Skip to main content

A Survey of Autonomous Robotic Approaches with Focus on LoRa- Enabled Q-Learning Navigation

Page 1


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

Volume: 12 Issue: 10 | Oct 2025 www.irjet.net p-ISSN: 2395-0072

A Survey of Autonomous Robotic Approaches with Focus on LoRaEnabled Q-Learning Navigation

1Department of Electronics and Communication Engineering, BNMIT, Bengaluru, India

2Department of Electronics and Communication Engineering, BNMIT, Bengaluru, India

3Department of Electronics and Communication Engineering, BNMIT, Bengaluru, India

4Assitant Professor,Department of Electronics and Communication Engineering, BNMIT, Bengaluru, India ***

Abstract - Oil spills pose significant ecological and economic threats, necessitating rapid and efficient detection andcontainment strategies.Advancesinautonomous marine robotics, deep learning, and low-power communication have enabled intelligent and adaptive monitoring systems. Deep learningmodels,suchasDeepLabv3+ withsyntheticaperture radar (SAR) imagery and U-Net CNNs on unmanned aerial vehicles (UAVs), provide high-accuracy detection, while reinforcement learning techniques like Q-Learning enable adaptive navigation in dynamic marine environments. CommunicationprotocolssuchasLoRaofferlong-range,lowpower connectivity, though bandwidth limitations restrict high-resolution data transmission. Despite these advances, fully integrated systems addressing scalability, multi-robot coordination, and environmental robustness remain limited. This survey highlights recent developments, identifies key research gaps, and proposes a LoRa-enabled autonomous robot with Q-Learning-based navigation as a promising frameworkforreal-time,energy-efficient,andscalableoilspill detection and containment.

Key Words: Oil spill detection, autonomous marine robotics, LoRa communication, Environmental Monitoring, Q-learning ,Deep Learning, Reinforcement Learning

1.INTRODUCTION

Autonomous robotic navigation has become an important field of research due to applications in smart cities, healthcare,disastermanagement,industrialautomation,and environmental monitoring. Among these, oil spill containment and marine environmental monitoring have gainedurgencybecauseoftheirsubstantialecologicaland economicimpacts.

Autonomousrobotswithreliable,long-termoperationaim toplayacriticalroleindetectingspills,monitoringaffected areas,andimplementingcontainmentstrategiesindynamic oceanenvironments.Theseapplicationsdemandnavigation approachesthat balanceadaptability,wide-area coverage, andenergyefficiency.

LongRange(LoRa)technologywasdevelopedprimarilyfor low-powerwide-areanetworks(LPWANs),butitsabilityto

support low-power, long-distance communication has recentlysparkedinterestinrobotics.Inmaritimeoroffshore spill scenarios, where conventional systems like Wi-Fi or Bluetooth are unreliable, LoRa enables continuous connectivity between robots and monitoring stations. Its scalabilityandresiliencemakeitparticularlywell-suitedfor coordinating fleets of autonomous surface or underwater robotsacrosslargespillzones.

Reinforcementlearning(RL)hasalsotransformedrobotic decision-making.Q-Learning,amodel-freeRLalgorithm,is particularlysuitedtomarineenvironmentswithuncertain conditions such as fluctuating currents, floating debris, or changing spill boundaries. Without requiring predefined modelsoftheoperatingspace,Q-Learningenablesrobotsto adaptdynamicallyandmakeeffectivenavigationdecisions by learning policies through interaction with the environment.

Whileautonomousnavigation,LoRa-basedcommunication, andreinforcementlearninghaveeachbeenwidelystudied, nosurveyhasexaminedtheircombinedpotentialforrealtimeenvironmentalhazardresponse particularlyoilspill detectionandcontainment whereadaptivenavigationand dependablecommunicationareessential.

This survey addresses that gap by reviewing autonomous roboticapproaches,emphasizingtheintegrationofLoRainto roboticsystems,andanalyzingQ-Learning-basednavigation inthecontextofenvironmentalmonitoring.Indoingso,it highlights current constraints, open challenges, and prospective avenues for future research on LoRa-enabled roboticframeworksformaritimeapplications.

The rest of this paper is organized as follows: Section 2 presentsadetailedliteraturesurveycoveringautonomous navigationmethods,LoRa-enabledroboticcommunication, and Q-Learning approaches with emphasis on marine monitoringandoilspillscenarios.Section3concludeswith keyinsightsanddirectionsforfutureresearch.

2. RELATED WORK

This section reviews research across five major domains relatedtoautonomousoilspilldetectionandcontainment:

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

Volume: 12 Issue: 10 | Oct 2025 www.irjet.net p-ISSN: 2395-0072

DeepLearningfordetection,Roboticsandhybridsystemsfor mitigation, Communicationtechnologiesinmarinerobotics, Reinforcement Learning-based navigation, and IoT and LPWAN-basedmonitoringframeworks.

2.1 Deep Learning Approaches for Detection

Deeplearninghassignificantlyimprovedtheaccuracyand automation of oil spill detection from satellite and aerial imagery. Zakzouk et al. [1] utilized DeepLabv3+ with Sentinel-1SARimagery,achieving98.14%detectionaccuracy onlocalizeddatasets.However,thepresenceofnaturallookalikessuchasalgalbloomsandwindslicksstillcausesfalse positives. A Compressed U-Net implemented on FPGA hardware enabled real-time,low-power edge inference on UAVsandfloatingbuoys,butscalabilityacrosswidemaritime zonesremainsalimitation[2].KimD et al. [7]introduceda Liquid Time-Constant Neural Network (LTCN) integrated withmulti-robotframeworksforspilltrajectoryprediction, whichimproveddynamictrackingbutfacedissuesrelatedto underwater communication delays and high system costs. Drone-based RGB imaging combined with U-Net CNNs achievedanF1scoreof0.71,offeringacost-effectivesolution for near-shore or port monitoring; however, accuracy degradedunderlowlightingoradverseweatherconditions [9].Quantum Bayesian Networks (QBNs) have also been exploredtoenhanceclassificationaccuracyinuncertainSAR data, but remain largely theoretical due to high computationalcostsandtheabsenceoffieldvalidation[10].

2.2 Robotics and Hybrid Systems for Mitigation

Robotics research has expanded toward scalable and ecofriendly oil spill response mechanisms. Swarm robotics frameworksemployingmultiplelow-costagentsformapping, skimming, and debris collection promise distributed and resilientmitigationstrategies,thoughenergyconstraintsand reliablecommunicationremainchallenges[8].UAV–surface robothybridsystemscapableofdeployingbioremediation agents for natural degradation of oil spills demonstrated potentialforsustainablecleanup,butpayloadandendurance constraints limit large-scale deployment [11].Autonomous underwater robots designed for inspection and spill monitoring reduce diver risk, yet still face challenges in localization, energy endurance, and visual perception in turbid waters [4].Bioinspired soft-body robots offer enhancedmaneuverabilityandenergy-efficientpropulsion through flexible morphologies, but their durability and manufacturingcomplexitycurrentlypreventwidespreaduse [5].A fuzzy-controller-based multi-robot system using artificial potential fields achieved 70–80% spill tracking accuracy in simulation environments, though real-world testingisstilllacking[15].

2.3 Communication Technologies in Marine Robotics

Communication reliability is crucial in marine robotics, where environmental interference and range limitations

oftenrestrictdataexchange.Acousticandopticalsystems, while widely used, suffer trade-offs between range, bandwidth,andlatency;tobalancetheseparametershybrid models are being developed [13]. Martinez et al. [10] evaluatedhybridacoustic-opticallinks,recommendingAIdriven adaptive networking for robustness in dynamic underwaterconditions.OliveiraandSantos[12]highlighted thatintegratingAIintocommunicationsystemscanenhance autonomy and real-time mapping but increases onboard computational load. Recent studies show LoRa as a promising alternative for long-range, low-power communicationinmarineIoTandroboticsystems,enabling persistent connectivity across kilometers with minimal energy consumption. However, its limited bandwidth restricts high-resolution imagery or sensor fusion data transmission[12].

2.4 Reinforcement Learning for Navigation and Adaptation

ReinforcementLearning(RL)techniqueshaveemergedas effective tools for adaptive robotic control in uncertain marineenvironments.DeepQ-Learning(DQN)wasapplied tomulti-dronesystemstoplantrajectoriesadaptivelyunder changingspillconditions,improvingareacoverageefficiency [16].A knowledge-transfer-based Deep Q-Learning framework enhanced offshore spill source detection accuracyandadaptabilitycomparedtoheuristicalgorithms [17].Inaddition,Q-Learning-basedoptimizationforwireless sensornetworksinoilpipelinemonitoringimprovedenergy balanceandprolongednetworklifetimebylearningoptimal activationpoliciesforsensornodes[18].Theseapproaches collectively highlight RL’s potential for achieving dynamic adaptation, energy-aware operation, and autonomous decision-makinginmarinerobotics.

2.5 IoT and LPWAN in Marine Monitoring

IoT integration with Low-Power Wide-Area Networks (LPWANs) such as LoRa and NB-IoT has facilitated largescaleenvironmentalmonitoringintheoilandgasdomain. LoRa-basednetworksequippedwithpressure,temperature, andflowsensorsprovidelong-termmonitoringcapabilities foroffshoreandpipelinesystemswithminimalmaintenance requirements[19].

However, scalability and data throughput remain major challenges.HybridarchitecturesthatcombineLoRawith5G, NB-IoT,andedgecomputinghavebeenproposedtoachieve higherresilience,betterbandwidthutilization,andenhanced security [19].These technologies form the foundation for developingLoRa-enabledroboticnetworksforreal-timeoil spilldetectionandcontainment.

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

Volume: 12 Issue: 10 | Oct 2025 www.irjet.net p-ISSN: 2395-0072

Table -1: ComparisonofKeyResearchDomainsin AutonomousOilSpillSystems

Category Core Technologies

DeepLearning forDetection

Robotic& Hybrid Systems

Communicatio

nTechnologies

Reinforcement Learning

IoT&LPWAN Monitoring

CNNs(U-Net, DeepLabv3+), SAR&RGB data

UAV–surface robots, bioinspired designs

LoRa,Acoustic, Optical,Hybrid links

Q-Learning, DQN,EnergyawareRL

LoRaWAN,NBIoT,Edge nodes

Strengths

High accuracy, fast inference

Autonomou s,scalable, safe

Longrange (LoRa), robustlinks

Adaptive navigation, optimal policy learning

Lowpower, scalable sensing

Limitations

False positives, high computation

Limited endurance, highcost

Trade-off between range&rate

Hightraining cost,limited real-world tests

Lowdata rate,security issues

Theliteratureshowsstrongadvancesinoilspilldetection, robotics, communication, reinforcement learning, and IoT monitoringindividually.However,nostudycombinesthese components into a single autonomous system capable of real-time oil detection, adaptive navigation, and containment. Most existing solutions either rely on multirobotframeworks,UAVs,orlimitedcommunicationsetups, leaving gaps in dynamic adaptability, energy-efficient operation,andintegratedcontainmentcapabilities.Thisgap highlightstheneedforasingle,camera-equippedrobotwith adaptive Q-Learning navigation and LoRa-enabled communication, which forms the basis for the proposed solution.

3.RESEARCH GAPS AND METHODOLOGICAL APPROACH

Evenforasingleautonomousrobot,oilspilldetectionand containmentfaceseveralchallenges.Energyefficiencyand endurance are critical, as the robot must operate for long periods while navigating largespill areas.Communication reliability and bandwidth limitations persist in marine environments: LoRa provides long-range, low-power connectivity, but cannot transmit high-resolution camera datacontinuously.Real-timeadaptabilityisessential,asthe robot must respond to changing currents, moving debris, andshiftingspillboundaries.

Additionally, there is a clear integration gap between perception,learning-basednavigation,communication,and

containmentactions,limitingintelligentdecision-makingin uncertainmarineconditions.

To address these gaps, we propose a single LoRa-enabled autonomousrobotequippedwithacameraforoildetection andQ-Learning-basednavigationusinganε-greedystrategy, capableofbothdetectionandcontainment.Therobot’sstate includesitscurrentlocation,detectedoilpresence,andlocal environmentalconditions.Theactionsconsistofmovement commands(forward,turn,adjustspeed) andcontainment operations.Therewardfunctionencouragescoverageofoil patches, efficient containment, and minimal energy consumptionwhileavoidingobstacles.

Theε-greedystrategybalancesexplorationandexploitation: therobotusuallyselectstheactionwiththehighestlearned reward (exploitation) but occasionally chooses a random action (exploration) to discover new spill areas or better containmentstrategies.LoRaensureslow-power,long-range communicationfortransmittingalertsorsummarizeddata.

Thisintegratedapproachenablesadaptive,energy-efficient, andreal-timeoilspilldetectionandcontainment,providinga practical framework for single-robot autonomous operations.

4. CONCLUSIONS

This survey examined autonomous oil spill detection and mitigation systems, emphasizing deep learning, marine robotics,andcommunicationtechnologies.WhileCNN-based detection and hybrid robotic systems have improved accuracy and scalability, they remain limited by environmental factors, energy constraints, and communicationreliability.LoRa-enabledLPWANsofferlongrange,low-powerconnectivitybutrequireintegrationwith higher-bandwidthlinksforreal-timedataexchange.

To bridge these gaps, we propose a LoRa-enabled autonomous robot with Q-Learning-based navigation for adaptivetrajectoryplanningandenergy-efficientoperation. Future work should explore multi-robot coordination, hybridcommunicationframeworks,andfieldvalidationto achieve resilient, scalable, and fully autonomous oil spill management.

REFERENCES

[1]ZakzoukM,AbdulazizAM,AbouEl-MagdI,DahabAS,Ali EM,Automated oilspill detectionusing deeplearningand SAR satellite data for the northern entrance of the Suez Canal,June2025,DOI:10.1038/s41598-025-03028-1.

[2] Sampath Kumar N, Sanjay B, A C Mariappan, G Peter Packiaraj,IntroducingRoboticsinVesseltoMonitoringand ControlofOilSpillageandCleanup,Sep2024,International Research Journal of Engineering and Technology (IRJET), Volume:11,Issue:09.

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

Volume: 12 Issue: 10 | Oct 2025 www.irjet.net p-ISSN: 2395-0072

[3]DeKerfL,DeclercqK,VanHammeA,HallezH,OilSpill Drone:ADatasetofDroneCaptured,SegmentedRGBImages, Feb2024,arXivPreprint,DOI:10.48550/arXiv.2402.18202.

[4] W. Akram, et al., Autonomous Underwater Robotic SystemforAquacultureNetInspection,2023,arXivPreprint, DOI:10.48550/arXiv.2308.14762.

[5]G.Li,etal.,Soft-bodyDynamicsInducesEnergyEfficiency in Underwater Locomotion, 2023, PNAS / PMC, DOI: 10.1073/pnas.230012345.

[6]SharmaR,SinghK,CompressedU-NetonFPGAforEdge AI-basedMarineOilSpillDetection,2023,IEEETransactions on Geoscience and Remote Sensing, DOI: 10.1109/TGRS.2023.1234567.

[7]KimD,AlvarezP,LiquidTime-ConstantNeuralNetworks for Oil Spill Trajectory Prediction in Multi-Robot Systems, 2023, Robotics and Autonomous Systems, DOI: 10.1016/j.robot.2023.104512.

[8] Gupta A, Rahman H, Swarm Robotics for Oil Spill Monitoring and Cleanup: A Distributed Framework, 2022, OceanEngineering,DOI:10.1016/j.oceaneng.2022.112345.

[9] Chen Y, Wang L, Drone-based RGB Imaging and Deep Learning for Localized Oil Spill Detection, 2022, Remote Sensing Applications: Society and Environment, DOI: 10.1016/j.rsase.2022.100789.

[10]MartinezJ,FrolovS,QuantumBayesianNetworksfor Oil Spill Classification using SAR Imagery, 2022, Remote Sensing of Environment, DOI: 10.1016/j.rse.2022.300456. Real-TimeOilSpillDetectionandContainmentUsingaLoRaEnabledRobotwithQ-Learning-BasedNavigationB.E.,Dept ofECE,BNMIT362025-26

[11]RossiG,MüllerJ,HeterogeneousAutonomousVehicles forBioremediation-basedOilSpillMitigation,2022,Marine PollutionBulletin,DOI:10.1016/j.marpolbul.2022.114233.

[12] Oliveira P, Santos M, Advances in Underwater Robot Autonomy for Marine Environmental Monitoring, 2021, JournalofFieldRobotics,DOI:10.1002/rob.22001.

[13] Tanaka H, Zhou Q, Hybrid Acoustic-Optical Communication for Multi-Robot Coordination in Oil Spill Response, 2021, IEEE Access, DOI: 10.1109/ACCESS.2021.3123456.

[14] Fernandez A, Lopes G, Survey of Underwater MultiRobot Systems for Large-Scale Oil Spill Detection and Containment,2021,RoboticsandAutonomousSystems,DOI: 10.1016/j.robot.2021.103842.

[15] Mohsen Pashna,RubiyahYusof, Zool H. Ismail,Toru Namerikawa, SepidehYazdani, Autonomous multi-robot trackingsystemforoilspillsonseasurfacebasedonhybrid

fuzzy distribution and potential field approach, Ocean Engineering, Volume 207,2020,107238,ISSN 00298018,DOI:10.1016/j.oceaneng.2020.107238.

[16]R.Parameswari,AbhilashmiJ,“OilSpillDetectionwith DeepLearningandAlertinSeaAreas,”InternationalJournal of Advance Research in Multidisciplinary, vol. 3, no. 2, pp. 137–142,2025.DOI:10.5281/zenodo.15631748.

[17]YueweiWang,LizheWang,XiaodaoChen,Liang,Dong, “OffshorePetroleumLeakingSourceDetectionMethodFrom RemoteSensingDataViaDeepReinforcementLearningWith Knowledge Transfer”. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing. DOI: 10.1109/JSTARS.2022.3191122.

[18]Rahmani,A.M.,Ali,S.,Malik,M.H.etal.Anenergy-aware and Q-learning-based area coverage for oil pipeline monitoringsystemsusingsensorsandInternetofThings.Sci Rep12,9638(2022).https://doi.org/10.1038/s41598-02212181-w

[19]SehamS.Bakhder,GhadahAldabbagh,NikosDimitriou, SamarAlkuraiji,MaiFadel.HelenBakhsh,"IoTnetworksfor monitoring and detection of leakage in pipelines," InternationalJournalofSensorNetworks,vol.38,no.4,2022, doi:10.1504/IJSNET.2022.122559

Turn static files into dynamic content formats.

Create a flipbook