How does a human being distinguish the motions of a piece of paper and a piece of cloth? A high-school physics teacher might answer that they are both tangentially inextensible but cloth cannot resist any bending force from the normal direction.

Recent advances in video generation models have sparked interest in world models capable of simulating realistic environments. While navigation has been well-explored, physically meaningful interactions that mimic real-world forces remain largely understudied. In this work, we investigate using physical forces as a control signal for video generation and propose force prompts which enable users to interact with images through both localized point forces, such as poking a plant, and global wind force fields, such as wind blowing on fabric. We demonstrate that these force prompts can enable videos to respond realistically to physical control signals by leveraging the visual and motion prior in the original pretrained model, without using any 3D asset or physics simulator at inference. The primary challenge of force prompting is the difficulty in obtaining high quality paired force-video training data, both in the real world due to the difficulty of obtaining force signals, and in synthetic data due to limitations in the visual quality and domain diversity of physics simulators. Our key finding is that video generation models can generalize remarkably well when adapted to follow physical force conditioning from videos synthesized by Blender, even with limited demonstrations of few objects. Our method can generate videos which simulate forces across diverse geometries, settings, and materials. We also try to understand the source of this generalization and perform ablations that reveal two key elements: visual diversity and the use of specific text keywords during training. Our approach is trained on only around 15k training examples for a single day on four A100 GPUs, and outperforms existing methods on force adherence and physics realism, bringing world models closer to real-world physics interactions. We release all datasets, code, weights, and interactive video demos at our project page.

From just a glance, humans can make rich predictions about the future of a wide range of physical systems. On the other hand, modern approaches from engineering, robotics, and graphics are often restricted to narrow domains or require information about the underlying state. We introduce the Visual Interaction Network, a general-purpose model for learning the dynamics of a physical system from raw visual observations. Our model consists of a perceptual front-end based on convolutional neural networks and a dynamics predictor based on interaction networks. Through joint training, the perceptual front-end learns to parse a dynamic visual scene into a set of factored latent object representations. The dynamics predictor learns to roll these states forward in time by computing their interactions, producing a predicted physical trajectory of arbitrary length. We found that from just six input video frames the Visual Interaction Network can generate accurate future trajectories of hundreds of time steps on a wide range of physical systems. Our model can also be applied to scenes with invisible objects, inferring their future states from their effects on the visible objects, and can implicitly infer the unknown mass of objects. This work opens new opportunities for model-based decision-making and planning from raw sensory observations in complex physical environments.

Abstract-- T ask and Motion Planning (T AMP) integrates high-level task planning with low-level motion feasibility, but existing methods are costly in long-horizon problems due to excessive motion sampling. While LLMs provide commonsense priors, they lack 3D spatial reasoning and cannot ensure geometric or dynamic feasibility. We propose a kinodynamic T AMP framework based on a hybrid state tree that uniformly represents symbolic and numeric states during planning, enabling task and motion decisions to be jointly decided. Kinodynamic constraints embedded in the T AMP problem are verified by an off-the-shelf motion planner and physics simulator, and a VLM guides exploring a T AMP solution and backtracks the search based on visual rendering of the states. I. INTRODUCTION Robotic manipulation tasks, such as tabletop manipulations, require reasoning over both symbolic task decisions and continuous geometric feasibility. A robot must decide which action to perform--such as picking, placing, or stacking-- and which object to grasp, which constitutes a discrete search process. Simultaneously, it must determine grasp poses, feasible end-effector configurations, and collision-free motion trajectories governed by continuous constraints. This class of problems is studied under the framework of Task and Motion Planning (i.e., T AMP), which combines high-level task planning with continuous action parameter binding and low-level motion planning [1], [2].

We propose a new family of neural networks to predict the behaviors of physical systems by learning their underpinning constraints.

--Deep object pose estimators are notoriously overconfident. A grasping agent that both estimates the 6-DoF pose of a target object and predicts the uncertainty of its own estimate could avoid task failure by choosing not to act under high uncertainty. Even though object pose estimation improves and uncertainty quantification research continues to make strides, few studies have connected them to the downstream task of robotic grasping. We propose a method for training lightweight, deep networks to predict whether a grasp guided by an image-based pose estimate will succeed before that grasp is attempted. We generate training data for our networks via object pose estimation on real images and simulated grasping. We also find that, despite high object variability in grasping trials, networks benefit from training on all objects jointly, suggesting that a diverse variety of objects can nevertheless contribute to the same goal. Remarkable progress in object pose estimation from single RGB images has been made in the past few years [1]-[4], primarily driven by deep learning and the ability to reduce the so-called sim2real gap . This has enabled end-to-end system training on large amounts of synthetic data with precise ground truth. Consider for example the pose estimates illustrated in Figure 1. These were made by current methods, yet all four caused grasping attempts to fail when used as guides. Motivated by this disconnect between pose evaluation and success in downstream grasping, we propose an approach to estimate the likelihood for success before a grasp is actually attempted.

Computer-aided design (CAD) is a promising application area for emerging artificial intelligence methods. Traditional workflows for cyberphysical systems create detailed digital models which can be evaluated by physics simulators in order to narrow the search space before creating physical prototypes. A major bottleneck of this approach is that the simulators are often computationally expensive and slow. Recent advancements in AI methods offer the possibility to accelerate these pipelines. We use the recently released AircraftVerse dataset, which is especially suited for developing and evaluating large language models for designs. AircraftVerse contains a diverse set of UAV designs represented via textual design trees together with detailed physics simulation results. Following the recent success of large language models (LLMs), we propose AGENT (Aircraft GENeraTor). AGENT is a comprehensive design tool built on the CodeT5+ LLM which learns powerful representations of aircraft textual designs directly from JSON files. We develop a curriculum of training tasks which imbues a single model with a suite of useful features. AGENT is able to generate designs conditioned on properties of flight dynamics (hover time, maximum speed, etc.). Additionally, AGENT can issue evaluations of designs allowing it to act as a surrogate model of the physics simulation that underlies the AircraftVerse dataset. We present a series of experiments which demonstrate our system's abilities. We are able to achieve strong performance using the smallest member of the CodeT5+ family (220M parameters). This allows for a flexible and powerful system which can be executed on a single GPU enabling a clear path toward future deployment.

Traffic simulation is essential for autonomous vehicle (AV) development, enabling comprehensive safety evaluation across diverse driving conditions. However, traditional rule-based simulators struggle to capture complex human interactions, while data-driven approaches often fail to maintain long-term behavioral realism or generate diverse safety-critical events. To address these challenges, we propose TeraSim, an open-source, high-fidelity traffic simulation platform designed to uncover unknown unsafe events and efficiently estimate AV statistical performance metrics, such as crash rates. TeraSim is designed for seamless integration with third-party physics simulators and standalone AV stacks, to construct a complete AV simulation system. Experimental results demonstrate its effectiveness in generating diverse safety-critical events involving both static and dynamic agents, identifying hidden deficiencies in AV systems, and enabling statistical performance evaluation. These findings highlight TeraSim's potential as a practical tool for AV safety assessment, benefiting researchers, developers, and policymakers. The code is available at https://github.com/mcity/TeraSim.

Meshes are used to represent complex objects in high fidelity physics simulators across a variety of domains, such as radar sensing and aerodynamics. There is growing interest in using neural networks to accelerate physics simulations, and also a growing body of work on applying neural networks directly to irregular mesh data. Since multiple mesh topologies can represent the same object, mesh augmentation is typically required to handle topological variation when training neural networks. Due to the sensitivity of physics simulators to small changes in mesh shape, it is challenging to use these augmentations when training neural network-based physics simulators. In this work, we show that variations in mesh topology can significantly reduce the performance of neural network simulators. We evaluate whether pretraining can be used to address this issue, and find that employing an established autoencoder pretraining technique with graph embedding models reduces the sensitivity of neural network simulators to variations in mesh topology. Finally, we highlight future research directions that may further reduce neural simulator sensitivity to mesh topology.

This paper introduces the TactiMesh Teleoperator Interface (TTI), a novel predictive visual and haptic system designed explicitly for human-in-the-loop robot control using a head-mounted display (HMD). By employing simultaneous localization and mapping (SLAM)in tandem with a space carving method (CARV), TTI creates a real time 3D surface mesh of remote environments from an RGB camera mounted on a Barrett WAM arm. The generated mesh is integrated into a physics simulator, featuring a digital twin of the WAM robot arm to create a virtual environment. In this virtual environment, TTI provides haptic feedback directly in response to the operator's movements, eliminating the problem with delayed response from the haptic follower robot. Furthermore, texturing the 3D mesh with keyframes from SLAM allows the operator to control the viewpoint of their Head Mounted Display (HMD) independently of the arm-mounted robot camera, giving a better visual immersion and improving manipulation speed. Incorporating predictive visual and haptic feedback significantly improves teleoperation in applications such as search and rescue, inspection, and remote maintenance.