Decision Process Model Research Articles

An Intrusion System for Internet of Things Security Breaches Using Machine Learning Techniques

Effective identification and categorization of network attacks are paramount for ensuring robust security. However, contemporary techniques often struggle to accurately discern and classify novel attack patterns. This research introduces an innovative framework designed to achieve reliable attack detection and classification by harnessing the synergistic capabilities of utilizing DenseNet convolutional neural networks and rap music analysis techniques. Our approach leverages feature extraction through the Attention Pyramid Network (RAPNet) framework, tailored to extract pertinent features from input data, alongside binary Pigeon optimization. Subsequently, we employ feature selection using the optimization algorithm (BPOA). Once the optimal features are identified, we employ the Densenet201 model to categorize attacks across various datasets, including Bot-IoT, CICIDS2017, and CICIDS2019, through deep learning methodologies. To address the challenge posed by imbalanced data, we introduce conditional generative adversarial networks for generating additional data samples for minority classes, thus mitigating the issue. In contrast to recent intrusion detection methods, our results showcase the model’s exceptional precision in detecting and categorizing achieving accuracy rates of 99.12%, 99.01%, and 99.18% for Bot-IoT, CICIDS2017, and CICIDS2019 datasets, respectively. Despite the potential benefits of a machine learning-based intrusion detection system (IDS) for Internet of Things (IoT) security, several limitations must be considered. These include the lack of standardized security protocols across various IoT devices and platforms, which makes it challenging to develop a uniform IDS. Furthermore, machine learning models, including those for intrusion detection, can be vulnerable to adversarial attacks that can circumvent or mislead the model’s decision-making process. Thus, the potential for sophisticated attacks on IoT systems must be considered when developing such a system.

Explainable ensemble technique for enhancing credit risk prediction

<span>Credit risk prediction is a critical task in financial institutions that can impact lending decisions and financial stability. While machine learning (ML) models have shown promise in accurately predicting credit risk, the complexity of these models often makes them difficult to interpret and explain. The paper proposes the explainable ensemble method to improve credit risk prediction while maintaining interpretability. In this study, an ensemble model is built by combining multiple base models that uses different ML algorithms. In addition, the model interpretation techniques to identify the most important features and visualize the model's decision-making process. Experimental results demonstrate that the proposed explainable ensemble model outperforms individual base models and achieves high accuracy with low loss. Additionally, the proposed model provides insights into the factors that contribute to credit risk, which can help financial institutions make more informed lending decisions. Overall, the study highlights the potential of explainable ensemble methods in enhancing credit risk prediction and promoting transparency and trust in financial decision-making.</span>

An intelligent multi-layer framework with SHAP integration for botnet detection and classification

In increasingly interconnected digital world, the threat of cyber-attacks and data breaches are a pervasive and growing concern. The Botnets are the most dangerous threats that launch a wide range of cyberattacks such as distributed denial of service (DDoS) attacks, sending spam emails, spreading malware, and stealing sensitive information. The identification and categorization of botnets has become a highly complex and critical process due to the massive amount of network traffic generated every second. This paper presents a multilayer botnet detection and classification framework based on explainable machine learning algorithms. The proposed approach comprises of three distinct layers: Feature Selection Layer, Botnet Detection Layer and Botnet Classification Layer. The Feature Selection Layer reduces the dataset into six essential features, that enhance the accuracy and efficiency of the framework. The botnet detection layer detects botnet activity by filtering the reduced dataset into normal packets and botnet packets. The filtered botnet packets further examined by botnet classification layer to classify the botnet packets into different types of botnets. Moreover, SHAP (SHapley Additive exPlanations) technique is utilized to provide transparency to the model's decision-making process. To evaluate the effectiveness of the proposed model, NCC-2 and CTU-13 datasets are examined. 10-fold cross-validation technique is applied to validate the experimental findings. The average accuracy achieved by proposed approach in botnet detection stands impressively at 99.98%, while the average accuracy achieved in classifying botnet families is 99.30%, exhibiting an exceptional level of performance. The comparative analysis is performed which demonstrates its superiority over existing methods in terms of accuracy, precision, recall, F1-Score.

Finding the optimal exploration-exploitation trade-off online through Bayesian risk estimation and minimization

We propose endogenous Bayesian risk minimization (EBRM) over policy sets as an approach to online learning across a wide range of settings. Many real-world online learning problems have complexities such as action- and belief-dependent rewards, time-discounting of reward, and heterogeneous costs for actions and feedback; we find that existing online learning heuristics cannot leverage most problem-specific information, to the detriment of their performance. We introduce a belief-space Markov decision process (BMDP) model that can capture these complexities, and further apply the concepts of aleatoric, epistemic, and process risks to online learning. These risk functions describe the risk inherent to the learning problem, the risk due to the agent's lack of knowledge, and the relative quality of its policy, respectively. We demonstrate how computing and minimizing these risk functions guides the online learning agent towards the optimal exploration-exploitation trade-off in any stochastic online learning problem, constituting the basis of the EBRM approach. We also show how Bayes' risk, the minimization objective in stochastic online learning problems, can be decomposed into the aforementioned aleatoric, epistemic, and process risks.In simulation experiments, EBRM algorithms achieve state-of-the-art performance across various classical online learning problems, including Gaussian and Bernoulli multi-armed bandits, best-arm identification, mixed objectives with action- and belief-dependent rewards, and dynamic pricing, a finite partial monitoring problem. To our knowledge, it is also the first computationally efficient online learning approach that can provide online bounds on an algorithm's Bayes' risk. Finally, because the EBRM approach is parameterized by a set of policy algorithms, it can be extended to incorporate new developments in online learning algorithms, and is thus well-suited as the foundation for developing real-world learning agents.

Exploring the performance and explainability of fine-tuned BERT models for neuroradiology protocol assignment

BackgroundDeep learning has demonstrated significant advancements across various domains. However, its implementation in specialized areas, such as medical settings, remains approached with caution. In these high-stake environments, understanding the model's decision-making process is critical. This study assesses the performance of different pretrained Bidirectional Encoder Representations from Transformers (BERT) models and delves into understanding its decision-making within the context of medical image protocol assignment.MethodsFour different pre-trained BERT models (BERT, BioBERT, ClinicalBERT, RoBERTa) were fine-tuned for the medical image protocol classification task. Word importance was measured by attributing the classification output to every word using a gradient-based method. Subsequently, a trained radiologist reviewed the resulting word importance scores to assess the model’s decision-making process relative to human reasoning.ResultsThe BERT model came close to human performance on our test set. The BERT model successfully identified relevant words indicative of the target protocol. Analysis of important words in misclassifications revealed potential systematic errors in the model.ConclusionsThe BERT model shows promise in medical image protocol assignment by reaching near human level performance and identifying key words effectively. The detection of systematic errors paves the way for further refinements to enhance its safety and utility in clinical settings.

Unveiling psychotic disorder patterns: A deep learning model analysing motor activity time-series data with explainable AI

Deep Learning (DL) holds immense potential in revolutionizing healthcare, offering robust support to clinicians and enhancing patient care. However, coming up with the right DL model is always challenging and depends on quality, quantity, and type of data. In this paper, two motor activity datasets are utilized: “Depresjon” dataset includes activity recordings from 32 healthy individuals (402 days) and 23 individuals with unipolar and bipolar depression (291 days) and the “Psykose” dataset consists of 22 schizophrenia subjects (285 days) and 32 healthy subjects (402 days). The motor activity data, represented as time-series signals, poses challenges due to variable lengths and non-uniform starting timestamps for each subject. Additionally, the daytime and nighttime distributions differ across samples, requiring explicit handling for Convolutional Neural Network models. To address this issue, a multi-branch DL architecture is employed with one branch fed with nighttime data and another with daytime data, capable of capturing features across various scales, accommodating patterns of different sizes. Moreover, the combined outputs of these branches are subjected to a self-attention-mechanism (MultiHeadAttention), which prioritizes essential features. The use of Gradient weighted Class Activation Map (Grad-CAM) technique aids in comprehending the model's decision-making process. The benchmark datasets were used to validate the model, which exhibited an accuracy of 0.94 for both classifying depressive and schizophrenic episodes from control subjects. An accuracy of 0.81 for classifying depressive episodes, and schizophrenics from control samples. This accuracy further increases when combining the control samples from both datasets, to 0.97 for depression and 0.98 for schizophrenia.

Glaucoma detection model by exploiting multi-region and multi-scan-pattern OCT images with dynamical region score.

Currently, deep learning-based methods have achieved success in glaucoma detection. However, most models focus on OCT images captured by a single scan pattern within a given region, holding the high risk of the omission of valuable features in the remaining regions or scan patterns. Therefore, we proposed a multi-region and multi-scan-pattern fusion model to address this issue. Our proposed model exploits comprehensive OCT images from three fundus anatomical regions (macular, middle, and optic nerve head regions) being captured by four scan patterns (radial, volume, single-line, and circular scan patterns). Moreover, to enhance the efficacy of integrating features across various scan patterns within a region and multiple regional features, we employed an attention multi-scan fusion module and an attention multi-region fusion module that auto-assign contribution to distinct scan-pattern features and region features adapting to characters of different samples, respectively. To alleviate the absence of available datasets, we have collected a specific dataset (MRMSG-OCT) comprising OCT images captured by four scan patterns from three regions. The experimental results and visualized feature maps both demonstrate that our proposed model achieves superior performance against the single scan-pattern models and single region-based models. Moreover, compared with the average fusion strategy, our proposed fusion modules yield superior performance, particularly reversing the performance degradation observed in some models relying on fixed weights, validating the efficacy of the proposed dynamic region scores adapted to different samples. Moreover, the derived region contribution scores enhance the interpretability of the model and offer an overview of the model's decision-making process, assisting ophthalmologists in prioritizing regions with heightened scores and increasing efficiency in clinical practice.

Integrating the Coupled Markov Chain and Fuzzy Analytic Hierarchy Process Model for Dynamic Decision Making

This paper introduces a pioneering model that merges coupled Markov chains (CMC) with the fuzzy analytic hierarchy process (FAHP) to enhance multi-criteria decision making (MCDM) amidst the dynamic interplay of criteria. Traditional MCDM frameworks often lack the granularity to manage the intricate and changing relationships among criteria. Our model addresses this gap by integrating fuzzy numbers into AHP, providing a nuanced means to handle the inherent uncertainty of decision criteria. The application of the Kronecker product in CMC enriches our approach, offering a data-driven analysis while mitigating excessive dependence on subjective expert opinion. A comprehensive numerical example underlines the model’s improved decision-making accuracy and efficiency, marking a substantial advancement in MCDM methodologies. This research contributes to the field by presenting a sophisticated yet practical framework for dynamic decision analysis in complex uncertain environments.

Power Allocation Based on Multi-Agent Deep Deterministic Policy Gradient for Underwater Acoustic Communication Networks

This paper proposes a reinforcement learning-based power allocation for underwater acoustic communication networks (UACNs). The objective function is formulated as maximizing channel capacity under constraints of maximum power and minimum channel capacity. To solve this problem, a multi-agent deep deterministic policy gradient (MADDPG) approach is introduced, where each transmitter node is considered as an agent. Given the definition of a Markov decision process (MDP) model for this problem, the agents learn to collaboratively maximize the channel capacity by deep deterministic policy gradient (DDPG) learning. Specifically, the power allocation of each agent is obtained by a centralized training and distributed execution (CTDE) method. Simulation results show the sum rate achieved by the proposed algorithm approximates that of the fractional programming (FP) algorithm and improves by at least 5% compared with the DQN (deep Q-learning network) -based power allocation algorithm.

Priority-Aware Actuation Update Scheme in Heterogeneous Industrial Networks.

In heterogeneous wireless networked control systems (WNCSs), the age of information (AoI) of the actuation update and actuation update cost are important performance metrics. To reduce the monetary cost, the control system can wait for the availability of a WiFi network for the actuator and then conduct the update using a WiFi network in an opportunistic manner, but this leads to an increased AoI of the actuation update. In addition, since there are different AoI requirements according to the control priorities (i.e., robustness of AoI of the actuation update), these need to be considered when delivering the actuation update. To jointly consider the monetary cost and AoI with priority, this paper proposes a priority-aware actuation update scheme (PAUS) where the control system decides whether to deliver or delay the actuation update to the actuator. For the optimal decision, we formulate a Markov decision process model and derive the optimal policy based on Q-learning, which aims to maximize the average reward that implies the balance between the monetary cost and AoI with priority. Simulation results demonstrate that the PAUS outperforms the comparison schemes in terms of the average reward under various settings.

SHELF FOR SERVING LINES DEVELOPMENT AND LAUNCHING TO THE MARKET

In today’s market, there is a very intense competition, as well as the unpredictability of customer desires and market trends that change at a rapid speed, resulting in a shorter product life cycle. The purpose of this article is to examine the industry of commercial kitchens and the main aspects could be the key of successfully launching shelf for serving lines to the market. The empirical research background for this study is defined by the performed survey-questionnaire. The results of the survey of enterprises that specialize in manufacturing professional kitchen equipment, as well as the results of the survey of customers who are interested in such products, are presented in this article. The aim of both surveys was to determine which success aspects the company should rely on, and which are the most important for successful product development and market launch. Based on the result of the research and the information which was gathered, a model for the successful development and launch of a shelf for serving lines was developed and proposed. It is a hybrid type model which is based on the principles of Stage-Gate model, Decision process model which was suggested by Kotler and Keller and Agile set of values and principles. It has 7 stages and, stages have 6 gates where the project is assessed every time. This model is a recombinational tool of launching the product to the market which could help to identify the problematic issues within the ongoing process in the company.

Policy evaluation for temporal and/or spatial dependent experiments

Abstract The aim of this article is to establish a causal link between the policies implemented by technology companies and the outcomes they yield within intricate temporal and/or spatial dependent experiments. We propose a novel temporal/spatio-temporal Varying Coefficient Decision Process model, capable of effectively capturing the evolving treatment effects in situations characterized by temporal and/or spatial dependence. Our methodology encompasses the decomposition of the average treatment effect into the direct effect (DE) and the indirect effect (IE). We subsequently devise comprehensive procedures for estimating and making inferences about both DE and IE. Additionally, we provide a rigorous analysis of the statistical properties of these procedures, such as asymptotic power. To substantiate the effectiveness of our approach, we carry out extensive simulations and real data analyses.

Stochastic Cyclic Inventory Routing with Supply Uncertainty: A Case in Green-Hydrogen Logistics

Hydrogen can be produced from water, using electricity. The hydrogen can subsequently be kept in inventory in large quantities, unlike the electricity itself. This enables solar and wind energy generation to occur asynchronously from its usage. For this reason, hydrogen is expected to be a key ingredient for reaching a climate-neutral economy. However, the logistics for hydrogen are complex. Inventory policies must be determined for multiple locations in the network, and transportation of hydrogen from the production location to customers must be scheduled. At the same time, production patterns of hydrogen are intermittent, which affects the possibilities to realize the planned transportation and inventory levels. To provide policies for efficient transportation and storage of hydrogen, this paper proposes a parameterized cost function approximation approach to the stochastic cyclic inventory routing problem. Firstly, our approach includes a parameterized mixed integer programming (MIP) model which yields fixed and repetitive schedules for vehicle transportation of hydrogen. Secondly, buying and selling decisions in case of underproduction or overproduction are optimized further via a Markov decision process (MDP) model, taking into account the uncertainties in production and demand quantities. To jointly optimize the parameterized MIP and the MDP model, our approach includes an algorithm that searches the parameter space by iteratively solving the MIP and MDP models. We conduct computational experiments to validate our model in various problem settings and show that it provides near-optimal solutions. Moreover, we test our approach on an expert-reviewed case study at two hydrogen production locations in the Netherlands. We offer insights for the stakeholders in the region and analyze the impact of various problem elements in these case studies. Funding: This project received funding from the Fuel Cells and Hydrogen 2 Joint Undertaking (now Clean Hydrogen Partnership) under [Grant Agreement 875090]. The Joint Undertaking receives support from the European Union’s Horizon 2020 research and innovation programme, Hydrogen Europe and Hydrogen Europe Research. A. H. Schrotenboer received support from the Dutch Science Foundation (Nederlandse Organisatie voor Wetenschappelijk Onderzoek; NWO) through [Grant VI.Veni.211E.043].

Proximal Policy Optimization for Efficient D2D-Assisted Computation Offloading and Resource Allocation in Multi-Access Edge Computing

In the advanced 5G and beyond networks, multi-access edge computing (MEC) is increasingly recognized as a promising technology, offering the dual advantages of reducing energy utilization in cloud data centers while catering to the demands for reliability and real-time responsiveness in end devices. However, the inherent complexity and variability of MEC networks pose significant challenges in computational offloading decisions. To tackle this problem, we propose a proximal policy optimization (PPO)-based Device-to-Device (D2D)-assisted computation offloading and resource allocation scheme. We construct a realistic MEC network environment and develop a Markov decision process (MDP) model that minimizes time loss and energy consumption. The integration of a D2D communication-based offloading framework allows for collaborative task offloading between end devices and MEC servers, enhancing both resource utilization and computational efficiency. The MDP model is solved using the PPO algorithm in deep reinforcement learning to derive an optimal policy for offloading and resource allocation. Extensive comparative analysis with three benchmarked approaches has confirmed our scheme’s superior performance in latency, energy consumption, and algorithmic convergence, demonstrating its potential to improve MEC network operations in the context of emerging 5G and beyond technologies.

A Causal Network-Based Markov Decision Process Model for Intervention Planning

Patients suffering from chronic diseases often receive interventions after they are discharged from hospitals for their health management. The intervention decisions can be made sequentially during the course of disease progression and treatment process, and timely and proper interventions can lead to better health condition and quality of life and reduce the rate of re-hospitalization. Moreover, to determine effective intervention plans, personalized care should be considered. However, in many cases, intervention plans are not tailored to patients’ health conditions or dynamically adjusted based on their status changes due to the lack of adequate analytical methods. To address this issue, we introduce a causal Bayesian network-based Markov decision process (CNMDP) model to dynamically evaluate the health status and effectiveness of interventions for each patient and update the corresponding intervention strategy. In such a model, a causal Bayesian network is integrated in order to predict the effect of interventions in each patient’s health status. The resulting risk at each decision point is considered in a Markov decision process model by examining diverse factors of a patient to make a personalized optimal decision on intervention planning to minimize the risk of health deterioration or readmission. To illustrate the applicability of the model and develop a specific formulation, an intervention decision model for chronic obstructive pulmonary disease (COPD) patients is presented using the proposed method. Such a model can be applied to other diseases as well to design personalized, prompt, and effective intervention plans. <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Note to Practitioners</i> —Substantial efforts have been devoted to treatment and care management of chronic diseases. Particularly, there has been an initiative to design appropriate follow-up plans and interventions for postdischarge care to reduce hospital readmissions. However, many intervention plans are not patient specific and do not change with patient conditions. To mitigate such an limitation, based on identified risk factors and causal relationships, a causal Bayesian network model is introduced to predict readmission risk and evaluate intervention effectiveness. The results are then embedded into a Markov decision process to derive optimal intervention plan to minimize readmission risk. Such a process is repeated dynamically to evaluate health status and update intervention strategies. Using chronic obstructive pulmonary disease (COPD) as an example of chronic disease, a case study at a community hospital is presented to illustrate the applicability of the model.

Enhancing substance identification by Raman spectroscopy using deep neural convolutional networks with an attention mechanism.

Raman spectroscopy is widely used for substance identification, providing molecular information from various components along with noise and instrument interference. Consequently, identifying components based on Raman spectra remains challenging. In this study, we collected Raman spectral data of 474 hazardous chemical substances using a portable Raman spectrometer, resulting in a dataset of 59 468 spectra. Our research employed a deep neural convolutional network based on the ResNet architecture, incorporating an attention mechanism called the SE module. By enhancing the weighting of certain spectral features, the performance of the model was significantly improved. We also investigated the classification predictive performance of the model under small-sample conditions, facilitating the addition of new hazardous chemical categories for future deployment on mobile devices. Additionally, we explored the features extracted by the convolutional neural network from Raman spectra, considering both Raman intensity and Raman shift aspects. We discovered that the neural network did not solely rely on intensity or shift for substance classification, but rather effectively combined both aspects. This research contributes to the advancement of Raman spectroscopy applications for hazardous chemical identification, particularly in scenarios with limited data availability. The findings shed light on the significance of spectral features in the model's decision-making process and have implications for broader applications of deep learning techniques in Raman spectroscopy-based substance identification.

Automated Bone Fracture Detection in X-ray Imaging to Improve Orthopaedic Diagnostics in Healthcare

The healthcare system in India faces an acute shortage of medical staff, significantly impacting the nation's ability to provide sufficient healthcare services. X-ray reports are paramount to orthopaedic surgeons for making critical patient care decisions. Implementing automated systems for detecting bone fractures offers a potentially efficient solution to supplement the limited workforce. However, these systems encounter substantial hurdles, such as inconsistent image quality and the risk of incorrect diagnosis due to the limitations of current technology. The research provides a specialised tool for identifying bone fractures with high precision (97% accuracy) and an excellent overall rating (F1 score of 0.98), showing it can reliably detect fractures. The assessment of this model is augmented by visual tools like heatmaps and performance graphs, providing an in-depth analysis of its operational effectiveness. Furthermore, creating a user-friendly interface is intended to streamline the adoption of this tool in clinical settings. Notwithstanding its demonstrated precision, the model must navigate data-related challenges, precisely data complexity and the representation of different categories, to prevent skewed outcomes. Clarifying the model's decision-making process and adaptability to various clinical situations are also critical. The study emphasises the need for thorough testing to ensure the model's dependability in diagnosing fractures before being deployed effectively in the medical field.

MARLens: Understanding Multi-Agent Reinforcement Learning for Traffic Signal Control Via Visual Analytics.

The issue of traffic congestion poses a significant obstacle to the development of global cities. One promising solution to tackle this problem is intelligent traffic signal control (TSC). Recently, TSC strategies leveraging reinforcement learning (RL) have garnered attention among researchers. However, the evaluation of these models has primarily relied on fixed metrics like reward and queue length. This limited evaluation approach provides only a narrow view of the model's decision-making process, impeding its practical implementation. Moreover, effective TSC necessitates coordinated actions across multiple intersections. Existing visual analysis solutions fall short when applied in multi-agent settings. In this study, we delve into the challenge of interpretability in multi-agent reinforcement learning (MARL), particularly within the context of TSC. We propose MARLens, a visual analytics system tailored to understand MARL-based TSC. Our system serves as a versatile platform for both RL and TSC researchers. It empowers them to explore the model's features from various perspectives, revealing its decision-making processes and shedding light on interactions among different agents. To facilitate quick identification of critical states, we have devised multiple visualization views, complemented by a traffic simulation module that allows users to replay specific training scenarios. To validate the utility of our proposed system, we present three comprehensive case studies, incorporate insights from domain experts through interviews, and conduct a user study. These collective efforts underscore the feasibility and effectiveness of MARLens in enhancing our understanding of MARL-based TSC systems and pave the way for more informed and efficient traffic management strategies.

Hamlet: Hierarchical Multi-Objective Reinforcement Learning Method for Charging Power Control of Electric Vehicles With Dynamic Quantity

Electric vehicle (EV) is emerging as an effective choice to reduce carbon emissions in the modern transportation system. However, the large access of EVs brings tremendous passive influence on the stability and performance of power grid operation. Thus, it is valuable to investigate the charging power management method to improve charging experience of EV drivers while maintaining normal bus voltage of the power system. This paper builds the charging power control problem of multiple EVs as a hierarchical Markov decision process (MDP) model and propose a hierarchical multi-objective reinforcement learning method (Hamlet) to obtain the real-time charging power control decisions. Specifically, by treating EVs with the same remaining charging time as the same agent, this paper addresses the dynamic state and action spaces issue and significantly reduce the input dimension of both state and action spaces. Besides, this paper quantifies drivers’ anxiety about the battery volume and distributes the comfort reward over the entire charging interval, which overcomes the delayed comfort reward issue. In order to alleviate the influence of multiple objectives on training stability, this paper develops a multi-objective learning method to dynamically adjust the optimization direction. Simulation results on IEEE 33 bus and 69 bus test feeders prove the validity of the proposed method in mitigating voltage fluctuation, minimizing bill payment and maximizing battery comfort.

A POMDP model-based online risk mitigation method for autonomous inland vessels

Autonomous surface vessels (ASVs) increasingly gain appeal in the maritime industry for their high efficiency and improved navigational capabilities. However, risks originating from various internal and external factors such as faults, traffic, harsh weather conditions, etc., can affect their guidance and control capabilities and impact nominal vessel operations. The existing risk mitigation methods mainly focus on the vessel's guidance system and do not consider unsafe actions due to the control system. In this paper, we propose a new method based on a partially observable Markov decision process (POMDP) model for the online risk mitigation of autonomous inland vessels. The POMDP model-based method utilizes information about situational awareness to assist the vessel's planning and control system in real-time decision-making during hazardous situations, thereby ensuring that the vessel remains in a minimum-risk condition. Based on the identified risk-influencing factors (RIFs), the transition probabilities are updated by a Bayesian belief network (BBN). A case study of an autonomous inland vessel navigating in a confined waterway is presented to demonstrate the capability of the proposed method.