arXiv subjects

Soumya Ranjan Sahoo

Publications and source records attributed to Soumya Ranjan Sahoo.

10 recordsLinked to original sources

arXiv2026

Barrier-Shaped Recurrent Reinforcement Learning for Autonomous Landing on a Heaving Ship Deck

This paper addresses autonomous landing of unmanned aerial vehicle (UAV) rotorcrafts on a heaving ship deck using a recurrent policy trained via asymmetric actor-critic reinforcement learning: the critic sees 6 s of future deck height during training, while the actor sees only what the onboard sensors provide in flight, a 17-dimensional state made of the vehicle's position, velocity and attitude relative to the deck and outputs world-frame velocity and yaw-rate commands. Training is performed across 4096 parallel simulated environments, followed by fine-tuning in 16 environments with an onboard vision pipeline in the loop. A control-barrier-function (CBF) stopping margin on the deck-relative vertical state is incorporated at two stages: as a reward term during training, where it halves the median simulated contact speed relative to a policy trained without it, and as a runtime safety filter at deployment, evaluated at every control step to abort and retry the descent when the margin is violated. Because the autopilot's disarm logic cannot detect the vehicle resting on a moving deck, proximity-based thrust cutoff at touchdown is commanded directly in the landing pipeline. The proposed approach is validated using a parallel-manipulator-platform-based deck emulator that reproduces the heaving motion of the ship deck (scaled to 0.70 m peak-to-peak, 7.5 s mean period) and a quadcopter UAV with an onboard camera. Across 31 motion-capture and 20 vision-based trials, the UAV landed every time, with median times to contact of 6.5 and 7.3 s; 77% and 50% landed on the first attempt, with mean deck-relative speeds of 0.29 and 0.27 m/s, respectively, at the instant of thrust cutoff, which is the last speed under the policy's control. Supplementary video: https://youtu.be/S5hDkrSZJt4

eess.SY↗

arXiv2026

Negation is Not Semantic: Diagnosing Dense Retrieval Failure Modes for Trade-offs in Contradiction-Aware Biomedical QA

Large Language Models (LLMs) have demonstrated strong capabilities in biomedical question answering, yet their tendency to generate plausible but unverified claims poses serious risks in clinical settings. To mitigate these risks, the TREC 2025 BioGen track mandates grounded answers that explicitly surface contradictory evidence (Task A) and the generation of narrative driven, fully attributed responses (Task B). Addressing the absence of target ground truth, we present a proxy-based development framework using the SciFact dataset to systematically optimize retrieval architectures. Our iterative evaluation revealed a "Simplicity Paradox": complex adversarial dense retrieval strategies failed catastrophically at contradiction detection (MRR 0.023) due to Semantic Collapse, where negation signals become indistinguishable in vector space. We further identify a Retrieval Asymmetry: filtering dense embeddings improves contradiction detection but degrades support recall, compromising reliability. We resolve this via a Decoupled Lexical Architecture built on a unified BM25 backbone, balancing semantic support recall (0.810) with precise contradiction surfacing (0.750). This approach achieves the highest Weighted MRR (0.790) on the proxy benchmark while remaining the only viable strategy for scaling to the 30 million document PubMed corpus. For answer generation, we introduce Narrative Aware Reranking and One-Shot In-Context Learning, improving citation coverage from 50% (zero-shot) to 100%. Official TREC results confirm our findings: our system ranks 2nd on Task A contradiction F1 and 3rd out of 50 runs on Task B citation coverage (98.77%), achieving zero citation contradict rate. Our work transforms LLMs from stochastic generators into honest evidence synthesizers, showing that epistemic integrity in biomedical AI requires precision and architectural scalability isolated metric optimization.

cs.IR↗

arXiv2024

ILLUMINER: Instruction-tuned Large Language Models as Few-shot Intent Classifier and Slot Filler

State-of-the-art intent classification (IC) and slot filling (SF) methods often rely on data-intensive deep learning models, limiting their practicality for industry applications. Large language models on the other hand, particularly instruction-tuned models (Instruct-LLMs), exhibit remarkable zero-shot performance across various natural language tasks. This study evaluates Instruct-LLMs on popular benchmark datasets for IC and SF, emphasizing their capacity to learn from fewer examples. We introduce ILLUMINER, an approach framing IC and SF as language generation tasks for Instruct-LLMs, with a more efficient SF-prompting method compared to prior work. A comprehensive comparison with multiple baselines shows that our approach, using the FLAN-T5 11B model, outperforms the state-of-the-art joint IC+SF method and in-context learning with GPT3.5 (175B), particularly in slot filling by 11.1--32.2 percentage points. Additionally, our in-depth ablation study demonstrates that parameter-efficient fine-tuning requires less than 6% of training data to yield comparable performance with traditional full-weight fine-tuning.

cs.CL↗

arXiv2023

InCA: Rethinking In-Car Conversational System Assessment Leveraging Large Language Models

The assessment of advanced generative large language models (LLMs) poses a significant challenge, given their heightened complexity in recent developments. Furthermore, evaluating the performance of LLM-based applications in various industries, as indicated by Key Performance Indicators (KPIs), is a complex undertaking. This task necessitates a profound understanding of industry use cases and the anticipated system behavior. Within the context of the automotive industry, existing evaluation metrics prove inadequate for assessing in-car conversational question answering (ConvQA) systems. The unique demands of these systems, where answers may relate to driver or car safety and are confined within the car domain, highlight the limitations of current metrics. To address these challenges, this paper introduces a set of KPIs tailored for evaluating the performance of in-car ConvQA systems, along with datasets specifically designed for these KPIs. A preliminary and comprehensive empirical evaluation substantiates the efficacy of our proposed approach. Furthermore, we investigate the impact of employing varied personas in prompts and found that it enhances the model's capacity to simulate diverse viewpoints in assessments, mirroring how individuals with different backgrounds perceive a topic.

cs.CL↗

arXiv2023

Cooperative Localization for Autonomous Underwater Vehicles -- a comprehensive review

Cooperative localization is an important technique in environments devoid of GPS-based localization, more so in underwater scenarios, where none of the terrestrial localization techniques based on radio frequency or optics are suitable due to severe attenuation. Given the large swaths of oceans and seas where autonomous underwater vehicles (AUVs) operate, traditional acoustic positioning systems fall short on many counts. Cooperative localization (CL), which involves sharing mutual information amongst the vehicles, has thus emerged as a viable option in the past decade. This paper assimilates the research carried out in AUV cooperative localization and presents a qualitative overview. The cooperative localization approaches are categorized by their cooperation and localization strategies, while the algorithms employed are reviewed on the various challenges posed by the underwater acoustic channel and environment. Furthermore, existing problems and future scope in the domain of underwater cooperative localization are discussed.

eess.SY↗

arXiv2023

Robust Short-term Operation of AC Power Network with Injection Uncertainties

With uncertain injections from Renewable Energy Sources (RESs) and loads, deterministic AC Optimal Power Flow (OPF) often fails to provide optimal setpoints of conventional generators. A computationally time-efficient, economical, and robust solution is essential for ACOPF with short-term injection uncertainties. Usually, applying Robust Optimization (RO) for conventional non-linear ACOPF results in computationally intractable Robust Counterpart (RC), which is undesirable as ACOPF is an operational problem. Hence, this paper proposes a single-stage non-integer non-recursive RC of ACOPF, using a dual transformation, for short-term injection uncertainties. The proposed RC is convex, tractable, and provides base-point active power generations and terminal voltage magnitudes (setpoints) of conventional generators that satisfy all constraints for all realizations of defined injection uncertainties. The non-linear impact of uncertainties on other variables is inherently modeled without using any affine policy. The proposed approach also includes the budget of uncertainty constraints for low conservatism of the obtained setpoints. Monte-Carlo Simulation (MCS) based participation factored AC power flows validate the robustness of the obtained setpoints on NESTA and case9241pegase systems for different injection uncertainties. Comparison with previous approaches indicates the efficacy of the proposed approach in terms of low operational cost and computation time.

eess.SY↗

arXiv2022

Data Augmentation using Feature Generation for Volumetric Medical Images

Medical image classification is one of the most critical problems in the image recognition area. One of the major challenges in this field is the scarcity of labelled training data. Additionally, there is often class imbalance in datasets as some cases are very rare to happen. As a result, accuracy in classification task is normally low. Deep Learning models, in particular, show promising results on image segmentation and classification problems, but they require very large datasets for training. Therefore, there is a need to generate more of synthetic samples from the same distribution. Previous work has shown that feature generation is more efficient and leads to better performance than corresponding image generation. We apply this idea in the Medical Imaging domain. We use transfer learning to train a segmentation model for the small dataset for which gold-standard class annotations are available. We extracted the learnt features and use them to generate synthetic features conditioned on class labels, using Auxiliary Classifier GAN (ACGAN). We test the quality of the generated features in a downstream classification task for brain tumors according to their severity level. Experimental results show a promising result regarding the validity of these generated features and their overall contribution to balancing the data and improving the classification class-wise accuracy.

eess.IV↗

arXiv2021

Subsystem decomposition and state estimation of nonlinear processes with implicit time-scale multiplicity

In this work, we propose a subsystem decomposition approach and a distributed estimation scheme for a class of implicit two-time-scale nonlinear systems. Taking the advantage of the two-time-scale separation, these processes are decomposed into smaller subsystems such as fast subsystem and slow subsystem. In the proposed method, an approach, composite solution, combines the approximate solutions obtained from both fast and slow subsystems. Based on the fast and slow subsystems, a distributed state estimation scheme is proposed to handle the implicit time-scale multiplicity. In the proposed design, an extended Kalman filter (EKF) is designed for the fast subsystem and a moving horizon estimator (MHE) is designed for the slow subsystem. There is a communication between the estimators: the slow subsystem is only required to send information to the fast subsystem one-directionally. The fast subsystem estimator does not send out any information. The estimators use different sampling of the state measurements, i.e., fast sampling of the fast state variables is considered in the fast EKF and slow sampling of the slow state variables is considered in the slow MHE. Extensive simulations based on a chemical process are performed to illustrate the effectiveness and applicability of the proposed subsystem decomposition and composite state estimation architecture.

eess.SY↗

arXiv2019

An Efficient Approach for obtaining Feasible solutions from SOCP formulation of ACOPF

Exact Second Order Conic Programming (SOCP) formulation of AC Optimal Power Flow (ACOPF) consists of non-convex arctangent constraints. Generally, these constraints have been ignored or approximated (at the expense of increased computational time) so as to solve the relaxed and convex SOCP formulation of ACOPF. As a consequence, retrieving unique and feasible bus voltage phasors for ACOPF of meshed networks is not always possible. In this letter, this issue has been addressed. The arctangent constraints have been represented by alternate linear constraints in the relaxed SOCP formulation of ACOPF, by exploiting the properties of the meshed power network. Numerical tests show that the proposed formulation gives an unique and feasible ACOPF solution, which is practically realizable from system operation perspective and with global optimality feature, as compared to other works reported in the literature. Moreover, the proposed formulation is extremely efficient as only one execution of formulation provides a feasible solution.

eess.SY↗

arXiv2018

Global Attitude Stabilization using Pseudo-Targets

The topological obstructions on the attitude space of a rigid body make global asymptotic stabilization impossible using continuous state-feedback. This paper presents novel algorithms to overcome such topological limitations and achieve arbitrary attitude maneuvers with only continuous, memory-less state-feedback. We first present nonlinear control laws using both rotation matrices and quaternions that give rise to one almost globally asymptotically stabilizable equilibrium along with a nowhere dense set of unstable equilibria. The unstable equilibria are uniquely identified in the attitude error space. Pseudo-targets are then designed to make the controller believe that the attitude error is within the region of attraction of the stable equilibrium. Further, the pseudo-target ensures that maximum control action is provided to push the closed-loop system toward the stable equilibrium. The proposed algorithms are validated using both numerical simulations and experiments to show their simplicity and effectiveness.

eess.SY↗