Search

Recovering metric depth from a single image remains a fundamental challenge in computer vision, requiring both scene understanding and accurate scaling. While deep learning has advanced monocular depth estimation, current models often struggle with unfamiliar scenes and layouts, particularly in zero-shot scenarios and when predicting scale-ergodic metric depth. We present MetricGold, a novel approach that harnesses generative diffusion model's rich priors to improve metric depth estimation. Building upon recent advances in MariGold, DDVM and Depth Anything V2 respectively, our method combines latent diffusion, log-scaled metric depth representation, and synthetic data training. MetricGold achieves efficient training on a single RTX 3090 within two days using photo-realistic synthetic data from HyperSIM, VirtualKitti, and TartanAir. Our experiments demonstrate robust generalization across diverse datasets, producing sharper and higher quality metric depth estimates compared to existing approaches., Comment: arXiv admin note: substantial text overlap with arXiv:2312.02145 by other authors

In autonomous driving with image based state space, accurate prediction of future events and modeling diverse behavioral modes are essential for safety and effective decision-making. World model-based Reinforcement Learning (WMRL) approaches offers a promising solution by simulating future states from current state and actions. However, utility of world models is often limited by typical RL policies being limited to deterministic or single gaussian distribution. By failing to capture the full spectrum of possible actions, reduces their adaptability in complex, dynamic environments. In this work, we introduce Imagine-2-Drive, a framework that consists of two components, VISTAPlan, a high-fidelity world model for accurate future prediction and Diffusion Policy Actor (DPA), a diffusion based policy to model multi-modal behaviors for trajectory prediction. We use VISTAPlan to simulate and evaluate trajectories from DPA and use Denoising Diffusion Policy Optimization (DDPO) to train DPA to maximize the cumulative sum of rewards over the trajectories. We analyze the benefits of each component and the framework as a whole in CARLA with standard driving metrics. As a consequence of our twin novelties- VISTAPlan and DPA, we significantly outperform the state of the art (SOTA) world models on standard driving metrics by 15% and 20% on Route Completion and Success Rate respectively., Comment: Submitted to ICRA 2025

Dual-arm manipulation is an area of growing interest in the robotics community. Enabling robots to perform tasks that require the coordinated use of two arms, is essential for complex manipulation tasks such as handling large objects, assembling components, and performing human-like interactions. However, achieving effective dual-arm manipulation is challenging due to the need for precise coordination, dynamic adaptability, and the ability to manage interaction forces between the arms and the objects being manipulated. We propose a novel pipeline that combines the advantages of policy learning based on environment feedback and gradient-based optimization to learn controller gains required for the control outputs. This allows the robotic system to dynamically modulate its impedance in response to task demands, ensuring stability and dexterity in dual-arm operations. We evaluate our pipeline on a trajectory-tracking task involving a variety of large, complex objects with different masses and geometries. The performance is then compared to three other established methods for controlling dual-arm robots, demonstrating superior results.

Visual servoing, the method of controlling robot motion through feedback from visual sensors, has seen significant advancements with the integration of optical flow-based methods. However, its application remains limited by inherent challenges, such as the necessity for a target image at test time, the requirement of substantial overlap between initial and target images, and the reliance on feedback from a single camera. This paper introduces Imagine2Servo, an innovative approach leveraging diffusion-based image editing techniques to enhance visual servoing algorithms by generating intermediate goal images. This methodology allows for the extension of visual servoing applications beyond traditional constraints, enabling tasks like long-range navigation and manipulation without predefined goal images. We propose a pipeline that synthesizes subgoal images grounded in the task at hand, facilitating servoing in scenarios with minimal initial and target image overlap and integrating multi-camera feedback for comprehensive task execution. Our contributions demonstrate a novel application of image generation to robotic control, significantly broadening the capabilities of visual servoing systems. Real-world experiments validate the effectiveness and versatility of the Imagine2Servo framework in accomplishing a variety of tasks, marking a notable advancement in the field of visual servoing.

Navigation amongst densely packed crowds remains a challenge for mobile robots. The complexity increases further if the environment layout changes, making the prior computed global plan infeasible. In this paper, we show that it is possible to dramatically enhance crowd navigation by just improving the local planner. Our approach combines generative modelling with inference time optimization to generate sophisticated long-horizon local plans at interactive rates. More specifically, we train a Vector Quantized Variational AutoEncoder to learn a prior over the expert trajectory distribution conditioned on the perception input. At run-time, this is used as an initialization for a sampling-based optimizer for further refinement. Our approach does not require any sophisticated prediction of dynamic obstacles and yet provides state-of-the-art performance. In particular, we compare against the recent DRL-VO approach and show a 40% improvement in success rate and a 6% improvement in travel time.

Re-identification (ReID) is a critical challenge in computer vision, predominantly studied in the context of pedestrians and vehicles. However, robust object-instance ReID, which has significant implications for tasks such as autonomous exploration, long-term perception, and scene understanding, remains underexplored. In this work, we address this gap by proposing a novel dual-path object-instance re-identification transformer architecture that integrates multimodal RGB and depth information. By leveraging depth data, we demonstrate improvements in ReID across scenes that are cluttered or have varying illumination conditions. Additionally, we develop a ReID-based localization framework that enables accurate camera localization and pose identification across different viewpoints. We validate our methods using two custom-built RGB-D datasets, as well as multiple sequences from the open-source TUM RGB-D datasets. Our approach demonstrates significant improvements in both object instance ReID (mAP of 75.18) and localization accuracy (success rate of 83% on TUM-RGBD), highlighting the essential role of object ReID in advancing robotic perception. Our models, frameworks, and datasets have been made publicly available., Comment: 8 pages, 5 figures, 3 tables. Submitted to ICRA 2025

Humans excel at forming mental maps of their surroundings, equipping them to understand object relationships and navigate based on language queries. Our previous work SI Maps [1] showed that having instance-level information and the semantic understanding of an environment helps significantly improve performance for language-guided tasks. We extend this instance-level approach to 3D while increasing the pipeline's robustness and improving quantitative and qualitative results. Our method leverages foundational models for object recognition, image segmentation, and feature extraction. We propose a representation that results in a 3D point cloud map with instance-level embeddings, which bring in the semantic understanding that natural language commands can query. Quantitatively, the work improves upon the success rate of language-guided tasks. At the same time, we qualitatively observe the ability to identify instances more clearly and leverage the foundational models and language and image-aligned embeddings to identify objects that, otherwise, a closed-set approach wouldn't be able to identify.

Efficiently generating grasp poses tailored to specific regions of an object is vital for various robotic manipulation tasks, especially in a dual-arm setup. This scenario presents a significant challenge due to the complex geometries involved, requiring a deep understanding of the local geometry to generate grasps efficiently on the specified constrained regions. Existing methods only explore settings involving table-top/small objects and require augmented datasets to train, limiting their performance on complex objects. We propose CGDF: Constrained Grasp Diffusion Fields, a diffusion-based grasp generative model that generalizes to objects with arbitrary geometries, as well as generates dense grasps on the target regions. CGDF uses a part-guided diffusion approach that enables it to get high sample efficiency in constrained grasping without explicitly training on massive constraint-augmented datasets. We provide qualitative and quantitative comparisons using analytical metrics and in simulation, in both unconstrained and constrained settings to show that our method can generalize to generate stable grasps on complex objects, especially useful for dual-arm manipulation settings, while existing methods struggle to do so., Comment: Project Page: https://constrained-grasp-diffusion.github.io/

Navigation of wheeled vehicles on uneven terrain necessitates going beyond the 2D approaches for trajectory planning. Specifically, it is essential to incorporate the full 6dof variation of vehicle pose and its associated stability cost in the planning process. To this end, most recent works aim to learn a neural network model to predict the vehicle evolution. However, such approaches are data-intensive and fraught with generalization issues. In this paper, we present a purely model-based approach that just requires the digital elevation information of the terrain. Specifically, we express the wheel-terrain interaction and 6dof pose prediction as a non-linear least squares (NLS) problem. As a result, trajectory planning can be viewed as a bi-level optimization. The inner optimization layer predicts the pose on the terrain along a given trajectory, while the outer layer deforms the trajectory itself to reduce the stability and kinematic costs of the pose. We improve the state-of-the-art in the following respects. First, we show that our NLS based pose prediction closely matches the output from a high-fidelity physics engine. This result coupled with the fact that we can query gradients of the NLS solver, makes our pose predictor, a differentiable wheel-terrain interaction model. We further leverage this differentiability to efficiently solve the proposed bi-level trajectory optimization problem. Finally, we perform extensive experiments, and comparison with a baseline to showcase the effectiveness of our approach in obtaining smooth, stable trajectories., Comment: 8 pages, 7 figures, submitted to IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS 2024)

Existing Vision-Language models (VLMs) estimate either long-term trajectory waypoints or a set of control actions as a reactive solution for closed-loop planning based on their rich scene comprehension. However, these estimations are coarse and are subjective to their "world understanding" which may generate sub-optimal decisions due to perception errors. In this paper, we introduce LeGo-Drive, which aims to address this issue by estimating a goal location based on the given language command as an intermediate representation in an end-to-end setting. The estimated goal might fall in a non-desirable region, like on top of a car for a parking-like command, leading to inadequate planning. Hence, we propose to train the architecture in an end-to-end manner, resulting in iterative refinement of both the goal and the trajectory collectively. We validate the effectiveness of our method through comprehensive experiments conducted in diverse simulated environments. We report significant improvements in standard autonomous driving metrics, with a goal reaching Success Rate of 81%. We further showcase the versatility of LeGo-Drive across different driving scenarios and linguistic inputs, underscoring its potential for practical deployment in autonomous vehicles and intelligent transportation systems.

Point cloud prediction is an important yet challenging task in the field of autonomous driving. The goal is to predict future point cloud sequences that maintain object structures while accurately representing their temporal motion. These predicted point clouds help in other subsequent tasks like object trajectory estimation for collision avoidance or estimating locations with the least odometry drift. In this work, we present ATPPNet, a novel architecture that predicts future point cloud sequences given a sequence of previous time step point clouds obtained with LiDAR sensor. ATPPNet leverages Conv-LSTM along with channel-wise and spatial attention dually complemented by a 3D-CNN branch for extracting an enhanced spatio-temporal context to recover high quality fidel predictions of future point clouds. We conduct extensive experiments on publicly available datasets and report impressive performance outperforming the existing methods. We also conduct a thorough ablative study of the proposed architecture and provide an application study that highlights the potential of our model for tasks like odometry estimation., Comment: Accepted for presentation at the 2024 IEEE International Conference on Robotics and Automation (ICRA)

Despite the technological advancements in the construction and surveying sector, the inspection of salient features like windows in an under-construction or existing building is predominantly a manual process. Moreover, the number of windows present in a building is directly related to the magnitude of deformation it suffers under earthquakes. In this research, a method to accurately detect and count the number of windows of a building by deploying an Unmanned Aerial Vehicle (UAV) based remote sensing system is proposed. The proposed two-stage method automates the identification and counting of windows by developing computer vision pipelines that utilize data from UAV's onboard camera and other sensors. Quantitative and Qualitative results show the effectiveness of our proposed approach in accurately detecting and counting the windows compared to the existing method.

In this paper we show an effective means of integrating data driven frameworks to sampling based optimal control to vastly reduce the compute time for easy adoption and adaptation to real time applications such as on-road autonomous driving in the presence of dynamic actors. Presented with training examples, a spatio-temporal CNN learns to predict the optimal mean control over a finite horizon that precludes further resampling, an iterative process that makes sampling based optimal control formulations difficult to adopt in real time settings. Generating control samples around the network-predicted optimal mean retains the advantage of sample diversity while enabling real time rollout of trajectories that avoids multiple dynamic obstacles in an on-road navigation setting. Further the 3D CNN architecture implicitly learns the future trajectories of the dynamic agents in the scene resulting in successful collision free navigation despite no explicit future trajectory prediction. We show performance gain over multiple baselines in a number of on-road scenes through closed loop simulations in CARLA. We also showcase the real world applicability of our system by running it on our custom Autonomous Driving Platform (AutoDP)., Comment: Published in 2023 IEEE 19th International Conference on Automation Science and Engineering (CASE)

Safe autonomous driving critically depends on how well the ego-vehicle can predict the trajectories of neighboring vehicles. To this end, several trajectory prediction algorithms have been presented in the existing literature. Many of these approaches output a multi-modal distribution of obstacle trajectories instead of a single deterministic prediction to account for the underlying uncertainty. However, existing planners cannot handle the multi-modality based on just sample-level information of the predictions. With this motivation, this paper proposes a trajectory optimizer that can leverage the distributional aspects of the prediction in a computationally tractable and sample-efficient manner. Our optimizer can work with arbitrarily complex distributions and thus can be used with output distribution represented as a deep neural network. The core of our approach is built on embedding distribution in Reproducing Kernel Hilbert Space (RKHS), which we leverage in two ways. First, we propose an RKHS embedding approach to select probable samples from the obstacle trajectory distribution. Second, we rephrase chance-constrained optimization as distribution matching in RKHS and propose a novel sampling-based optimizer for its solution. We validate our approach with hand-crafted and neural network-based predictors trained on real-world datasets and show improvement over the existing stochastic optimization approaches in safety metrics.

Semantic segmentation in adverse weather scenarios is a critical task for autonomous driving systems. While foundation models have shown promise, the need for specialized adaptors becomes evident for handling more challenging scenarios. We introduce DiffPrompter, a novel differentiable visual and latent prompting mechanism aimed at expanding the learning capabilities of existing adaptors in foundation models. Our proposed $\nabla$HFC image processing block excels particularly in adverse weather conditions, where conventional methods often fall short. Furthermore, we investigate the advantages of jointly training visual and latent prompts, demonstrating that this combined approach significantly enhances performance in out-of-distribution scenarios. Our differentiable visual prompts leverage parallel and series architectures to generate prompts, effectively improving object segmentation tasks in adverse conditions. Through a comprehensive series of experiments and evaluations, we provide empirical evidence to support the efficacy of our approach. Project page at https://diffprompter.github.io.

Talk2BEV is a large vision-language model (LVLM) interface for bird's-eye view (BEV) maps in autonomous driving contexts. While existing perception systems for autonomous driving scenarios have largely focused on a pre-defined (closed) set of object categories and driving scenarios, Talk2BEV blends recent advances in general-purpose language and vision models with BEV-structured map representations, eliminating the need for task-specific models. This enables a single system to cater to a variety of autonomous driving tasks encompassing visual and spatial reasoning, predicting the intents of traffic actors, and decision-making based on visual cues. We extensively evaluate Talk2BEV on a large number of scene understanding tasks that rely on both the ability to interpret free-form natural language queries, and in grounding these queries to the visual context embedded into the language-enhanced BEV map. To enable further research in LVLMs for autonomous driving scenarios, we develop and release Talk2BEV-Bench, a benchmark encompassing 1000 human-annotated BEV scenarios, with more than 20,000 questions and ground-truth responses from the NuScenes dataset., Comment: Project page at https://llmbev.github.io/talk2bev/

Neural Radiance Fields (NeRF) have become an increasingly popular representation to capture high-quality appearance and shape of scenes and objects. However, learning generalizable NeRF priors over categories of scenes or objects has been challenging due to the high dimensionality of network weight space. To address the limitations of existing work on generalization, multi-view consistency and to improve quality, we propose HyP-NeRF, a latent conditioning method for learning generalizable category-level NeRF priors using hypernetworks. Rather than using hypernetworks to estimate only the weights of a NeRF, we estimate both the weights and the multi-resolution hash encodings resulting in significant quality gains. To improve quality even further, we incorporate a denoise and finetune strategy that denoises images rendered from NeRFs estimated by the hypernetwork and finetunes it while retaining multiview consistency. These improvements enable us to use HyP-NeRF as a generalizable prior for multiple downstream tasks including NeRF reconstruction from single-view or cluttered scenes and text-to-NeRF. We provide qualitative comparisons and evaluate HyP-NeRF on three tasks: generalization, compression, and retrieval, demonstrating our state-of-the-art results., Comment: Project Page: https://hyp-nerf.github.io

Autonomous driving requires accurate reasoning of the location of objects from raw sensor data. Recent end-to-end learning methods go from raw sensor data to a trajectory output via Bird's Eye View(BEV) segmentation as an interpretable intermediate representation. Motion planning over cost maps generated via Birds Eye View (BEV) segmentation has emerged as a prominent approach in autonomous driving. However, the current approaches have two critical gaps. First, the optimization process is simplistic and involves just evaluating a fixed set of trajectories over the cost map. The trajectory samples are not adapted based on their associated cost values. Second, the existing cost maps do not account for the uncertainty in the cost maps that can arise due to noise in RGB images, and BEV annotations. As a result, these approaches can struggle in challenging scenarios where there is abrupt cut-in, stopping, overtaking, merging, etc from the neighboring vehicles. In this paper, we propose UAP-BEV: A novel approach that models the noise in Spatio-Temporal BEV predictions to create an uncertainty-aware occupancy grid map. Using queries of the distance to the closest occupied cell, we obtain a sample estimate of the collision probability of the ego-vehicle. Subsequently, our approach uses gradient-free sampling-based optimization to compute low-cost trajectories over the cost map. Importantly, the sampling distribution is adapted based on the optimal cost values of the sampled trajectories. By explicitly modeling probabilistic collision avoidance in the BEV space, our approach is able to outperform the cost-map-based baselines in collision avoidance, route completion, time to completion, and smoothness. To further validate our method, we also show results on the real-world dataset NuScenes, where we report improvements in collision avoidance and smoothness., Comment: Accepted to CASE 2023. Project video available at https://vikr-182.github.io/UAP-BEV

Humans have a natural ability to perform semantic associations with the surrounding objects in the environment. This allows them to create a mental map of the environment, allowing them to navigate on-demand when given linguistic instructions. A natural goal in Vision Language Navigation (VLN) research is to impart autonomous agents with similar capabilities. Recent works take a step towards this goal by creating a semantic spatial map representation of the environment without any labeled data. However, their representations are limited for practical applicability as they do not distinguish between different instances of the same object. In this work, we address this limitation by integrating instance-level information into spatial map representation using a community detection algorithm and utilizing word ontology learned by large language models (LLMs) to perform open-set semantic associations in the mapping representation. The resulting map representation improves the navigation performance by two-fold (233%) on realistic language commands with instance-specific descriptions compared to the baseline. We validate the practicality and effectiveness of our approach through extensive qualitative and quantitative experiments.

We focus on the problem of LiDAR point cloud based loop detection (or Finding) and closure (LDC) in a multi-agent setting. State-of-the-art (SOTA) techniques directly generate learned embeddings of a given point cloud, require large data transfers, and are not robust to wide variations in 6 Degrees-of-Freedom (DOF) viewpoint. Moreover, absence of strong priors in an unstructured point cloud leads to highly inaccurate LDC. In this original approach, we propose independent roll and pitch canonicalization of the point clouds using a common dominant ground plane. Discretization of the canonicalized point cloud along the axis perpendicular to the ground plane leads to an image similar to Digital Elevation Maps (DEMs), which exposes strong spatial priors in the scene. Our experiments show that LDC based on learnt embeddings of such DEMs is not only data efficient but also significantly more robust, and generalizable than the current SOTA. We report significant performance gain in terms of Average Precision for loop detection and absolute translation/rotation error for relative pose estimation (or loop closure) on Kitti, GPR and Oxford Robot Car over multiple SOTA LDC methods. Our encoder technique allows to compress the original point cloud by over 830 times. To further test the robustness of our technique we create and opensource a custom dataset called Lidar-UrbanFly Dataset (LUF) which consists of point clouds obtained from a LiDAR mounted on a quadrotor.

In this paper, we propose and showcase, for the first time, monocular multi-view layout estimation for warehouse racks and shelves. Unlike typical layout estimation methods, MVRackLay estimates multi-layered layouts, wherein each layer corresponds to the layout of a shelf within a rack. Given a sequence of images of a warehouse scene, a dual-headed Convolutional-LSTM architecture outputs segmented racks, the front and the top view layout of each shelf within a rack. With minimal effort, such an output is transformed into a 3D rendering of all racks, shelves and objects on the shelves, giving an accurate 3D depiction of the entire warehouse scene in terms of racks, shelves and the number of objects on each shelf. MVRackLay generalizes to a diverse set of warehouse scenes with varying number of objects on each shelf, number of shelves and in the presence of other such racks in the background. Further, MVRackLay shows superior performance vis-a-vis its single view counterpart, RackLay, in layout accuracy, quantized in terms of the mean IoU and mAP metrics. We also showcase a multi-view stitching of the 3D layouts resulting in a representation of the warehouse scene with respect to a global reference frame akin to a rendering of the scene from a SLAM pipeline. To the best of our knowledge, this is the first such work to portray a 3D rendering of a warehouse scene in terms of its semantic components - Racks, Shelves and Objects - all from a single monocular camera.

Detecting objects under adverse weather and lighting conditions is crucial for the safe and continuous operation of an autonomous vehicle, and remains an unsolved problem. We present a Gated Differentiable Image Processing (GDIP) block, a domain-agnostic network architecture, which can be plugged into existing object detection networks (e.g., Yolo) and trained end-to-end with adverse condition images such as those captured under fog and low lighting. Our proposed GDIP block learns to enhance images directly through the downstream object detection loss. This is achieved by learning parameters of multiple image pre-processing (IP) techniques that operate concurrently, with their outputs combined using weights learned through a novel gating mechanism. We further improve GDIP through a multi-stage guidance procedure for progressive image enhancement. Finally, trading off accuracy for speed, we propose a variant of GDIP that can be used as a regularizer for training Yolo, which eliminates the need for GDIP-based image enhancement during inference, resulting in higher throughput and plausible real-world deployment. We demonstrate significant improvement in detection performance over several state-of-the-art methods through quantitative and qualitative studies on synthetic datasets such as PascalVOC, and real-world foggy (RTTS) and low-lighting (ExDark) datasets., Comment: Submitted to ICRA2023. More information at https://gatedip.github.io

Unmanned Aerial Vehicle (UAV) based remote sensing system incorporated with computer vision has demonstrated potential for assisting building construction and in disaster management like damage assessment during earthquakes. The vulnerability of a building to earthquake can be assessed through inspection that takes into account the expected damage progression of the associated component and the component's contribution to structural system performance. Most of these inspections are done manually, leading to high utilization of manpower, time, and cost. This paper proposes a methodology to automate these inspections through UAV-based image data collection and a software library for post-processing that helps in estimating the seismic structural parameters. The key parameters considered here are the distances between adjacent buildings, building plan-shape, building plan area, objects on the rooftop and rooftop layout. The accuracy of the proposed methodology in estimating the above-mentioned parameters is verified through field measurements taken using a distance measuring sensor and also from the data obtained through Google Earth. Additional details and code can be accessed from https://uvrsabi.github.io/ ., Comment: Paper accepted at CVCIE Workshop at ECCV, 2022 and the project page is https://uvrsabi.github.io/

We investigate the Vision-and-Language Navigation (VLN) problem in the context of autonomous driving in outdoor settings. We solve the problem by explicitly grounding the navigable regions corresponding to the textual command. At each timestamp, the model predicts a segmentation mask corresponding to the intermediate or the final navigable region. Our work contrasts with existing efforts in VLN, which pose this task as a node selection problem, given a discrete connected graph corresponding to the environment. We do not assume the availability of such a discretised map. Our work moves towards continuity in action space, provides interpretability through visual feedback and allows VLN on commands requiring finer manoeuvres like "park between the two cars". Furthermore, we propose a novel meta-dataset CARLA-NAV to allow efficient training and validation. The dataset comprises pre-recorded training sequences and a live environment for validation and testing. We provide extensive qualitative and quantitive empirical results to validate the efficacy of the proposed approach.

The world we live in is full of technology and with each passing day the advancement and usage of UAVs increases efficiently. As a result of the many application scenarios, there are some missions where the UAVs are vulnerable to external disruptions, such as a ground station's loss of connectivity, security missions, safety concerns, and delivery-related missions. Therefore, depending on the scenario, this could affect the operations and result in the safe landing of UAVs. Hence, this paper presents a heuristic approach towards safe landing of multi-rotor UAVs in the dynamic environments. The aim of this approach is to detect safe potential landing zones - PLZ, and find out the best one to land in. The PLZ is initially, detected by processing an image through the canny edge algorithm, and then the diameter-area estimation is applied for each region with minimal edges. The spots that have a higher area than the vehicle's clearance are labeled as safe PLZ. Onto the second phase of this approach, the velocities of dynamic obstacles that are moving towards the PLZs are calculated and their time to reach the zones are taken into consideration. The ETA of the UAV is calculated and during the descending of UAV, the dynamic obstacle avoidance is executed. The approach tested on the real-world environments have shown better results from existing work., Comment: 8 pages, 6 figures, 36 references

Many commodity sensors that measure the robot and dynamic obstacle's state have non-Gaussian noise characteristics. Yet, many current approaches treat the underlying-uncertainty in motion and perception as Gaussian, primarily to ensure computational tractability. On the other hand, existing planners working with non-Gaussian uncertainty do not shed light on leveraging distributional characteristics of motion and perception noise, such as bias for efficient collision avoidance. This paper fills this gap by interpreting reactive collision avoidance as a distribution matching problem between the collision constraint violations and Dirac Delta distribution. To ensure fast reactivity in the planner, we embed each distribution in Reproducing Kernel Hilbert Space and reformulate the distribution matching as minimizing the Maximum Mean Discrepancy (MMD) between the two distributions. We show that evaluating the MMD for a given control input boils down to just matrix-matrix products. We leverage this insight to develop a simple control sampling approach for reactive collision avoidance with dynamic and uncertain obstacles. We advance the state-of-the-art in two respects. First, we conduct an extensive empirical study to show that our planner can infer distributional bias from sample-level information. Consequently, it uses this insight to guide the robot to good homotopy. We also highlight how a Gaussian approximation of the underlying uncertainty can lose the bias estimate and guide the robot to unfavorable states with a high collision probability. Second, we show tangible comparative advantages of the proposed distribution matching approach for collision avoidance with previous non-parametric and Gaussian approximated methods of reactive collision avoidance.

We propose a novel flow synthesis based visual servoing framework enabling long-range obstacle avoidance for Micro Air Vehicles (MAV) flying amongst tall skyscrapers. Recent deep learning based frameworks use optical flow to do high-precision visual servoing. In this paper, we explore the question: can we design a surrogate flow for these high-precision visual-servoing methods, which leads to obstacle avoidance? We revisit the concept of saliency for identifying high-rise structures in/close to the line of attack amongst other competing skyscrapers and buildings as a collision obstacle. A synthesised flow is used to displace the salient object segmentation mask. This flow is so computed that the visual servoing controller maneuvers the MAV safely around the obstacle. In this approach, we use a multi-step Cross-Entropy Method (CEM) based servo control to achieve flow convergence, resulting in obstacle avoidance. We use this novel pipeline to successfully and persistently maneuver high-rises and reach the goal in simulated and photo-realistic real-world scenes. We conduct extensive experimentation and compare our approach with optical flow and short-range depth-based obstacle avoidance methods to demonstrate the proposed framework's merit. Additional Visualisation can be found at https://sites.google.com/view/monocular-obstacle/home, Comment: Accepted to IEEE International Conference on Automation Science and Engineering (CASE), 2022

Table-top Rearrangement and Planning is a challenging problem that relies heavily on an excellent perception stack. The perception stack involves observing and registering the 3D scene on the table, detecting what objects are on the table, and how to manipulate them. Consequently, it greatly influences the system's task-planning and motion-planning stacks that follow. We present a comprehensive overview and discuss the different challenges associated with the perception module. This work is a result of our extensive involvement in the ICRA 2022 Open Cloud Robot Table Organization Challenge, in which we stood third in the final rankings., Comment: 5 pages including references, 3 figures

We present UrbanFly: an uncertainty-aware real-time planning framework for quadrotor navigation in urban high-rise environments. A core aspect of UrbanFly is its ability to robustly plan directly on the sparse point clouds generated by a Monocular Visual Inertial SLAM (VINS) backend. It achieves this by using the sparse point clouds to build an uncertainty-integrated cuboid representation of the environment through a data-driven monocular plane segmentation network. Our chosen world model provides faster distance queries than the more common voxel-grid representation, and UrbanFly leverages this capability in two different ways leading to two trajectory optimizers. The first optimizer uses a gradient-free cross-entropy method to compute trajectories that minimize collision probability and smoothness cost. Our second optimizer is a simplified version of the first and uses a sequential convex programming optimizer initialized based on probabilistic safety estimates on a set of randomly drawn trajectories. Both our trajectory optimizers are made computationally tractable and independent of the nature of underlying uncertainty by embedding the distribution of collision violations in Reproducing Kernel Hilbert Space. Empowered by the algorithmic innovation, UrbanFly outperforms competing baselines in metrics such as collision rate, trajectory length, etc., on a high-fidelity AirSim simulator augmented with synthetic and real-world dataset scenes., Comment: Submitted to ACC 2023, Code available at https://github.com/sudarshan-s-harithas/UrbanFly

Modern autonomous vehicles (AVs) often rely on vision, LIDAR, and even radar-based simultaneous localization and mapping (SLAM) frameworks for precise localization and navigation. However, modern SLAM frameworks often lead to unacceptably high levels of drift (i.e., localization error) when AVs observe few visually distinct features or encounter occlusions due to dynamic obstacles. This paper argues that minimizing drift must be a key desiderata in AV motion planning, which requires an AV to take active control decisions to move towards feature-rich regions while also minimizing conventional control cost. To do so, we first introduce a novel data-driven perception module that observes LIDAR point clouds and estimates which features/regions an AV must navigate towards for drift minimization. Then, we introduce an interpretable model predictive controller (MPC) that moves an AV toward such feature-rich regions while avoiding visual occlusions and gracefully trading off drift and control cost. Our experiments on challenging, dynamic scenarios in the state-of-the-art CARLA simulator indicate our method reduces drift up to 76.76% compared to benchmark approaches., Comment: Accepted in IROS 2022

Sparse local feature matching is pivotal for many computer vision and robotics tasks. To improve their invariance to challenging appearance conditions and viewing angles, and hence their usefulness, existing learning-based methods have primarily focused on data augmentation-based training. In this work, we propose an alternative, complementary approach that centers on inducing bias in the model architecture itself to generate `rotation-specific' features using Steerable E2-CNNs, that are then group-pooled to achieve rotation-invariant local features. We demonstrate that this high performance, rotation-specific coverage from the steerable CNNs can be expanded to all rotation angles by combining it with augmentation-trained standard CNNs which have broader coverage but are often inaccurate, thus creating a state-of-the-art rotation-robust local feature matcher. We benchmark our proposed methods against existing techniques on HPatches and a newly proposed UrbanScenes3D-Air dataset for visual place recognition. Furthermore, we present a detailed analysis of the performance effects of ensembling, robust estimation, network architecture variations, and the use of rotation priors.

We consider the problem of an agent/robot with non-holonomic kinematics avoiding many dynamic obstacles. State and velocity noise of both the robot and obstacles as well as the robot's control noise are modelled as non-parametric distributions as often the Gaussian assumptions of noise models are violated in real-world scenarios. Under these assumptions, we formulate a robust MPC that samples robotic controls effectively in a manner that aligns the robot to the goal state while avoiding obstacles under the duress of such non-parametric noise. In particular, the MPC incorporates a distribution matching cost that effectively aligns the distribution of the current collision cone to a certain desired distribution whose samples are collision-free. This cost is posed as a distance function in the Hilbert Space, whose minimization typically results in the collision cone samples becoming collision-free. We compare and show tangible performance gain with methods that model the collision cone distribution by linearizing the Gaussian approximations of the original non-parametric state and obstacle distributions. We also show superior performance with methods that pose a chance constraint formulation of the Gaussian approximations of non-parametric noise without subjecting such approximations to further linearizations. The performance gain is shown both in terms of trajectory length and control costs that vindicates the efficacy of the proposed method. To the best of our knowledge, this is the first presentation of non-holonomic collision avoidance of moving obstacles in the presence of non-parametric state, velocity and actuator noise models.

Humans have a natural ability to effortlessly comprehend linguistic commands such as "park next to the yellow sedan" and instinctively know which region of the road the vehicle should navigate. Extending this ability to autonomous vehicles is the next step towards creating fully autonomous agents that respond and act according to human commands. To this end, we propose the novel task of Referring Navigable Regions (RNR), i.e., grounding regions of interest for navigation based on the linguistic command. RNR is different from Referring Image Segmentation (RIS), which focuses on grounding an object referred to by the natural language expression instead of grounding a navigable region. For example, for a command "park next to the yellow sedan," RIS will aim to segment the referred sedan, and RNR aims to segment the suggested parking region on the road. We introduce a new dataset, Talk2Car-RegSeg, which extends the existing Talk2car dataset with segmentation masks for the regions described by the linguistic commands. A separate test split with concise manoeuvre-oriented commands is provided to assess the practicality of our dataset. We benchmark the proposed dataset using a novel transformer-based architecture. We present extensive ablations and show superior performance over baselines on multiple evaluation metrics. A downstream path planner generating trajectories based on RNR outputs confirms the efficacy of the proposed framework.

This paper presents a novel passive three-output differential with three degrees of freedom (3DOF), that translates motion and torque from a single input to three outputs. The proposed Three-Output Open Differential is designed such that its functioning is analogous to the functioning of a traditional two-output open differential. That is, the differential translates equal motion and torque to all its three outputs when the outputs are unconstrained or are subjected to equivalent load conditions. The introduced design is the first differential with three outputs to realise this outcome. The differential action between the three outputs is realised passively by a symmetric arrangement of three two-output open differentials and three two-input open differentials. The resulting differential mechanism achieves the novel result of equivalent input to output angular velocity and torque relations for all its three outputs. Furthermore, Three-Output Open Differential achieves the novel result for differentials with more than two outputs where each of its outputs shares equivalent angular velocity and torque relations with all the other outputs. The kinematics and dynamics of the Three-Output Open Differential are derived using the bond graph method. In addition, the merits of the differential mechanism along with its current and potential applications are presented., Comment: 8 pages, 9 figures, to be published in proceedings of the ASME 2021 International Design Engineering Technical Conferences and Computers and Information in Engineering Conference IDETC/CIE 2021

The paper introduces the novel Modular Pipe Climber III with a Three-Output Open Differential (3-OOD) mechanism to eliminate slipping of the tracks due to the changing cross-sections of the pipe. This will be achieved in any orientation of the robot. Previous pipe climbers use three-wheel/track modules, each with an individual driving mechanism to achieve stable traversing. Slipping of tracks is prevalent in such robots when it encounters the pipe turns. Thus, active control of each module's speed is employed to mitigate the slip, thereby requiring substantial control effort. The proposed pipe climber implements the 3-OOD to address this issue by allowing the robot to mechanically modulate the track speeds as it encounters a turn. The proposed 3-OOD is the first three-output differential to realize the functional abilities of a traditional two-output differential.

We embark on a hitherto unreported problem of an autonomous robot (self-driving car) navigating in dynamic scenes in a manner that reduces its localization error and eventual cumulative drift or Absolute Trajectory Error, which is pronounced in such dynamic scenes. With the hugely popular Velodyne-16 3D LIDAR as the main sensing modality, and the accurate LIDAR-based Localization and Mapping algorithm, LOAM, as the state estimation framework, we show that in the absence of a navigation policy, drift rapidly accumulates in the presence of moving objects. To overcome this, we learn actions that lead to drift-minimized navigation through a suitable set of reward and penalty functions. We use Proximal Policy Optimization, a class of Deep Reinforcement Learning methods, to learn the actions that result in drift-minimized trajectories. We show by extensive comparisons on a variety of synthetic, yet photo-realistic scenes made available through the CARLA Simulator the superior performance of the proposed framework vis-a-vis methods that do not adopt such policies., Comment: Accepted in American Control Conference 2022

We present CCO-VOXEL: the very first chance-constrained optimization (CCO) algorithm that can compute trajectory plans with probabilistic safety guarantees in real-time directly on the voxel-grid representation of the world. CCO-VOXEL maps the distribution over the distance to the closest obstacle to a distribution over collision-constraint violation and computes an optimal trajectory that minimizes the violation probability. Importantly, unlike existing works, we never assume the nature of the sensor uncertainty or the probability distribution of the resulting collision-constraint violations. We leverage the notion of Hilbert Space embedding of distributions and Maximum Mean Discrepancy (MMD) to compute a tractable surrogate for the original chance-constrained optimization problem and employ a combination of A* based graph-search and Cross-Entropy Method for obtaining its minimum. We show tangible performance gain in terms of collision avoidance and trajectory smoothness as a consequence of our probabilistic formulation vis a vis state-of-the-art planning methods that do not account for such nonparametric noise. Finally, we also show how a combination of low-dimensional feature embedding and pre-caching of Kernel Matrices of MMD allows us to achieve real-time performance in simulations as well as in implementations on on-board commodity hardware that controls the quadrotor flight, Comment: Accepted at ICRA 2022 , Code available at https://github.com/sudarshan-s-harithas/CCO-VOXEL

Standard Model Predictive Control (MPC) or trajectory optimization approaches perform only a local search to solve a complex non-convex optimization problem. As a result, they cannot capture the multi-modal characteristic of human driving. A global optimizer can be a potential solution but is computationally intractable in a real-time setting. In this paper, we present a real-time MPC capable of searching over different driving modalities. Our basic idea is simple: we run several goal-directed parallel trajectory optimizations and score the resulting trajectories based on user-defined meta cost functions. This allows us to perform a global search over several locally optimal motion plans. Although conceptually straightforward, realizing this idea in real-time with existing optimizers is highly challenging from technical and computational standpoints. With this motivation, we present a novel batch non-holonomic trajectory optimization whose underlying matrix algebra is easily parallelizable across problem instances and reduces to computing large batch matrix-vector products. This structure, in turn, is achieved by deriving a linearization-free multi-convex reformulation of the non-holonomic kinematics and collision avoidance constraints. We extensively validate our approach using both synthetic and real data sets (NGSIM) of traffic scenarios. We highlight how our algorithm automatically takes lane-change and overtaking decisions based on the defined meta cost function. Our batch optimizer achieves trajectories with lower meta cost, up to 6x faster than competing baselines., Comment: Published IEEE Robotics and Automation Letters (RA-L)

Given an image or a video captured from a monocular camera, amodal layout estimation is the task of predicting semantics and occupancy in bird's eye view. The term amodal implies we also reason about entities in the scene that are occluded or truncated in image space. While several recent efforts have tackled this problem, there is a lack of standardization in task specification, datasets, and evaluation protocols. We address these gaps with AutoLay, a dataset and benchmark for amodal layout estimation from monocular images. AutoLay encompasses driving imagery from two popular datasets: KITTI and Argoverse. In addition to fine-grained attributes such as lanes, sidewalks, and vehicles, we also provide semantically annotated 3D point clouds. We implement several baselines and bleeding edge approaches, and release our data and code., Comment: published in 2020 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS)

Modern visual-inertial navigation systems (VINS) are faced with a critical challenge in real-world deployment: they need to operate reliably and robustly in highly dynamic environments. Current best solutions merely filter dynamic objects as outliers based on the semantics of the object category. Such an approach does not scale as it requires semantic classifiers to encompass all possibly-moving object classes; this is hard to define, let alone deploy. On the other hand, many real-world environments exhibit strong structural regularities in the form of planes such as walls and ground surfaces, which are also crucially static. We present RP-VIO, a monocular visual-inertial odometry system that leverages the simple geometry of these planes for improved robustness and accuracy in challenging dynamic environments. Since existing datasets have a limited number of dynamic elements, we also present a highly-dynamic, photorealistic synthetic dataset for a more effective evaluation of the capabilities of modern VINS systems. We evaluate our approach on this dataset, and three diverse sequences from standard datasets including two real-world dynamic sequences and show a significant improvement in robustness and accuracy over a state-of-the-art monocular visual-inertial odometry system. We also show in simulation an improvement over a simple dynamic-features masking approach. Our code and dataset are publicly available., Comment: Presented at IROS 2021, code and dataset available at https://karnikram.info/rp-vio

Given a monocular colour image of a warehouse rack, we aim to predict the bird's-eye view layout for each shelf in the rack, which we term as multi-layer layout prediction. To this end, we present RackLay, a deep neural network for real-time shelf layout estimation from a single image. Unlike previous layout estimation methods, which provide a single layout for the dominant ground plane alone, RackLay estimates the top-view and front-view layout for each shelf in the considered rack populated with objects. RackLay's architecture and its variants are versatile and estimate accurate layouts for diverse scenes characterized by varying number of visible shelves in an image, large range in shelf occupancy factor and varied background clutter. Given the extreme paucity of datasets in this space and the difficulty involved in acquiring real data from warehouses, we additionally release a flexible synthetic dataset generation pipeline WareSynth which allows users to control the generation process and tailor the dataset according to contingent application. The ablations across architectural variants and comparison with strong prior baselines vindicate the efficacy of RackLay as an apt architecture for the novel problem of multi-layered layout estimation. We also show that fusing the top-view and front-view enables 3D reasoning applications such as metric free space estimation for the considered rack., Comment: Visit our project repository at https://github.com/Avinash2468/RackLay

The use of local detectors and descriptors in typical computer vision pipelines work well until variations in viewpoint and appearance change become extreme. Past research in this area has typically focused on one of two approaches to this challenge: the use of projections into spaces more suitable for feature matching under extreme viewpoint changes, and attempting to learn features that are inherently more robust to viewpoint change. In this paper, we present a novel framework that combines learning of invariant descriptors through data augmentation and orthographic viewpoint projection. We propose rotation-robust local descriptors, learnt through training data augmentation based on rotation homographies, and a correspondence ensemble technique that combines vanilla feature correspondences with those obtained through rotation-robust features. Using a range of benchmark datasets as well as contributing a new bespoke dataset for this research domain, we evaluate the effectiveness of the proposed approach on key tasks including pose estimation and visual place recognition. Our system outperforms a range of baseline and state-of-the-art techniques, including enabling higher levels of place recognition precision across opposing place viewpoints and achieves practically-useful performance levels even under extreme viewpoint changes., Comment: Accepted to IROS 2021. Project Page: https://uditsinghparihar.github.io/RoRD/

We present DRACO, a method for Dense Reconstruction And Canonicalization of Object shape from one or more RGB images. Canonical shape reconstruction, estimating 3D object shape in a coordinate space canonicalized for scale, rotation, and translation parameters, is an emerging paradigm that holds promise for a multitude of robotic applications. Prior approaches either rely on painstakingly gathered dense 3D supervision, or produce only sparse canonical representations, limiting real-world applicability. DRACO performs dense canonicalization using only weak supervision in the form of camera poses and semantic keypoints at train time. During inference, DRACO predicts dense object-centric depth maps in a canonical coordinate-space, solely using one or more RGB images of an object. Extensive experiments on canonical shape reconstruction and pose estimation show that DRACO is competitive or superior to fully-supervised methods., Comment: Preprint. For project page and code, see https://aadilmehdis.github.io/DRACO-Project-Page/

In this paper, we present BirdSLAM, a novel simultaneous localization and mapping (SLAM) system for the challenging scenario of autonomous driving platforms equipped with only a monocular camera. BirdSLAM tackles challenges faced by other monocular SLAM systems (such as scale ambiguity in monocular reconstruction, dynamic object localization, and uncertainty in feature representation) by using an orthographic (bird's-eye) view as the configuration space in which localization and mapping are performed. By assuming only the height of the ego-camera above the ground, BirdSLAM leverages single-view metrology cues to accurately localize the ego-vehicle and all other traffic participants in bird's-eye view. We demonstrate that our system outperforms prior work that uses strictly greater information, and highlight the relevance of each design decision via an ablation analysis., Comment: Accepted in VISIGRAPP (VISAPP) 2021

Joint space trajectory optimization under end-effector task constraints leads to a challenging non-convex problem. Thus, a real-time adaptation of prior computed trajectories to perturbation in task constraints often becomes intractable. Existing works use the so-called warm-starting of trajectory optimization to improve computational performance. We present a fundamentally different approach that relies on deriving analytical gradients of the optimal solution with respect to the task constraint parameters. This gradient map characterizes the direction in which the prior computed joint trajectories need to be deformed to comply with the new task constraints. Subsequently, we develop an iterative line-search algorithm for computing the scale of deformation. Our algorithm provides near real-time adaptation of joint trajectories for a diverse class of task perturbations such as (i) changes in initial and final joint configurations of end-effector orientation-constrained trajectories and (ii) changes in end-effector goal or way-points under end-effector orientation constraints. We relate each of these examples to real-world applications ranging from learning from demonstration to obstacle avoidance. We also show that our algorithm produces trajectories with quality similar to what one would obtain by solving the trajectory optimization from scratch with warm-start initialization. But most importantly, our algorithm achieves a worst-case speed-up of 160x over the latter approach., Comment: 6+1 pages (last page references)

Significant advances have been made recently in Visual Place Recognition (VPR), feature correspondence, and localization due to the proliferation of deep-learning-based methods. However, existing approaches tend to address, partially or fully, only one of two key challenges: viewpoint change and perceptual aliasing. In this paper, we present novel research that simultaneously addresses both challenges by combining deep-learned features with geometric transformations based on reasonable domain assumptions about navigation on a ground-plane, whilst also removing the requirement for specialized hardware setup (e.g. lighting, downwards facing cameras). In particular, our integration of VPR with SLAM by leveraging the robustness of deep-learned features and our homography-based extreme viewpoint invariance significantly boosts the performance of VPR, feature correspondence, and pose graph submodules of the SLAM pipeline. For the first time, we demonstrate a localization system capable of state-of-the-art performance despite perceptual aliasing and extreme 180-degree-rotated viewpoint change in a range of real-world and simulated experiments. Our system is able to achieve early loop closures that prevent significant drifts in SLAM trajectories. We also compare extensively several deep architectures for VPR and descriptor matching. We also show that superior place recognition and descriptor matching across opposite views results in a similar performance gain in back-end pose graph optimization., Comment: Accepted to VISAPP 2021. Video Link: https://youtu.be/q6cKYW0kX4s

In this paper, we present a simple baseline for visual grounding for autonomous driving which outperforms the state of the art methods, while retaining minimal design choices. Our framework minimizes the cross-entropy loss over the cosine distance between multiple image ROI features with a text embedding (representing the give sentence/phrase). We use pre-trained networks for obtaining the initial embeddings and learn a transformation layer on top of the text embedding. We perform experiments on the Talk2Car dataset and achieve 68.7% AP50 accuracy, improving upon the previous state of the art by 8.6%. Our investigation suggests reconsideration towards more approaches employing sophisticated attention mechanisms or multi-stage reasoning or complex metric learning loss functions by showing promise in simpler alternatives.

Classical Image-Based Visual Servoing (IBVS) makes use of geometric image features like point, straight line and image moments to control a robotic system. Robust extraction and real-time tracking of these features are crucial to the performance of the IBVS. Moreover, such features can be unsuitable for real world applications where it might not be easy to distinguish a target from the rest of the environment. Alternatively, an approach based on complete photometric data can avoid the requirement of feature extraction, tracking and object detection. In this work, we propose one such probabilistic model based approach which uses entire photometric data for the purpose of visual servoing. A novel image modelling method has been proposed using Student Mixture Model (SMM), which is based on Multivariate Student's t-Distribution. Consequently, a vision-based control law is formulated as a least squares minimisation problem. Efficacy of the proposed framework is demonstrated for 2D and 3D positioning tasks showing favourable error convergence and acceptable camera trajectories. Numerical experiments are also carried out to show robustness to distinct image scenes and partial occlusion., Comment: 35 pages, 17 figures

We present a novel Multi-Relational Graph Convolutional Network (MRGCN) based framework to model on-road vehicle behaviors from a sequence of temporally ordered frames as grabbed by a moving monocular camera. The input to MRGCN is a multi-relational graph where the graph's nodes represent the active and passive agents/objects in the scene, and the bidirectional edges that connect every pair of nodes are encodings of their Spatio-temporal relations. We show that this proposed explicit encoding and usage of an intermediate spatio-temporal interaction graph to be well suited for our tasks over learning end-end directly on a set of temporally ordered spatial relations. We also propose an attention mechanism for MRGCNs that conditioned on the scene dynamically scores the importance of information from different interaction types. The proposed framework achieves significant performance gain over prior methods on vehicle-behavior classification tasks on four datasets. We also show a seamless transfer of learning to multiple datasets without resorting to fine-tuning. Such behavior prediction methods find immediate relevance in a variety of navigation tasks such as behavior planning, state estimation, and applications relating to the detection of traffic violations over videos., Comment: To appear at IROS 2020