Search

We propose Wolf, a WOrLd summarization Framework for accurate video captioning. Wolf is an automated captioning framework that adopts a mixture-of-experts approach, leveraging complementary strengths of Vision Language Models (VLMs). By utilizing both image and video models, our framework captures different levels of information and summarizes them efficiently. Our approach can be applied to enhance video understanding, auto-labeling, and captioning. To evaluate caption quality, we introduce CapScore, an LLM-based metric to assess the similarity and quality of generated captions compared to the ground truth captions. We further build four human-annotated datasets in three domains: autonomous driving, general scenes, and robotics, to facilitate comprehensive comparisons. We show that Wolf achieves superior captioning performance compared to state-of-the-art approaches from the research community (VILA1.5, CogAgent) and commercial solutions (Gemini-Pro-1.5, GPT-4V). For instance, in comparison with GPT-4V, Wolf improves CapScore both quality-wise by 55.6% and similarity-wise by 77.4% on challenging driving videos. Finally, we establish a benchmark for video captioning and introduce a leaderboard, aiming to accelerate advancements in video understanding, captioning, and data alignment. Leaderboard: https://wolfv0.github.io/leaderboard.html.

Visual language models (VLMs) have rapidly progressed, driven by the success of large language models (LLMs). While model architectures and training infrastructures advance rapidly, data curation remains under-explored. When data quantity and quality become a bottleneck, existing work either directly crawls more raw data from the Internet that does not have a guarantee of data quality or distills from black-box commercial models (e.g., GPT-4V / Gemini) causing the performance upper bounded by that model. In this work, we introduce a novel approach that includes a self-augment step and a specialist-augment step to iteratively improve data quality and model performance. In the self-augment step, a VLM recaptions its own pretraining data to enhance data quality, and then retrains from scratch using this refined dataset to improve model performance. This process can iterate for several rounds. Once self-augmentation saturates, we employ several specialist VLMs finetuned from the self-augmented VLM with domain-specific expertise, to further infuse specialist knowledge into the generalist VLM through task-oriented recaptioning and retraining. With the combined self-augmented and specialist-augmented training, we introduce $VILA^2$ (VILA-augmented-VILA), a VLM family that consistently improves the accuracy on a wide range of tasks over prior art, and achieves new state-of-the-art results on MMMU leaderboard among open-sourced models.

Foundation models, e.g., large language models (LLMs), trained on internet-scale data possess zero-shot generalization capabilities that make them a promising technology towards detecting and mitigating out-of-distribution failure modes of robotic systems. Fully realizing this promise, however, poses two challenges: (i) mitigating the considerable computational expense of these models such that they may be applied online, and (ii) incorporating their judgement regarding potential anomalies into a safe control framework. In this work, we present a two-stage reasoning framework: First is a fast binary anomaly classifier that analyzes observations in an LLM embedding space, which may then trigger a slower fallback selection stage that utilizes the reasoning capabilities of generative LLMs. These stages correspond to branch points in a model predictive control strategy that maintains the joint feasibility of continuing along various fallback plans to account for the slow reasoner's latency as soon as an anomaly is detected, thus ensuring safety. We show that our fast anomaly classifier outperforms autoregressive reasoning with state-of-the-art GPT models, even when instantiated with relatively small language models. This enables our runtime monitor to improve the trustworthiness of dynamic robotic systems, such as quadrotors or autonomous vehicles, under resource and time constraints. Videos illustrating our approach in both simulation and real-world experiments are available on this project page: https://sites.google.com/view/aesop-llm., Comment: Accepted to Robotics: Science and Systems (RSS) 2024

Understanding road geometry is a critical component of the autonomous vehicle (AV) stack. While high-definition (HD) maps can readily provide such information, they suffer from high labeling and maintenance costs. Accordingly, many recent works have proposed methods for estimating HD maps online from sensor data. The vast majority of recent approaches encode multi-camera observations into an intermediate representation, e.g., a bird's eye view (BEV) grid, and produce vector map elements via a decoder. While this architecture is performant, it decimates much of the information encoded in the intermediate representation, preventing downstream tasks (e.g., behavior prediction) from leveraging them. In this work, we propose exposing the rich internal features of online map estimation methods and show how they enable more tightly integrating online mapping with trajectory forecasting. In doing so, we find that directly accessing internal BEV features yields up to 73% faster inference speeds and up to 29% more accurate predictions on the real-world nuScenes dataset., Comment: 21 pages, 10 figures, 6 tables. ECCV 2024

Caves and lava tubes on the Moon and Mars are sites of geological and astrobiological interest but consist of terrain that is inaccessible with traditional robot locomotion. To support the exploration of these sites, we present ReachBot, a robot that uses extendable booms as appendages to manipulate itself with respect to irregular rock surfaces. The booms terminate in grippers equipped with microspines and provide ReachBot with a large workspace, allowing it to achieve force closure in enclosed spaces such as the walls of a lava tube. To propel ReachBot, we present a contact-before-motion planner for non-gaited legged locomotion that utilizes internal force control, similar to a multi-fingered hand, to keep its long, slender booms in tension. Motion planning also depends on finding and executing secure grips on rock features. We use a Monte Carlo simulation to inform gripper design and predict grasp strength and variability. Additionally, we use a two-step perception system to identify possible grasp locations. To validate our approach and mechanisms under realistic conditions, we deployed a single ReachBot arm and gripper in a lava tube in the Mojave Desert. The field test confirmed that ReachBot will find many targets for secure grasps with the proposed kinematic design.

The autonomous driving industry is increasingly adopting end-to-end learning from sensory inputs to minimize human biases in system design. Traditional end-to-end driving models, however, suffer from long-tail events due to rare or unseen inputs within their training distributions. To address this, we propose TOKEN, a novel Multi-Modal Large Language Model (MM-LLM) that tokenizes the world into object-level knowledge, enabling better utilization of LLM's reasoning capabilities to enhance autonomous vehicle planning in long-tail scenarios. TOKEN effectively alleviates data scarcity and inefficient tokenization by leveraging a traditional end-to-end driving model to produce condensed and semantically enriched representations of the scene, which are optimized for LLM planning compatibility through deliberate representation and reasoning alignment training stages. Our results demonstrate that TOKEN excels in grounding, reasoning, and planning capabilities, outperforming existing frameworks with a 27% reduction in trajectory L2 error and a 39% decrease in collision rates in long-tail scenarios. Additionally, our work highlights the importance of representation alignment and structured reasoning in sparking the common-sense reasoning capabilities of MM-LLMs for effective planning.

Benchmarking vision-based driving policies is challenging. On one hand, open-loop evaluation with real data is easy, but these results do not reflect closed-loop performance. On the other, closed-loop evaluation is possible in simulation, but is hard to scale due to its significant computational demands. Further, the simulators available today exhibit a large domain gap to real data. This has resulted in an inability to draw clear conclusions from the rapidly growing body of research on end-to-end autonomous driving. In this paper, we present NAVSIM, a middle ground between these evaluation paradigms, where we use large datasets in combination with a non-reactive simulator to enable large-scale real-world benchmarking. Specifically, we gather simulation-based metrics, such as progress and time to collision, by unrolling bird's eye view abstractions of the test scenes for a short simulation horizon. Our simulation is non-reactive, i.e., the evaluated policy and environment do not influence each other. As we demonstrate empirically, this decoupling allows open-loop metric computation while being better aligned with closed-loop evaluations than traditional displacement errors. NAVSIM enabled a new competition held at CVPR 2024, where 143 teams submitted 463 entries, resulting in several new insights. On a large set of challenging scenarios, we observe that simple methods with moderate compute requirements such as TransFuser can match recent large-scale end-to-end driving architectures such as UniAD. Our modular framework can potentially be extended with new datasets, data curation strategies, and metrics, and will be continually maintained to host future challenges. Our code is available at https://github.com/autonomousvision/navsim.

As natural access points to the subsurface, lava tubes and other caves have become premier targets of planetary missions for astrobiological analyses. Few existing robotic paradigms, however, are able to explore such challenging environments. ReachBot is a robot that enables navigation in planetary caves by using extendable and retractable limbs to locomote. This paper outlines the potential science return and mission operations for a notional mission that deploys ReachBot to a martian lava tube. In this work, the motivating science goals and science traceability matrix are provided to guide payload selection. A Concept of Operations (ConOps) is also developed for ReachBot, providing a framework for deployment and activities on Mars, analyzing mission risks, and developing mitigation strategies, Comment: In International Conference on Space Robotics 2024

We propose DistillNeRF, a self-supervised learning framework addressing the challenge of understanding 3D environments from limited 2D observations in autonomous driving. Our method is a generalizable feedforward model that predicts a rich neural scene representation from sparse, single-frame multi-view camera inputs, and is trained self-supervised with differentiable rendering to reconstruct RGB, depth, or feature images. Our first insight is to exploit per-scene optimized Neural Radiance Fields (NeRFs) by generating dense depth and virtual camera targets for training, thereby helping our model to learn 3D geometry from sparse non-overlapping image inputs. Second, to learn a semantically rich 3D representation, we propose distilling features from pre-trained 2D foundation models, such as CLIP or DINOv2, thereby enabling various downstream tasks without the need for costly 3D human annotations. To leverage these two insights, we introduce a novel model architecture with a two-stage lift-splat-shoot encoder and a parameterized sparse hierarchical voxel representation. Experimental results on the NuScenes dataset demonstrate that DistillNeRF significantly outperforms existing comparable self-supervised methods for scene reconstruction, novel view synthesis, and depth estimation; and it allows for competitive zero-shot 3D semantic occupancy prediction, as well as open-world scene understanding through distilled foundation model features. Demos and code will be available at https://distillnerf.github.io/.

The increasing rate of road accidents worldwide results not only in significant loss of life but also imposes billions financial burdens on societies. Current research in traffic crash frequency modeling and analysis has predominantly approached the problem as classification tasks, focusing mainly on learning-based classification or ensemble learning methods. These approaches often overlook the intricate relationships among the complex infrastructure, environmental, human and contextual factors related to traffic crashes and risky situations. In contrast, we initially propose a large-scale traffic crash language dataset, named CrashEvent, summarizing 19,340 real-world crash reports and incorporating infrastructure data, environmental and traffic textual and visual information in Washington State. Leveraging this rich dataset, we further formulate the crash event feature learning as a novel text reasoning problem and further fine-tune various large language models (LLMs) to predict detailed accident outcomes, such as crash types, severity and number of injuries, based on contextual and environmental factors. The proposed model, CrashLLM, distinguishes itself from existing solutions by leveraging the inherent text reasoning capabilities of LLMs to parse and learn from complex, unstructured data, thereby enabling a more nuanced analysis of contributing factors. Our experiments results shows that our LLM-based approach not only predicts the severity of accidents but also classifies different types of accidents and predicts injury outcomes, all with averaged F1 score boosted from 34.9% to 53.8%. Furthermore, CrashLLM can provide valuable insights for numerous open-world what-if situational-awareness traffic safety analyses with learned reasoning features, which existing models cannot offer. We make our benchmark, datasets, and model public available for further exploration.

The past few years have seen immense progress on two fronts that are critical to safe, widespread mobile robot deployment: predicting uncertain motion of multiple agents, and planning robot motion under uncertainty. However, the numerical methods required on each front have resulted in a mismatch of representation for prediction and planning. In prediction, numerical tractability is usually achieved by coarsely discretizing time, and by representing multimodal multi-agent interactions as distributions with infinite support. On the other hand, safe planning typically requires very fine time discretization, paired with distributions with compact support, to reduce conservativeness and ensure numerical tractability. The result is, when existing predictors are coupled with planning and control, one may often find unsafe motion plans. This paper proposes ZAPP (Zonotope Agreement of Prediction and Planning) to resolve the representation mismatch. ZAPP unites a prediction-friendly coarse time discretization and a planning-friendly zonotope uncertainty representation; the method also enables differentiating through a zonotope collision check, allowing one to integrate prediction and planning within a gradient-based optimization framework. Numerical examples show how ZAPP can produce safer trajectories compared to baselines in interactive scenes., Comment: 8 pages, 3 figures, 1 table, submitted to 2024 IEEE International Conference on Robotics and Automation (ICRA)

Humans naturally retain memories of permanent elements, while ephemeral moments often slip through the cracks of memory. This selective retention is crucial for robotic perception, localization, and mapping. To endow robots with this capability, we introduce 3D Gaussian Mapping (3DGM), a self-supervised, camera-only offline mapping framework grounded in 3D Gaussian Splatting. 3DGM converts multitraverse RGB videos from the same region into a Gaussian-based environmental map while concurrently performing 2D ephemeral object segmentation. Our key observation is that the environment remains consistent across traversals, while objects frequently change. This allows us to exploit self-supervision from repeated traversals to achieve environment-object decomposition. More specifically, 3DGM formulates multitraverse environmental mapping as a robust differentiable rendering problem, treating pixels of the environment and objects as inliers and outliers, respectively. Using robust feature distillation, feature residuals mining, and robust optimization, 3DGM jointly performs 2D segmentation and 3D mapping without human intervention. We build the Mapverse benchmark, sourced from the Ithaca365 and nuPlan datasets, to evaluate our method in unsupervised 2D segmentation, 3D reconstruction, and neural rendering. Extensive results verify the effectiveness and potential of our method for self-driving and robotics., Comment: Project page: https://3d-gaussian-mapping.github.io; Code and data: https://github.com/NVlabs/3DGM

Ensuring robust 3D object detection and localization is crucial for many applications in robotics and autonomous driving. Recent models, however, face difficulties in maintaining high performance when applied to domains with differing sensor setups or geographic locations, often resulting in poor localization accuracy due to domain shift. To overcome this challenge, we introduce a novel diffusion-based box refinement approach. This method employs a domain-agnostic diffusion model, conditioned on the LiDAR points surrounding a coarse bounding box, to simultaneously refine the box's location, size, and orientation. We evaluate this approach under various domain adaptation settings, and our results reveal significant improvements across different datasets, object classes and detectors.

ReachBot is a robot concept for the planetary exploration of caves and lava tubes, which are often inaccessible with traditional robot locomotion methods. It uses extendable booms as appendages, with grippers mounted at the end, to grasp irregular rock surfaces and traverse these difficult terrains. We have built a partial ReachBot prototype consisting of a single boom and gripper, mounted on a tripod. We present the details on the design and field test of this partial ReachBot prototype in a lava tube in the Mojave Desert. The technical requirements of the field testing, implementation details, and grasp performance results are discussed. The planning and preparation of the field test and lessons learned are also given., Comment: Accepted to the IEEE ICRA Workshop on Field Robotics 2024; 4 pages

Modern neural trajectory predictors in autonomous driving are developed using imitation learning (IL) from driving logs. Although IL benefits from its ability to glean nuanced and multi-modal human driving behaviors from large datasets, the resulting predictors often struggle with out-of-distribution (OOD) scenarios and with traffic rule compliance. On the other hand, classical rule-based predictors, by design, can predict traffic rule satisfying behaviors while being robust to OOD scenarios, but these predictors fail to capture nuances in agent-to-agent interactions and human driver's intent. In this paper, we present RuleFuser, a posterior-net inspired evidential framework that combines neural predictors with classical rule-based predictors to draw on the complementary benefits of both, thereby striking a balance between performance and traffic rule compliance. The efficacy of our approach is demonstrated on the real-world nuPlan dataset where RuleFuser leverages the higher performance of the neural predictor in in-distribution (ID) scenarios and the higher safety offered by the rule-based predictor in OOD scenarios., Comment: 9 pages, 3 figures

Multi-modal large language models (MLLMs) have shown incredible capabilities in a variety of 2D vision and language tasks. We extend MLLMs' perceptual capabilities to ground and reason about images in 3-dimensional space. To that end, we first develop a large-scale pre-training dataset for 2D and 3D called LV3D by combining multiple existing 2D and 3D recognition datasets under a common task formulation: as multi-turn question-answering. Next, we introduce a new MLLM named Cube-LLM and pre-train it on LV3D. We show that pure data scaling makes a strong 3D perception capability without 3D specific architectural design or training objective. Cube-LLM exhibits intriguing properties similar to LLMs: (1) Cube-LLM can apply chain-of-thought prompting to improve 3D understanding from 2D context information. (2) Cube-LLM can follow complex and diverse instructions and adapt to versatile input and output formats. (3) Cube-LLM can be visually prompted such as 2D box or a set of candidate 3D boxes from specialists. Our experiments on outdoor benchmarks demonstrate that Cube-LLM significantly outperforms existing baselines by 21.3 points of AP-BEV on the Talk2Car dataset for 3D grounded reasoning and 17.7 points on the DriveLM dataset for complex reasoning about driving scenarios, respectively. Cube-LLM also shows competitive results in general MLLM benchmarks such as refCOCO for 2D grounding with (87.0) average score, as well as visual question answering benchmarks such as VQAv2, GQA, SQA, POPE, etc. for complex reasoning. Our project is available at https://janghyuncho.github.io/Cube-LLM., Comment: Project page: https://janghyuncho.github.io/Cube-LLM

In recent years, deep reinforcement learning (RL) has shown its effectiveness in solving complex continuous control tasks like locomotion and dexterous manipulation. However, this comes at the cost of an enormous amount of experience required for training, exacerbated by the sensitivity of learning efficiency and the policy performance to hyperparameter selection, which often requires numerous trials of time-consuming experiments. This work introduces a Population-Based Reinforcement Learning (PBRL) approach that exploits a GPU-accelerated physics simulator to enhance the exploration capabilities of RL by concurrently training multiple policies in parallel. The PBRL framework is applied to three state-of-the-art RL algorithms - PPO, SAC, and DDPG - dynamically adjusting hyperparameters based on the performance of learning agents. The experiments are performed on four challenging tasks in Isaac Gym - Anymal Terrain, Shadow Hand, Humanoid, Franka Nut Pick - by analyzing the effect of population size and mutation mechanisms for hyperparameters. The results demonstrate that PBRL agents outperform non-evolutionary baseline agents across tasks essential for humanoid robots, such as bipedal locomotion, manipulation, and grasping in unstructured environments. The trained agents are finally deployed in the real world for the Franka Nut Pick manipulation task. To our knowledge, this is the first sim-to-real attempt for successfully deploying PBRL agents on real hardware. Code and videos of the learned policies are available on our project website (https://sites.google.com/view/pbrl)., Comment: Submitted for publication to IEEE-RAS 23rd International Conference on Humanoid Robots

In Model Predictive Control (MPC), discrepancies between the actual system and the predictive model can lead to substantial tracking errors and significantly degrade performance and reliability. While such discrepancies can be alleviated with more complex models, this often complicates controller design and implementation. By leveraging the fact that many trajectories of interest are periodic, we show that perfect tracking is possible when incorporating a simple observer that estimates and compensates for periodic disturbances. We present the design of the observer and the accompanying tracking MPC scheme, proving that their combination achieves zero tracking error asymptotically, regardless of the complexity of the unmodelled dynamics. We validate the effectiveness of our method, demonstrating asymptotically perfect tracking on a high-dimensional soft robot with nearly 10,000 states and a fivefold reduction in tracking errors compared to a baseline MPC on small-scale autonomous race car experiments., Comment: 8 pages, 3 figures, Submitted to the 2024 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS 2024)

While novel view synthesis (NVS) from a sparse set of images has advanced significantly in 3D computer vision, it relies on precise initial estimation of camera parameters using Structure-from-Motion (SfM). For instance, the recently developed Gaussian Splatting depends heavily on the accuracy of SfM-derived points and poses. However, SfM processes are time-consuming and often prove unreliable in sparse-view scenarios, where matched features are scarce, leading to accumulated errors and limited generalization capability across datasets. In this study, we introduce a novel and efficient framework to enhance robust NVS from sparse-view images. Our framework, InstantSplat, integrates multi-view stereo(MVS) predictions with point-based representations to construct 3D Gaussians of large-scale scenes from sparse-view data within seconds, addressing the aforementioned performance and efficiency issues by SfM. Specifically, InstantSplat generates densely populated surface points across all training views and determines the initial camera parameters using pixel-alignment. Nonetheless, the MVS points are not globally accurate, and the pixel-wise prediction from all views results in an excessive Gaussian number, yielding a overparameterized scene representation that compromises both training speed and accuracy. To address this issue, we employ a grid-based, confidence-aware Farthest Point Sampling to strategically position point primitives at representative locations in parallel. Next, we enhance pose accuracy and tune scene parameters through a gradient-based joint optimization framework from self-supervision. By employing this simplified framework, InstantSplat achieves a substantial reduction in training time, from hours to mere seconds, and demonstrates robust performance across various numbers of views in diverse datasets., Comment: Project Page: https://instantsplat.github.io/

High-definition (HD) maps have played an integral role in the development of modern autonomous vehicle (AV) stacks, albeit with high associated labeling and maintenance costs. As a result, many recent works have proposed methods for estimating HD maps online from sensor data, enabling AVs to operate outside of previously-mapped regions. However, current online map estimation approaches are developed in isolation of their downstream tasks, complicating their integration in AV stacks. In particular, they do not produce uncertainty or confidence estimates. In this work, we extend multiple state-of-the-art online map estimation methods to additionally estimate uncertainty and show how this enables more tightly integrating online mapping with trajectory forecasting. In doing so, we find that incorporating uncertainty yields up to 50% faster training convergence and up to 15% better prediction performance on the real-world nuScenes driving dataset., Comment: 14 pages, 14 figures, 6 tables. CVPR 2024

Tolling, or congestion pricing, offers a promising traffic management policy for regulating congestion, but has also attracted criticism for placing outsized financial burdens on low-income users. Credit-based congestion pricing (CBCP) and discount-based congestion pricing (DBCP) policies, which respectively provide travel credits and toll discounts to low-income users on tolled roads, have emerged as promising mechanisms for reducing traffic congestion without worsening societal inequities. However, the optimal design of CBCP and DBCP policies, as well as their relative advantages and disadvantages, remain poorly understood. To address this, we study the effects of implementing CBCP and DBCP policies to route users on a network of multi-lane highways with tolled express lanes. We formulate a non-atomic routing game framework in which a subset of eligible users is granted toll relief in the form of a fixed budget or toll discount, while the remaining ineligible users must pay out-of-pocket. We prove the existence of Nash equilibrium traffic flow patterns corresponding to any given CBCP or DBCP policy. Under the additional assumption that eligible users have time-invariant VoTs, we provide a convex program to efficiently compute these equilibria. For networks consisting of a single edge, we identify conditions under which CBCP policies outperform DBCP policies (and vice versa), in the sense of improving eligible users' access to the express lane. Finally, we present empirical results from a CBCP pilot study of the San Mateo 101 Express Lane Project in California. Our empirical results corroborate our theoretical analysis of the impact of deploying credit-based and discount-based policies, and lend insights into the sensitivity of their impact with respect to the travel demand and users' VoTs.

ReachBot, a proposed robotic platform, employs extendable booms as limbs for mobility in challenging environments, such as martian caves. When attached to the environment, ReachBot acts as a parallel robot, with reconfiguration driven by the ability to detach and re-place the booms. This ability enables manipulation-focused scientific objectives: for instance, through operating tools, or handling and transporting samples. To achieve these capabilities, we develop a two-part solution, optimizing for robustness against task uncertainty and stochastic failure modes. First, we present a mixed-integer stance planner to determine the positioning of ReachBot's booms to maximize the task wrench space about the nominal point(s). Second, we present a convex tension planner to determine boom tensions for the desired task wrenches, accounting for the probabilistic nature of microspine grasping. We demonstrate improvements in key robustness metrics from the field of dexterous manipulation, and show a large increase in the volume of the manipulation workspace. Finally, we employ Monte-Carlo simulation to validate the robustness of these methods, demonstrating good performance across a range of randomized tasks and environments, and generalization to cable-driven morphologies. We make our code available at our project webpage, https://stanfordasl.github.io/reachbot_manipulation/

Monocular SLAM has long grappled with the challenge of accurately modeling 3D geometries. Recent advances in Neural Radiance Fields (NeRF)-based monocular SLAM have shown promise, yet these methods typically focus on novel view synthesis rather than precise 3D geometry modeling. This focus results in a significant disconnect between NeRF applications, i.e., novel-view synthesis and the requirements of SLAM. We identify that the gap results from the volumetric representations used in NeRF, which are often dense and noisy. In this study, we propose a novel approach that reimagines volumetric representations through the lens of quadric forms. We posit that most scene components can be effectively represented as quadric planes. Leveraging this assumption, we reshape the volumetric representations with million of cubes by several quadric planes, which leads to more accurate and efficient modeling of 3D scenes in SLAM contexts. Our method involves two key steps: First, we use the quadric assumption to enhance coarse depth estimations obtained from tracking modules, e.g., Droid-SLAM. This step alone significantly improves depth estimation accuracy. Second, in the subsequent mapping phase, we diverge from previous NeRF-based SLAM methods that distribute sampling points across the entire volume space. Instead, we concentrate sampling points around quadric planes and aggregate them using a novel quadric-decomposed Transformer. Additionally, we introduce an end-to-end joint optimization strategy that synchronizes pose estimation with 3D reconstruction.

Robots require a semantic understanding of their surroundings to operate in an efficient and explainable way in human environments. In the literature, there has been an extensive focus on object labeling and exhaustive scene graph generation; less effort has been focused on the task of purely identifying and mapping large semantic regions. The present work proposes a method for semantic region mapping via embodied navigation in indoor environments, generating a high-level representation of the knowledge of the agent. To enable region identification, the method uses a vision-to-language model to provide scene information for mapping. By projecting egocentric scene understanding into the global frame, the proposed method generates a semantic map as a distribution over possible region labels at each location. This mapping procedure is paired with a trained navigation policy to enable autonomous map generation. The proposed method significantly outperforms a variety of baselines, including an object-based system and a pretrained scene classifier, in experiments in a photorealistic simulator., Comment: Accepted by IEEE International Conference on Robotics and Automation (ICRA 2024)

Recent years have seen a surge of artificial currency-based mechanisms in contexts where monetary instruments are deemed unfair or inappropriate, e.g., in allocating food donations to food banks, course seats to students, and, more recently, even for traffic congestion management. Yet the applicability of these mechanisms remains limited in repeated auction settings, as it is challenging for users to learn how to bid an artificial currency that has no value outside the auctions. Indeed, users must jointly learn the value of the currency in addition to how to spend it optimally. In this work, we study the problem of learning to bid in two prominent classes of artificial currency auctions: those in which currency, which users spend to obtain public resources, is only issued at the beginning of a finite period; and those where, in addition to the initial currency endowment, currency payments are redistributed to users at each time step. In the latter class, the currency has been referred to as karma, since users do not only spend karma to obtain public resources but also gain karma for yielding them. In both classes, we propose a simple learning strategy, called adaptive karma pacing, and show that this strategy a) is asymptotically optimal for a single user bidding against competing bids drawn from a stationary distribution; b) leads to convergent learning dynamics when all users adopt it; and c) constitutes an approximate Nash equilibrium as the number of users grows. Our results require a novel analysis in comparison to adaptive pacing strategies in monetary auctions, since we depart from the classical assumption that the currency has known value outside the auctions, and moreover consider that the currency is both spent and gained in the class of auctions with redistribution., Comment: Manuscript submitted for review to the 25th ACM Conference on Economics & Computation (EC'24)

The ability to differentiate through optimization problems has unlocked numerous applications, from optimization-based layers in machine learning models to complex design problems formulated as bilevel programs. It has been shown that exploiting problem structure can yield significant computation gains for optimization and, in some cases, enable distributed computation. One should expect that this structure can be similarly exploited for gradient computation. In this work, we discuss a decentralized framework for computing gradients of constraint-coupled optimization problems. First, we show that this framework results in significant computational gains, especially for large systems, and provide sufficient conditions for its validity. Second, we leverage exponential decay of sensitivities in graph-structured problems towards building a fully distributed algorithm with convergence guarantees. Finally, we use the methodology to rigorously estimate marginal emissions rates in power systems models. Specifically, we demonstrate how the distributed scheme allows for accurate and efficient estimation of these important emissions metrics on large dynamic power system models.

This work focuses on autonomous contingency planning for scientific missions by enabling rapid policy computation from any off-nominal point in the state space in the event of a delay or deviation from the nominal mission plan. Successful contingency planning involves managing risks and rewards, often probabilistically associated with actions, in stochastic scenarios. Markov Decision Processes (MDPs) are used to mathematically model decision-making in such scenarios. However, in the specific case of planetary rover traverse planning, the vast action space and long planning time horizon pose computational challenges. A bi-level MDP framework is proposed to improve computational tractability, while also aligning with existing mission planning practices and enhancing explainability and trustworthiness of AI-driven solutions. We discuss the conversion of a mission planning MDP into a bi-level MDP, and test the framework on RoverGridWorld, a modified GridWorld environment for rover mission planning. We demonstrate the computational tractability and near-optimal policies achievable with the bi-level MDP approach, highlighting the trade-offs between compute time and policy optimality as the problem's complexity grows. This work facilitates more efficient and flexible contingency planning in the context of scientific missions.

While modern best practices advocate for scalable architectures that support long-range interactions, object-centric models are yet to fully embrace these architectures. In particular, existing object-centric models for handling sequential inputs, due to their reliance on RNN-based implementation, show poor stability and capacity and are slow to train on long sequences. We introduce Parallelizable Spatiotemporal Binder or PSB, the first temporally-parallelizable slot learning architecture for sequential inputs. Unlike conventional RNN-based approaches, PSB produces object-centric representations, known as slots, for all time-steps in parallel. This is achieved by refining the initial slots across all time-steps through a fixed number of layers equipped with causal attention. By capitalizing on the parallelism induced by our architecture, the proposed model exhibits a significant boost in efficiency. In experiments, we test PSB extensively as an encoder within an auto-encoding framework paired with a wide variety of decoder options. Compared to the state-of-the-art, our architecture demonstrates stable training on longer sequences, achieves parallelization that results in a 60% increase in training speed, and yields performance that is on par with or better on unsupervised 2D and 3D object-centric scene decomposition and understanding., Comment: See project page at https://parallel-st-binder.github.io

Artificial currencies have grown in popularity in many real-world resource allocation settings, gaining traction in government benefits programs like food assistance and transit benefits programs. However, such programs are susceptible to misreporting fraud, wherein users can misreport their private attributes to gain access to more artificial currency (credits) than they are entitled to. To address the problem of misreporting fraud in artificial currency based benefits programs, we introduce an audit mechanism that induces a two-stage game between an administrator and users. In our proposed mechanism, the administrator running the benefits program can audit users at some cost and levy fines against them for misreporting their information. For this audit game, we study the natural solution concept of a signaling game equilibrium and investigate conditions on the administrator budget to establish the existence of equilibria. The computation of equilibria can be done via linear programming in our problem setting through an appropriate design of the audit rules. Our analysis also provides upper bounds that hold in any signaling game equilibrium on the expected excess payments made by the administrator and the probability that users misreport their information. We further show that the decrease in misreporting fraud corresponding to our audit mechanism far outweighs the administrator spending to run it by establishing that its total costs are lower than that of the status quo with no audits. Finally, to highlight the practical viability of our audit mechanism in mitigating misreporting fraud, we present a case study based on the Washington D.C. federal transit benefits program. In this case study, the proposed audit mechanism achieves several orders of magnitude improvement in total cost compared to a no-audit strategy for some parameter ranges.

Fraudulent or illegal activities are ubiquitous across applications and involve users bypassing the rule of law, often with the strategic aim of obtaining some benefit that would otherwise be unattainable within the bounds of lawful conduct. However, user fraud is detrimental, as it may compromise safety or impose disproportionate negative externalities on particular population groups. To mitigate the potential harms of user fraud, we study the problem of policing such fraud as a security game between an administrator and users. In this game, an administrator deploys $R$ security resources (e.g., police officers) across $L$ locations and levies fines against users engaging in fraud at those locations. For this security game, we study both welfare and revenue maximization administrator objectives. In both settings, we show that computing the optimal administrator strategy is NP-hard and develop natural greedy algorithm variants for the respective settings that achieve at least half the welfare or revenue as the welfare-maximizing or revenue-maximizing solutions, respectively. We also establish a resource augmentation guarantee that our proposed greedy algorithms with one extra resource, i.e., $R+1$ resources, achieve at least the same welfare (revenue) as the welfare-maximizing (revenue-maximizing) outcome with $R$ resources. Finally, since the welfare and revenue-maximizing solutions can differ significantly, we present a framework inspired by contract theory, wherein a revenue-maximizing administrator is compensated through contracts for the welfare it contributes. Beyond extending our theoretical results in the welfare and revenue maximization settings to studying equilibrium strategies in the contract game, we also present numerical experiments highlighting the efficacy of contracts in bridging the gap between the revenue and welfare-maximizing administrator outcomes.

We discuss guidelines for evaluating the performance of parameterized stochastic solvers for optimization problems, with particular attention to systems that employ novel hardware, such as digital quantum processors running variational algorithms, analog processors performing quantum annealing, or coherent Ising Machines. We illustrate through an example a benchmarking procedure grounded in the statistical analysis of the expectation of a given performance metric measured in a test environment. In particular, we discuss the necessity and cost of setting parameters that affect the algorithm's performance. The optimal value of these parameters could vary significantly between instances of the same target problem. We present an open-source software package that facilitates the design, evaluation, and visualization of practical parameter tuning strategies for complex use of the heterogeneous components of the solver. We examine in detail an example using parallel tempering and a simulator of a photonic Coherent Ising Machine computing and display the scoring of an illustrative baseline family of parameter-setting strategies that feature an exploration-exploitation trade-off., Comment: 13 pages, 6 figures

Adapting driving behavior to new environments, customs, and laws is a long-standing problem in autonomous driving, precluding the widespread deployment of autonomous vehicles (AVs). In this paper, we present LLaDA, a simple yet powerful tool that enables human drivers and autonomous vehicles alike to drive everywhere by adapting their tasks and motion plans to traffic rules in new locations. LLaDA achieves this by leveraging the impressive zero-shot generalizability of large language models (LLMs) in interpreting the traffic rules in the local driver handbook. Through an extensive user study, we show that LLaDA's instructions are useful in disambiguating in-the-wild unexpected situations. We also demonstrate LLaDA's ability to adapt AV motion planning policies in real-world datasets; LLaDA outperforms baseline planning approaches on all our metrics. Please check our website for more details: https://boyiliee.github.io/llada., Comment: CVPR 2024, featured in GTC 2024: https://www.youtube.com/watch?v=t-UPlPlrYgQ&t=51s

Generalization remains a central challenge in machine learning. In this work, we propose Learning from Teaching (LoT), a novel regularization technique for deep neural networks to enhance generalization. Inspired by the human ability to capture concise and abstract patterns, we hypothesize that generalizable correlations are expected to be easier to imitate. LoT operationalizes this concept to improve generalization of the main model with auxiliary student learners. The student learners are trained by the main model and, in turn, provide feedback to help the main model capture more generalizable and imitable correlations. Our experimental results across several domains, including Computer Vision, Natural Language Processing, and methodologies like Reinforcement Learning, demonstrate that the introduction of LoT brings significant benefits compared to training models on the original dataset. The results suggest the effectiveness and efficiency of LoT in identifying generalizable information at the right scales while discarding spurious data correlations, thus making LoT a valuable addition to current machine learning. Code is available at https://github.com/jincan333/LoT.

The Coherent Ising Machine (CIM) is a non-conventional architecture that takes inspiration from physical annealing processes to solve Ising problems heuristically. Its dynamics are naturally continuous and described by a set of ordinary differential equations that have been proven to be useful for the optimization of continuous variables non-convex quadratic optimization problems. The dynamics of such Continuous Variable CIMs (CV-CIM) encourage optimization via optical pulses whose amplitudes are determined by the negative gradient of the objective; however, standard gradient descent is known to be trapped by local minima and hampered by poor problem conditioning. In this work, we propose to modify the CV-CIM dynamics using more sophisticated pulse injections based on tried-and-true optimization techniques such as momentum and Adam. Through numerical experiments, we show that the momentum and Adam updates can significantly speed up the CV-CIM's convergence and improve sample diversity over the original CV-CIM dynamics. We also find that the Adam-CV-CIM's performance is more stable as a function of feedback strength, especially on poorly conditioned instances, resulting in an algorithm that is more robust, reliable, and easily tunable. More broadly, we identify the CIM dynamical framework as a fertile opportunity for exploring the intersection of classical optimization and modern analog computing.

To extend the limited scope of autonomy used in prior missions for operation in distant and complex environments, there is a need to further develop and mature autonomy that jointly reasons over multiple subsystems, which we term system-level autonomy. System-level autonomy establishes situational awareness that resolves conflicting information across subsystems, which may necessitate the refinement and interconnection of the underlying spacecraft and environment onboard models. However, with a limited understanding of the assumptions and tradeoffs of modeling to arbitrary extents, designing onboard models to support system-level capabilities presents a significant challenge. In this paper, we provide a detailed analysis of the increasing levels of model fidelity for several key spacecraft subsystems, with the goal of informing future spacecraft functional- and system-level autonomy algorithms and the physics-based simulators on which they are validated. We do not argue for the adoption of a particular fidelity class of models but, instead, highlight the potential tradeoffs and opportunities associated with the use of models for onboard autonomy and in physics-based simulators at various fidelity levels. We ground our analysis in the context of deep space exploration of small bodies, an emerging frontier for autonomous spacecraft operation in space, where the choice of models employed onboard the spacecraft may determine mission success. We conduct our experiments in the Multi-Spacecraft Concept and Autonomy Tool (MuSCAT), a software suite for developing spacecraft autonomy algorithms., Comment: Project page: https://sites.google.com/stanford.edu/spacecraft-models. 20 pages, 8 figures. Accepted to the IEEE Conference on Aerospace (AeroConf) 2024