Ramesh S

Dr. Ramesh S


Research Focus & Metrics

Academic Career Length

20Years (2007-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
9 February 2023 in IEEE Transactions on Software Engineering
Verified
Journal Article
Open access
ZAZohreh Aghababaeyan
Zohreh Aghababaeyan
MAManel Abdellatif
Manel Abdellatif
LBLionel Briand
Lionel Briand
RSRamesh S
Ramesh S
MBMojtaba Bagherzadeh
Mojtaba Bagherzadeh
Deep Neural Networks (DNNs) have been extensively used in many areas including image processing, medical diagnostics and autonomous driving. However, DNNs can exhibit erroneous behaviours that may lead to critical errors, especially when used in safety-critical systems. Inspired by testing techniques for traditional software systems, researchers have proposed neuron coverage criteria, as an analogy to source code coverage, to guide the testing of DNNs. Despite very active research on DNN coverage, several recent studies have questioned the usefulness of such criteria in guiding DNN testing. Further, from a practical standpoint, these criteria are white-box as they require access to the internals or training data of DNNs, which is often not feasible or convenient. Measuring such coverage requires executing DNNs with candidate inputs to guide testing, which is not an option in many practical contexts. In this paper, we investigate diversity metrics as an alternative to white-box coverage criteria. For the previously mentioned reasons, we require such metrics to be black-box and not rely on the execution and outputs of DNNs under test. To this end, we first select and adapt three diversity metrics and study, in a controlled manner, their capacity to measure actual diversity in input sets. We then analyze their statistical association with fault detection using four datasets and five DNNs. We further compare diversity with state-of-the-art white-box coverage criteria. As a mechanism to enable such analysis, we also propose a novel way to estimate fault detection in DNNs. Our experiments show that relying on the diversity of image features embedded in test input sets is a more reliable indicator than coverage criteria to effectively guide DNN testing. Indeed, we found that one of our selected black-box diversity metrics far outperforms existing coverage criteria in terms of fault-revealing capability and computational time. Results also confirm the suspicions that state-of-the-art coverage criteria are not adequate to guide the construction of test input sets to detect as many faults as possible using natural inputs.

25 April 2023 in IEEE Transactions on Software Engineering
Verified
Journal Article
Open access
AZAmirhossein Zolfagharian
Amirhossein Zolfagharian
MAManel Abdellatif
Manel Abdellatif
LBLionel C. Briand
Lionel C. Briand
MBMojtaba Bagherzadeh
Mojtaba Bagherzadeh
RSRamesh S
Ramesh S
Deep Reinforcement Learning (DRL) algorithms have been increasingly employed during the last decade to solve various decision-making problems such as autonomous driving, trading decisions, and robotics. However, these algorithms have faced great challenges when deployed in safety-critical environments since they often exhibit erroneous behaviors that can lead to potentially critical errors. One of the ways to assess the safety of DRL agents is to test them to detect possible faults leading to critical failures during their execution. This raises the question of how we can efficiently test DRL policies to ensure their correctness and adherence to safety requirements. Most existing works on testing DRL agents use adversarial attacks that perturb states or actions of the agent. However, such attacks often lead to unrealistic states of the environment. Furthermore, their main goal is to test the robustness of DRL agents rather than testing the compliance of the agents' policies with respect to requirements. Due to the huge state space of DRL environments, the high cost of test execution, and the black-box nature of DRL algorithms, exhaustive testing of DRL agents is impossible. In this paper, we propose a Search-based Testing Approach of Reinforcement Learning Agents (STARLA) to test the policy of a DRL agent by effectively searching for failing executions of the agent within a limited testing budget. We rely on machine learning models and a dedicated genetic algorithm to narrow the search toward faulty episodes (i.e., sequences of states and actions produced by the DRL agent). We apply STARLA on Deep-Q-Learning agents trained on two different RL problems widely used as benchmarks and show that STARLA significantly outperforms Random Testing by detecting more faults related to the agent's policy. We also investigate how to extract rules that characterize faulty episodes of the DRL agent using our search results. Such rules can be used to understand the conditions under which the agent fails and thus assess the risks of deploying it.

10 January 2019 in Systems
Verified
Journal Article
Open access
PDF file
JDJoseph D'Ambrosio
Joseph D'Ambrosio
AAArun Adiththan
Arun Adiththan
EOEdwin Ordoukhanian
Edwin Ordoukhanian
PPPrakash Peranandam
Prakash Peranandam
RSRamesh S
Ramesh S
PSPadma Sundaram
Padma Sundaram
Advanced driver assistance and automated driving systems must operate in complex environments and make safety-critical decisions. Resilient behavior of these systems in their targeted operation design domain is essential. In this paper, we describe developments in our Model-Based Systems Engineering (MBSE) approach to develop resilient safety-critical automated systems. An MBSE approach provides the ability to provide guarantees about system behavior and potentially reduces dependence on in-vehicle testing through the use of rigorous models and extensive simulation. We are applying MBSE methods to two key aspects of developing resilient systems: (1) ensuring resilient behavior through the use of Resilience Contracts for system decision making; and (2) applying simulation-based testing methods to verify the system handles all known scenarios and to validate the system against potential unknown scenarios. Resilience Contracts make use of contract-based design methods and Partially Observable Markov Decision Processes (POMDP), which allow the system to model potential uncertainty in the sensed environment and thus make more resilient decisions. The simulation-based testing methodology provides a structured approach to evaluate the operation of the target system in a wide variety of operating conditions and thus confirm that the expected resilient behavior has indeed been achieved. This paper provides details on the development of a utility function to support Resilience Contracts and outlines the specific test methods used to evaluate known and unknown operating scenarios.

6 November 2024 in IEEE Transactions on Software Engineering
Verified
Journal Article
Closed access
AZAmirhossein Zolfagharian
Amirhossein Zolfagharian
MAManel Abdellatif
Manel Abdellatif
LBLionel C. Briand
Lionel C. Briand
RSRamesh S
Ramesh S
Deep Reinforcement Learning (DRL) has made significant advancements in various fields, such as autonomous driving, healthcare, and robotics, by enabling agents to learn optimal policies through interactions with their environments. However, the application of DRL in safety-critical domains presents challenges, particularly concerning the safety of the learned policies. DRL agents, which are focused on maximizing rewards, may select unsafe actions, leading to safety violations. Runtime safety monitoring is thus essential to ensure the safe operation of these agents, especially in unpredictable and dynamic environments. This paper introduces SMARLA , a black-box safety monitoring approach specifically designed for DRL agents. SMARLA utilizes machine learning to predict safety violations by observing the agent’s behavior during execution. The approach is based on Q-values, which reflect the expected reward for taking actions in specific states. SMARLA employs state abstraction to reduce the complexity of the state space, enhancing the predictive capabilities of the monitoring model. Such abstraction enables the early detection of unsafe states, allowing for the implementation of corrective and preventive measures before incidents occur. We quantitatively and qualitatively validated SMARLA on three well-known case studies widely used in DRL research. Empirical results reveal that SMARLA is accurate at predicting safety violations, with a low false positive rate, and can predict violations at an early stage, approximately halfway through the execution of the agent, before violations occur. We also discuss different decision criteria, based on confidence intervals of the predicted violation probabilities, to trigger safety mechanisms aiming at a trade-off between early detection and low false positive rates.

30 May 2022 in Empirical Software Engineering
Verified
Journal Article
Open access
PDF file
SASofia Ananieva
Sofia Ananieva
SGSandra Greiner
Sandra Greiner
TKTimo Kehrer
Timo Kehrer
JKJacob Krüger
Jacob Krüger
TKThomas Kühn
Thomas Kühn
RRRalf Reussner
Ralf Reussner
With the increasing demand for customized systems and rapidly evolving technology, software engineering faces many challenges. A particular challenge is the development and maintenance of systems that are highly variable both in space (concurrent variations of the system at one point in time) and time (sequential variations of the system, due to its evolution). Recent research aims to address this challenge by managing variability in space and time simultaneously. However, this research originates from two different areas, software product line engineering and software configuration management, resulting in non-uniform terminologies and a varying understanding of concepts. These problems hamper the communication and understanding of involved concepts, as well as the development of techniques that unify variability in space and time. To tackle these problems, we performed an iterative, expert-driven analysis of existing tools from both research areas to derive a conceptual model that integrates and unifies concepts of both dimensions of variability. In this article, we first explain the construction process and present the resulting conceptual model. We validate the model and discuss its coverage and granularity with respect to established concepts of variability in space and time. Furthermore, we perform a formal concept analysis to discuss the commonalities and differences among the tools we considered. Finally, we show illustrative applications to explain how the conceptual model can be used in practice to derive conforming tools. The conceptual model unifies concepts and relations used in software product line engineering and software configuration management, provides a unified terminology and common ground for researchers and developers for comparing their works, clarifies communication, and prevents redundant developments.

3 August 2016 in Entropy
Verified
Journal Article
Open access
PDF file
GNGanesh Khandu Narwane
Ganesh Khandu Narwane
JDJosé Ángel Galindo Duarte
José Ángel Galindo Duarte
SKShankara Narayanan Krishna
Shankara Narayanan Krishna
DBDavid Benavides
David Benavides
JMJean-Vivien Millo
Jean-Vivien Millo
RSRamesh S
Ramesh S
In a Software Product Line (SPL), the central notion of implementability provides the requisite connection between specifications and their implementations, leading to the definition of products. While it appears to be a simple extension of the traceability relation between components and features, it involves several subtle issues that were overlooked in the existing literature. In this paper, we have introduced a precise and formal definition of implementability over a fairly expressive traceability relation. The consequent definition of products in the given SPL naturally entails a set of useful analysis problems that are either refinements of known problems or are completely novel. We also propose a new approach to solve these analysis problems by encoding them as Quantified Boolean Formulae (QBF) and solving them through Quantified Satisfiability (QSAT) solvers. QBF can represent more complex analysis operations, which cannot be represented by using propositional formulae. The methodology scales much better than the SAT-based solutions hinted in the literature and were demonstrated through a tool called SPLAnE (SPL Analysis Engine) on a large set of SPL models.