Mihir Kulkarni

Mayank Mittal, Pascal Roth, James Tigue et al. · nvidia

We present Isaac Lab, the natural successor to Isaac Gym, which extends the paradigm of GPU-native robotics simulation into the era of large-scale multi-modal learning. Isaac Lab combines high-fidelity GPU parallel physics, photorealistic rendering, and a modular, composable architecture for designing environments and training robot policies. Beyond physics and rendering, the framework integrates actuator models, multi-frequency sensor simulation, data collection pipelines, and domain randomization tools, unifying best practices for reinforcement and imitation learning at scale within a single extensible platform. We highlight its application to a diverse set of challenges, including whole-body control, cross-embodiment mobility, contact-rich and dexterous manipulation, and the integration of human demonstrations for skill acquisition. Finally, we discuss upcoming integration with the differentiable, GPU-accelerated Newton physics engine, which promises new opportunities for scalable, data-efficient, and gradient-based approaches to robot learning. We believe Isaac Lab's combination of advanced simulation capabilities, rich sensing, and data-center scale execution will help unlock the next generation of breakthroughs in robotics research.

Helen Shao, Francisco Villaescusa-Navarro, Pablo Villanueva-Domingo et al.

We train graph neural networks on halo catalogues from Gadget N-body simulations to perform field-level likelihood-free inference of cosmological parameters. The catalogues contain $\lesssim$5,000 halos with masses $\gtrsim 10^{10}~h^{-1}M_\odot$ in a periodic volume of $(25~h^{-1}{\rm Mpc})^3$; every halo in the catalogue is characterized by several properties such as position, mass, velocity, concentration, and maximum circular velocity. Our models, built to be permutationally, translationally, and rotationally invariant, do not impose a minimum scale on which to extract information and are able to infer the values of $Ω_{\rm m}$ and $σ_8$ with a mean relative error of $\sim6\%$, when using positions plus velocities and positions plus masses, respectively. More importantly, we find that our models are very robust: they can infer the value of $Ω_{\rm m}$ and $σ_8$ when tested using halo catalogues from thousands of N-body simulations run with five different N-body codes: Abacus, CUBEP$^3$M, Enzo, PKDGrav3, and Ramses. Surprisingly, the model trained to infer $Ω_{\rm m}$ also works when tested on thousands of state-of-the-art CAMELS hydrodynamic simulations run with four different codes and subgrid physics implementations. Using halo properties such as concentration and maximum circular velocity allow our models to extract more information, at the expense of breaking the robustness of the models. This may happen because the different N-body codes are not converged on the relevant scales corresponding to these parameters.

Mihir Dharmadhikari, Nikhil Khedekar, Mihir Kulkarni et al.

We introduce and open-source the Unified Autonomy Stack, a system-level solution that enables resilient autonomy across diverse aerial and ground robot morphologies. The architecture centers on three synergistic modules -- multi-modal perception, multi-behavior planning, and multi-layered safe navigation -- that together deliver comprehensive mission autonomy. The stack fuses data from LiDAR, radar, vision, and inertial sensing, enabling (a) robust localization and mapping through factor graph-based fusion, (b) semantic scene understanding, (c) motion and informative path planning through sampling-based techniques adaptive across spatial scales, as well as (d) multi-layered safe navigation both through planning on the online reconstructed map and deep learning-driven exteroceptive policies alongside last-resort safety filters using control barrier functions. The resulting behaviors include safe GNSS-denied navigation into unknown and perceptually-degraded regions, exploration of complex environments, object discovery, and efficient inspection planning. The stack has been field-tested and validated on both aerial (rotorcraft) and ground (legged) robots operating in a host of demanding environments, including self-similar and smoke-filled settings, with complex geometries and high obstacle clutter. These tests demonstrate resilient performance in challenging conditions. To facilitate ease of adoption, we open-source the implementation alongside supporting documentation, validation, and evaluation datasets https://github.com/ntnu-arl/unified_autonomy_stack. A video giving the overview of the paper and the field experiments is available at https://youtu.be/l8Su8OXsM-E.

Mihir Kulkarni, Kostas Alexis

This paper contributes a novel learning-based method for aggressive task-driven compression of depth images and their encoding as images tailored to collision prediction for robotic systems. A novel 3D image processing methodology is proposed that accounts for the robot's size in order to appropriately "inflate" the obstacles represented in the depth image and thus obtain the distance that can be traversed by the robot in a collision-free manner along any given ray within the camera frustum. Such depth-and-collision image pairs are used to train a neural network that follows the architecture of Variational Autoencoders to compress-and-transform the information in the original depth image to derive a latent representation that encodes the collision information for the given depth image. We compare our proposed task-driven encoding method with classical task-agnostic methods and demonstrate superior performance for the task of collision image prediction from extremely low-dimensional latent spaces. A set of comparative studies show that the proposed approach is capable of encoding depth image-and-collision image tuples from complex scenes with thin obstacles at long distances better than the classical methods at compression ratios as high as 4050:1.

Mihir Kulkarni, Satvik Golechha, Rishi Raj et al.

Tuberculosis (TB), an infectious bacterial disease, is a significant cause of death, especially in low-income countries, with an estimated ten million new cases reported globally in $2020$. While TB is treatable, non-adherence to the medication regimen is a significant cause of morbidity and mortality. Thus, proactively identifying patients at risk of dropping off their medication regimen enables corrective measures to mitigate adverse outcomes. Using a proxy measure of extreme non-adherence and a dataset of nearly $700,000$ patients from four states in India, we formulate and solve the machine learning (ML) problem of early prediction of non-adherence based on a custom rank-based metric. We train ML models and evaluate against baselines, achieving a $\sim 100\%$ lift over rule-based baselines and $\sim 214\%$ over a random classifier, taking into account country-wide large-scale future deployment. We deal with various issues in the process, including data quality, high-cardinality categorical data, low target prevalence, distribution shift, variation across cohorts, algorithmic fairness, and the need for robustness and explainability. Our findings indicate that risk stratification of non-adherent patients is a viable, deployable-at-scale ML solution. As the official AI partner of India's Central TB Division, we are working on multiple city and state-level pilots with the goal of pan-India deployment.

Ashwath Vaithinathan Aravindan, Abha Jha, Mihir Kulkarni

Vision-Language Models (VLMs) have shown remarkable performance in integrating visual and textual information for tasks such as image captioning and visual question answering. However, these models struggle with compositional generalization and object binding, which limit their ability to handle novel combinations of objects and their attributes. Our work explores the root causes of these failures using mechanistic interpretability techniques. We show evidence that individual neurons in the MLP layers of CLIP's vision encoder represent multiple features, and this "superposition" directly hinders its compositional feature representation which consequently affects compositional reasoning and object binding capabilities. We hope this study will serve as an initial step toward uncovering the mechanistic roots of compositional failures in VLMs. The code and supporting results can be found https://github.com/Mystic-Slice/Do-VLMs-Have-Bad-Eyes.

Zekun Li, Malcolm Grossman, Eric et al.

Geospatial question answering (QA) is a fundamental task in navigation and point of interest (POI) searches. While existing geospatial QA datasets exist, they are limited in both scale and diversity, often relying solely on textual descriptions of geo-entities without considering their geometries. A major challenge in scaling geospatial QA datasets for reasoning lies in the complexity of geospatial relationships, which require integrating spatial structures, topological dependencies, and multi-hop reasoning capabilities that most text-based QA datasets lack. To address these limitations, we introduce MapQA, a novel dataset that not only provides question-answer pairs but also includes the geometries of geo-entities referenced in the questions. MapQA is constructed using SQL query templates to extract question-answer pairs from OpenStreetMap (OSM) for two study regions: Southern California and Illinois. It consists of 3,154 QA pairs spanning nine question types that require geospatial reasoning, such as neighborhood inference and geo-entity type identification. Compared to existing datasets, MapQA expands both the number and diversity of geospatial question types. We explore two approaches to tackle this challenge: (1) a retrieval-based language model that ranks candidate geo-entities by embedding similarity, and (2) a large language model (LLM) that generates SQL queries from natural language questions and geo-entity attributes, which are then executed against an OSM database. Our findings indicate that retrieval-based methods effectively capture concepts like closeness and direction but struggle with questions that require explicit computations (e.g., distance calculations). LLMs (e.g., GPT and Gemini) excel at generating SQL queries for one-hop reasoning but face challenges with multi-hop reasoning, highlighting a key bottleneck in advancing geospatial QA systems.

Paolo De Petris, Huan Nguyen, Mihir Dharmadhikari et al.

This work presents the design, hardware realization, autonomous exploration and object detection capabilities of RMF-Owl, a new collision-tolerant aerial robot tailored for resilient autonomous subterranean exploration. The system is custom built for underground exploration with focus on collision tolerance, resilient autonomy with robust localization and mapping, alongside high-performance exploration path planning in confined, obstacle-filled and topologically complex underground environments. Moreover, RMF-Owl offers the ability to search, detect and locate objects of interest which can be particularly useful in search and rescue missions. A series of results from field experiments are presented in order to demonstrate the system's ability to autonomously explore challenging unknown underground environments.

Marco Tranzatto, Frank Mascarich, Lukas Bernreiter et al.

Autonomous exploration of subterranean environments constitutes a major frontier for robotic systems as underground settings present key challenges that can render robot autonomy hard to achieve. This has motivated the DARPA Subterranean Challenge, where teams of robots search for objects of interest in various underground environments. In response, the CERBERUS system-of-systems is presented as a unified strategy towards subterranean exploration using legged and flying robots. As primary robots, ANYmal quadruped systems are deployed considering their endurance and potential to traverse challenging terrain. For aerial robots, both conventional and collision-tolerant multirotors are utilized to explore spaces too narrow or otherwise unreachable by ground systems. Anticipating degraded sensing conditions, a complementary multi-modal sensor fusion approach utilizing camera, LiDAR, and inertial data for resilient robot pose estimation is proposed. Individual robot pose estimates are refined by a centralized multi-robot map optimization approach to improve the reported location accuracy of detected objects of interest in the DARPA-defined coordinate frame. Furthermore, a unified exploration path planning policy is presented to facilitate the autonomous operation of both legged and aerial robots in complex underground networks. Finally, to enable communication between the robots and the base station, CERBERUS utilizes a ground rover with a high-gain antenna and an optical fiber connection to the base station, alongside breadcrumbing of wireless nodes by our legged robots. We report results from the CERBERUS system-of-systems deployment at the DARPA Subterranean Challenge Tunnel and Urban Circuits, along with the current limitations and the lessons learned for the benefit of the community.

Mihir Kulkarni, Mihir Dharmadhikari, Marco Tranzatto et al.

This paper presents a novel strategy for autonomous teamed exploration of subterranean environments using legged and aerial robots. Tailored to the fact that subterranean settings, such as cave networks and underground mines, often involve complex, large-scale and multi-branched topologies, while wireless communication within them can be particularly challenging, this work is structured around the synergy of an onboard exploration path planner that allows for resilient long-term autonomy, and a multi-robot coordination framework. The onboard path planner is unified across legged and flying robots and enables navigation in environments with steep slopes, and diverse geometries. When a communication link is available, each robot of the team shares submaps to a centralized location where a multi-robot coordination framework identifies global frontiers of the exploration space to inform each system about where it should re-position to best continue its mission. The strategy is verified through a field deployment inside an underground mine in Switzerland using a legged and a flying robot collectively exploring for 45 min, as well as a longer simulation study with three systems.

Mihir Kulkarni, Huan Nguyen, Kostas Alexis

This paper presents the design concept, modeling and motion planning solution for the aerial robotic chain. This design represents a configurable robotic system of systems, consisting of multi-linked micro aerial vehicles that simultaneously presents the ability to cross narrow sections, morph its shape, ferry significant payloads, offer the potential of distributed sensing and processing, and allow system extendability. We contribute an approach to address the motion planning problem of such a connected robotic system of systems, making full use of its reconfigurable nature, to find collision free paths in a fast manner despite the increased number of degrees of freedom. The presented approach exploits a library of aerial robotic chain configurations, optimized either for cross-section size or sensor coverage, alongside a probabilistic strategy to sample random shape configurations that may be needed to facilitate continued collision-free navigation. Evaluation studies in simulation involve traversal of constrained and obstacle-laden environments, having narrow corridors and cross sections.

Jaydeep Chipalkatti, Mihir Kulkarni

We carry out a comprehensive analysis of letter frequencies in contemporary written Marathi. We determine sets of letters which statistically predominate any large generic Marathi text, and use these sets to estimate the entropy of Marathi.