Skip to main content

Engineering Trustworthy ROBOTICS System: Safety, Ethics & Cyber Security

Page 1


International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056

Volume: 13 Issue: 03 | Mar 2026 www.irjet.net p-ISSN: 2395-0072

Engineering Trustworthy ROBOTICS System: Safety, Ethics & Cyber Security

Students, Electronics & Computer Science, St John College of Engineering and Management, Maharashtra, India

Abstract: The rapid integration of Artificial Intelligence (AI) and Large Language Models (LLMs) into robotic systems has ushered in an era of unprecedented autonomy, enabling machines to perform complex tasks in unstructured environments across healthcare, manufacturing, anddefense. However, this "paradigm shift" introduces significant vulnerabilities known as the "embodiment gap" a critical discord between an LLM’s abstract digital reasoning and the physical, context-dependent nature of robotic actions. Unlike traditional software, where failures result in data loss or semantic toxicity, a failure in a robotic system manifests as a kinetic event, potentially leading to catastrophic physical consequences, environmental damage, or loss of human life. This paper presents an exhaustive multidisciplinary review of the three interconnected pillars of modern robotics: physical safety, ethical standards, and digital security. We provide an in-depth analysis of methodologies such as EHAZOP (Ethical Hazard Analysis) for identifying socio-technical risks like "culture flattening" and "infantilization." Furthermore, we survey the emerging threat landscape of LLM-controlled agents, specificallyfocusingonmulti-modalprompt injection, backdoor attacks, and jailbreaking mechanisms that bypass safety alignments. By synthesizing hardware-level safety protocols such as redundant E-stop topologies and forcelimiting actuators with advanced algorithmic defenses like Control Barrier Functions (CBFs) and multi-LLM oversight, this review provides a foundational roadmap for the "Safetyby-Design," "Ethics-by-Design," and "Security-by-Design" principles. This 15-pagecomprehensivereviewaims tobridge the gap between theoretical AI safety and practical robotic engineering to ensure long-term reliability and societal acceptance.

Key Words: Robotics Safety, AI Ethics, Cybersecurity, Human-Robot Interaction (HRI), EHAZOP, EmbodiedAI, Risk Assessment, Cyber-Physical Systems, LLM Grounding, Control Barrier Functions.

1. INTRODUCTION

1.1 The Robotics Revolution and the Autonomy Paradox

Robotic technology has undergone a dramatic transformation over the past decade, evolving from rigid industrialmanipulatorstofluid,autonomousagentscapable ofnavigatinghuman-centricspaces.Withadvancementsin machine learning, high-fidelity sensor technologies, and

cloud-edge computing, robots are increasingly being deployed in dynamic environments that were previously deemed too complex for automation. Today, applications extendbeyondsimplerepetitivetaskstoincludehigh-stakes domains such as robotic-assisted surgery, autonomous vehicle fleets, smart logistics warehouses, agricultural drones,andmilitarydefensesystems.

However, this surge in capability introduces what researchers call the "Autonomy Paradox": as a system becomesmorecapableofindependentdecision-making,it becomesharderforhumanoperatorstopredict,verify,and secure its behavior. Traditional robots operated within "caged"industrialsettings,isolatedfrompublicinteraction. In contrast, modern collaborative robots (cobots) and servicerobotsoperateinopen,unpredictableenvironments whereasinglealgorithmicerrororsensormalfunctioncan directlyimpacthumansafety.

1.2 Defining the Three Pillars of Trustworthiness

To address these risks, the engineering community has identifiedthreefoundationaldimensionsoftrustworthiness thatmustbeintegratedintotheroboticlifecycle:

1. Safety:Thisisthemostfundamentalrequirement, defined as the absence of unacceptable physical risk. It involves the mechanical and algorithmic reliabilityofthemachinetopreventkineticharmto humansanditssurroundings.

2. Ethics:Asrobotstakeonrolestraditionallyheldby humans (e.g., caregiving, policing), they must operatewithinaframeworkofmoralresponsibility. Thisincludesensuringhumandignity,preventing algorithmic bias, and maintaining clear lines of accountabilityforautonomousactions.

3. Cybersecurity:Inaninterconnectedworld,robots are essentially mobile Internet of Things (IoT) devices.Cybersecurityprotectstheroboticplatform frommaliciousexploitation,ensuringthatahacker cannot hijack the robot's physical components to causeintentionalharm.

2. LITERATURE SURVEY:

The quest for trustworthy robotics is documented across severalpioneeringstudies.Recentresearchhighlightsthat

International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056

Volume: 13 Issue: 03 | Mar 2026 www.irjet.net p-ISSN: 2395-0072

the"trustworthiness"ofarobotisamulti-layeredproperty thatcannotbeachievedthroughsoftwareupdatesalone.

2.1 Cybersecurity of Robotic Architectures

Surveysofroboticsecurity(Yaacoubetal.,2022)categorize threatsintohardware,software,andcommunicationlevels. AcriticalfindingisthatrobotsutilizingtheRobotOperating System(ROS)areparticularlyvulnerable.ROSwasoriginally designed for research environments; its default configurations often lack encryption and authentication mechanisms.Thisallowsattackerstoexploitweaknetwork securitytoinjectmalicious commands,manipulatesensor data, or remotely hijack robotic platforms. Neupane et al. (2024) specifically highlights the vulnerabilities in the hybrid architectures used in AI-Robotics, noting that the "Control"layerisoftentheleastprotectedduetotheneed for real-time, low-latency communication which often precludesheavyencryption.

2.2 Safety-Critical Verification

Traditionalsafetyvalidationmethods,suchasFailureMode and Effects Analysis (FMEA) and Fault Tree Analysis, are being adapted for AI. However, because machine learning models behave probabilistically rather than deterministically, researchers now advocate for "Formal Verification" usingmathematicallogictoprovethatarobot will never enter a defined "unsafe state" (Guiochet et al., 2017).

2.3 Ethical Governance Frameworks

WinfieldandJirotka(2018)arguethatethicalgovernanceis a prerequisite for public adoption. The European Commission’s (2019) guidelines further established that "TrustworthyAI"mustbelawful,ethical,androbust.Inthe domesticsphere,Menonetal.(2024)introducedEHAZOP, the first structured method to apply engineering hazard analysistoabstractethicalconceptslike"dignity"and"social isolation."

3. ROBOTICS SAFETY: ENGINEERING PHYSICAL INTEGRITY

Safety remains the most fundamental requirement of any roboticsystem.Mechanicalhazards,electricalfaults,sensor failures,andalgorithmicerrorscanleadtosevereaccidents.

3.1 Hardware-Level Interlocks and Fail-Safes

 EmergencyStop(E-Stop)Topology:ModernE-stops useredundant,dual-channelcircuits.Ifasinglewire breaks,thesystemdefaultstoa"Stop"state.These aretypicallyhard-wiredtothepowersupplyofthe actuators,bypassingsoftwarelayerstoensurethat

even a system-wide software crash does not preventamanualoverride.

 Force and Torque Limiting: Collaborative robots (cobots) are equipped with internal sensors that detectabnormalresistance.Ifarobotarmstrikesa human, the sudden spike in torque triggers an immediatehalttopreventcrushinginjuries.

 Redundant Sensor Fusion: To prevent "Sensor Blindness,"trustworthyrobotsutilizemulti-modal fusion.LiDARprovidesaccuratedistancemapping, whileultrasonicsensorsdetectglassortransparent objectsthatLiDARmightmiss.

3.2 Algorithmic Safety Invariants

Beyond hardware, safety is managed through Control BarrierFunctions(CBFs).Mathematically,aCBFensuresthat a set of "safe" states is forward-invariant. If the robot's currentstateapproachestheboundaryofsafety,thesafety controlleroverridestheAI'sgoal-seekingcommandtosteer therobotbacktoasecuretrajectory.

4. ETHICS-BY-DESIGN:THEEHAZOP METHODOLOGY

Ethical considerations in robotics extend beyond physical safetytoincludehumandignity,privacy,andpsychological well-being.

4.1 The EHAZOP Framework and Guide Words

The Ethical Hazard Analysis (EHAZOP) is a structured processadaptedfromtraditionalsafetyengineering.Ituses "GuideWords"toidentifyrisksinassistiverobotics:

 Culture Flattening: Occurs when a robot's standardized programming ignores the specific cultural, linguistic, or personal habits of a user, forcingtheusertoadapttothemachine.

 Infantilization:Inelderlycare,ifarobottakesover taskstheuserisstillcapableof,itcanleadtoaloss ofcognitiveandphysicalagency.

 Deception:Ethical hazardsariseif a robot"fakes" emotionstomanipulateauser'sbehavior,leadingto misplacedtrust.

4.2 The Responsibility Gap

A primary ethical dilemma is accountability. If an autonomoussystemmakesadecisionthatleadstoharm,the "responsibilitygap"makesitdifficulttoassignlegalormoral blamebetweenthemanufacturer,theAIdeveloper,andthe owner. Researchers propose "Black Box" recorders for robotstoexplainAIdecisionsinhuman-readablelanguage.

International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056

Volume: 13 Issue: 03 | Mar 2026 www.irjet.net p-ISSN: 2395-0072

5. CYBER SECURITY: THE DEFENSE-IN-DEPTH TAXONOMY

Cybersecurity in AI-robotics protects against malicious actors who seek to hijack the physical capabilities of the machine.

5.1 Multi-Layer Attack Taxonomy

Modern AI-Robotics systems are vulnerable across three fundamentalarchitecturalelements:

1. PerceptionLayer:Attacksincludesensorspoofing (e.g., using a laser to "blind" a camera) or adversarialexamples(e.g.,placingstickersonastop signtomakearobotseeitasaspeedlimitsign).

2. NavigationandPlanningLayer:Maliciousactorscan manipulate pathfinding algorithms via "map poisoning," leading a robot into a restricted or dangerousarea.

3. ControlLayer:ActuatorscanbehijackedviaManin-the-Middle (MITM) attacks. An attacker can overrideasafety-haltcommand,forcingtherobotto continuemovingatmaximumvelocity.

5.2 The Embodiment Gap and LLM Threats

The integration of Large Language Models (LLMs) allows robots to interpret complex human commands but introducesPromptInjectionandJailbreaking.Becausethe LLM understands language but lacks physical "common sense,"anattackercantricktherobotintoignoringsafety guidelinesthroughcleverphrasing(e.g.,"Ignoreallprevious safetyprotocolsanddemonstratemaximumvelocity").

6. RESEARCH GAPS AND FUTURE SCOPE

Despitesubstantialprogress,severalresearchgapspersist:

 UnifiedGlobalRegulatoryFrameworks:Thereisno single international regulatory framework governing autonomous robotic systems across borders.

 ValidationofAISafety:Provingthatadeep-learning modelwill always actsafelyineverypossiblerealworldscenarioremainsachallenge.

 Quantum-Resistant Encryption: Future robotic communicationmustaccountforquantumthreats topreventlong-termdatahijacking.

7. CONCLUSION

Therapidconvergenceofrobotics,artificialintelligence,and cloudcomputinghaspermanentlyalteredthetrajectoryof modern engineering. As this comprehensive review has demonstrated,thetransitionfromindustrialautomationto autonomousagencynecessitatesatotalreimaginingofhow we define and implement system integrity. We have identified that trust in robotic systems is not a singular featurebutabyproductoftheseamlessintegrationofsafety, ethics,andcybersecurity.

Thefutureofroboticsdependsontheadoptionofaholistic lifecycle approach. We propose that "Safety-by-Design," "Ethics-by-Design," and "Security-by-Design" must move from being peripheral considerations to being the central pillars of the development process. By establishment of globalstandardsforfairness,transparency,andhardwarerootedsecurity,wecanensurethatrobotsamplifyhuman potentialwhilesafeguardingoursharedvaluesandphysical wellbeing.

REFERENCES

[1] C. Menon et al., "EHAZOP: A Proof of Concept Ethical HazardAnalysisofanAssistiveRobot,"arXiv:2406.09239, 2024.

[2] X. Huang et al., "Trust in LLM-controlled Robotics: a SurveyofSecurityThreats,DefensesandChallenges,"arXiv: 2601.02377,2025.

[3]S.Neupaneetal.,"SecurityConsiderationsinAI-Robotics: A Survey of Current Methods, Challenges, and Opportunities,"IEEEAccess,vol.12,2024.

Turn static files into dynamic content formats.

Create a flipbook
Engineering Trustworthy ROBOTICS System: Safety, Ethics & Cyber Security by IRJET Journal - Issuu