desired pose, subject to movement and safety constraints. Traditional motion planning finds collision-free paths, but this is overly restrictive in clutter, where it may not be possible for a robot to accomplish a task without contact. In addition, contacts range from relatively benign (e.g., brushing a soft pillow) to more dangerous (e.g., toppling a glass vase). Due to this diversity, it is difficult to characterize which contacts may be acceptable or unacceptable. In this paper, we propose IMPACT, a novel motion planning framework that uses Vision-Language Models (VLMs) to infer environment semantics, identifying which parts of the environment can best tolerate contact based on object properties and locations. Our approach uses the VLM's outputs to produce a dense 3D "cost map" that encodes contact tolerances and seamlessly integrates with standard motion planners. We perform experiments using 20 simulation and 10 real-world scenes and assess using task success rate, object displacements, and feedback from human evaluators. Our results over 3620 simulation and 200 real-world trials suggest that IMPACT enables efficient contact-rich motion planning in cluttered settings while outperforming alternative methods and ablations. Supplementary material is available at https://impact-planning.github.io/.

Sequential Multi-Object Grasping with One Dexterous Hand

Sequentially grasping multiple objects with multi-fingered hands is common in daily life, where humans can fully leverage the dexterity of their hands to enclose multiple objects. However, the diversity of object geometries and the complex contact interactions required for high-DOF hands to grasp one object while enclosing another make sequential multi-object grasping challenging for robots. In this paper, we propose SeqMultiGrasp, a system for sequentially grasping objects with a four-fingered Allegro Hand. We focus on sequentially grasping two objects, ensuring that the hand fully encloses one object before lifting it and then grasps the second object without dropping the first. Our system first synthesizes single-object grasp candidates, where each grasp is constrained to use only a subset of the hand's links. These grasps are then validated in a physics simulator to ensure stability and feasibility. Next, we merge the validated single-object grasp poses to construct multi-object grasp configurations. For real-world deployment, we train a diffusion model conditioned on point clouds to propose grasp poses, followed by a heuristic-based execution strategy. We test our system using $8 \times 8$ object combinations in simulation and $6 \times 3$ object combinations in real. Our diffusion-based grasp model obtains an average success rate of 65.8% over 1600 simulation trials and 56.7% over 90 real-world trials, suggesting that it is a promising approach for sequential multi-object grasping with multi-fingered hands. Supplementary material is available on our project website: https://hesic73.github.io/SeqMultiGrasp.

Cross-domain Few-shot Object Detection with Multi-modal Textual Enrichment

Advancements in cross-modal feature extraction and integration have significantly enhanced performance in few-shot learning tasks. However, current multi-modal object detection (MM-OD) methods often experience notable performance degradation when encountering substantial domain shifts. We propose that incorporating rich textual information can enable the model to establish a more robust knowledge relationship between visual instances and their corresponding language descriptions, thereby mitigating the challenges of domain shift. Specifically, we focus on the problem of Cross-Domain Multi-Modal Few-Shot Object Detection (CDMM-FSOD) and introduce a meta-learning-based framework designed to leverage rich textual semantics as an auxiliary modality to achieve effective domain adaptation. Our new architecture incorporates two key components: (i) A multi-modal feature aggregation module, which aligns visual and linguistic feature embeddings to ensure cohesive integration across modalities. (ii) A rich text semantic rectification module, which employs bidirectional text feature generation to refine multi-modal feature alignment, thereby enhancing understanding of language and its application in object detection. We evaluate the proposed method on common cross-domain object detection benchmarks and demonstrate that it significantly surpasses existing few-shot object detection approaches.

PhysBench: Benchmarking and Enhancing Vision-Language Models for Physical World Understanding

Understanding the physical world is a fundamental challenge in embodied AI, critical for enabling agents to perform complex tasks and operate safely in real-world environments. While Vision-Language Models (VLMs) have shown great promise in reasoning and task planning for embodied agents, their ability to comprehend physical phenomena remains extremely limited. To close this gap, we introduce PhysBench, a comprehensive benchmark designed to evaluate VLMs' physical world understanding capability across a diverse set of tasks. PhysBench contains 10,002 entries of interleaved video-image-text data, categorized into four major domains: physical object properties, physical object relationships, physical scene understanding, and physics-based dynamics, further divided into 19 subclasses and 8 distinct capability dimensions. Our extensive experiments, conducted on 75 representative VLMs, reveal that while these models excel in common-sense reasoning, they struggle with understanding the physical world -- likely due to the absence of physical knowledge in their training data and the lack of embedded physical priors. To tackle the shortfall, we introduce PhysAgent, a novel framework that combines the generalization strengths of VLMs with the specialized expertise of vision models, significantly enhancing VLMs' physical understanding across a variety of tasks, including an 18.4\% improvement on GPT-4o. Furthermore, our results demonstrate that enhancing VLMs' physical world understanding capabilities can help embodied agents such as MOKA. We believe that PhysBench and PhysAgent offer valuable insights and contribute to bridging the gap between VLMs and physical world understanding.

Learning to Singulate Objects in Packed Environments using a Dexterous Hand

Robotic object singulation, where a robot must isolate, grasp, and retrieve a target object in a cluttered environment, is a fundamental challenge in robotic manipulation. This task is difficult due to occlusions and how other objects act as obstacles for manipulation. A robot must also reason about the effect of object-object interactions as it tries to singulate the target. Prior work has explored object singulation in scenarios where there is enough free space to perform relatively long pushes to separate objects, in contrast to when space is tight and objects have little separation from each other. In this paper, we propose the Singulating Objects in Packed Environments (SOPE) framework. We propose a novel method that involves a displacement-based state representation and a multi-phase reinforcement learning procedure that enables singulation using the 16-DOF Allegro Hand. We demonstrate extensive experiments in Isaac Gym simulation, showing the ability of our system to singulate a target object in clutter. We directly transfer the policy trained in simulation to the real world. Over 250 physical robot manipulation trials, our method obtains success rates of 79.2%, outperforming alternative learning and non-learning methods.

VoxAct-B: Voxel-Based Acting and Stabilizing Policy for Bimanual Manipulation

Bimanual manipulation is critical to many robotics applications. In contrast to single-arm manipulation, bimanual manipulation tasks are challenging due to higher-dimensional action spaces. Prior works leverage large amounts of data and primitive actions to address this problem, but may suffer from sample inefficiency and limited generalization across various tasks. To this end, we propose VoxAct-B, a language-conditioned, voxel-based method that leverages Vision Language Models (VLMs) to prioritize key regions within the scene and reconstruct a voxel grid. We provide this voxel grid to our bimanual manipulation policy to learn acting and stabilizing actions. This approach enables more efficient policy learning from voxels and is generalizable to different tasks. In simulation, we show that VoxAct-B outperforms strong baselines on fine-grained bimanual manipulation tasks. Furthermore, we demonstrate VoxAct-B on real-world $\texttt{Open Drawer}$ and $\texttt{Open Jar}$ tasks using two UR5s. Code, data, and videos will be available at https://voxact-b.github.io.

Learning Granular Media Avalanche Behavior for Indirectly Manipulating Obstacles on a Granular Slope

Legged robot locomotion on sand slopes is challenging due to the complex dynamics of granular media and how the lack of solid surfaces can hinder locomotion. A promising strategy, inspired by ghost crabs and other organisms in nature, is to strategically interact with rocks, debris, and other obstacles to facilitate movement. To provide legged robots with this ability, we present a novel approach that leverages avalanche dynamics to indirectly manipulate objects on a granular slope. We use a Vision Transformer (ViT) to process image representations of granular dynamics and robot excavation actions. The ViT predicts object movement, which we use to determine which leg excavation action to execute. We collect training data from 100 real physical trials and, at test time, deploy our trained model in novel settings. Experimental results suggest that our model can accurately predict object movements and achieve a success rate $\geq 80\%$ in a variety of manipulation tasks with up to four obstacles, and can also generalize to objects with different physics properties. To our knowledge, this is the first paper to leverage granular media avalanche dynamics to indirectly manipulate objects on granular slopes. Supplementary material is available at https://sites.google.com/view/grain-corl2024/home.

GPT-Fabric: Folding and Smoothing Fabric by Leveraging Pre-Trained Foundation Models

Fabric manipulation has applications in folding blankets, handling patient clothing, and protecting items with covers. It is challenging for robots to perform fabric manipulation since fabrics have infinite-dimensional configuration spaces, complex dynamics, and may be in folded or crumpled configurations with severe self-occlusions. Prior work on robotic fabric manipulation relies either on heavily engineered setups or learning-based approaches that create and train on robot-fabric interaction data. In this paper, we propose GPT-Fabric for the canonical tasks of fabric folding and smoothing, where GPT directly outputs an action informing a robot where to grasp and pull a fabric. We perform extensive experiments in simulation to test GPT-Fabric against prior state of the art methods for folding and smoothing. We obtain comparable or better performance to most methods even without explicitly training on a fabric-specific dataset (i.e., zero-shot manipulation). Furthermore, we apply GPT-Fabric in physical experiments over 12 folding and 10 smoothing rollouts. Our results suggest that GPT-Fabric is a promising approach for high-precision fabric manipulation tasks.

Self-Supervised Learning of Dynamic Planar Manipulation of Free-End Cables

Dynamic manipulation of free-end cables has applications for cable management in homes, warehouses and manufacturing plants. We present a supervised learning approach for dynamic manipulation of free-end cables, focusing on the problem of getting the cable endpoint to a designated target position, which may lie outside the reachable workspace of the robot end effector. We present a simulator, tune it to closely match experiments with physical cables, and then collect training data for learning dynamic cable manipulation. We evaluate with 3 cables and a physical UR5 robot. Results over 32x5 trials on 3 cables suggest that a physical UR5 robot can attain a median error distance ranging from 22% to 35% of the cable length among cables, outperforming an analytic baseline by 21% and a Gaussian Process baseline by 7% with lower interquartile range (IQR).

Cross-domain Multi-modal Few-shot Object Detection via Rich Text

Cross-modal feature extraction and integration have led to steady performance improvements in few-shot learning tasks due to generating richer features. However, existing multi-modal object detection (MM-OD) methods degrade when facing significant domain-shift and are sample insufficient. We hypothesize that rich text information could more effectively help the model to build a knowledge relationship between the vision instance and its language description and can help mitigate domain shift. Specifically, we study the Cross-Domain few-shot generalization of MM-OD (CDMM-FSOD) and propose a meta-learning based multi-modal few-shot object detection method that utilizes rich text semantic information as an auxiliary modality to achieve domain adaptation in the context of FSOD. Our proposed network contains (i) a multi-modal feature aggregation module that aligns the vision and language support feature embeddings and (ii) a rich text semantic rectify module that utilizes bidirectional text feature generation to reinforce multi-modal feature alignment and thus to enhance the model's language understanding capability. We evaluate our model on common standard cross-domain object detection datasets and demonstrate that our approach considerably outperforms existing FSOD methods.