Robust Decision Research Articles

A data-driven distributionally robust optimization approach for the core acquisition problem

Reusing electric vehicles (EV) batteries that reach the end of their useful first life is an environmental and cost-competitive option; however, the process of recycling EV batteries is not yet mature. Due to complex electrochemical reactions and physical conditions, the quality of used EV batteries (cores) is highly uncertain. The remanufacturer needs to make the acquisition decision under quality distributional ambiguity. Perfect quality distribution of cores cannot be known to the remanufacturer in practice. We develop distributionally robust optimization models based on phi-divergence measures and the imprecise Dirichlet model (DRO-IDM) to derive robust decisions. First, we find that the bounds of quality probability intervals are identified solely based on the collected data by introducing the imprecise Dirichlet model. The derived finite-sample boundary can reduce the scope of the uncertainty set and avoid the no-direction search issue. Second, our models can hedge against distributional uncertainty, reduce the probability of a robust solution that deviates from the optimal solution, and correct bias in decision making. Third, we extend the DRO-IDM to develop data-driven models, that can reassess the value of multisource quality information to improve the estimation accuracy of core quality and maximize the remanufacturer’s profit. Our study provides new insights for remanufacturers: the new remanufacturing process proposed in our work can assist remanufacturers in utilizing the values of cores without disassembly; the information-aware algorithm can offer the remanufacturing sector a valuable tool for efficiently filtering out invalid information in optimizing acquisition decisions; this capability empowers decision-makers to leverage multiple sources of information and expedite the process of digital transformation in remanufacturing; our approach can also provide a manner of integrating information fusion and distribution learning into remanufacturing.

A two-stage robust decision-making framework (2S-RDM) for flood risk adaptation under deep uncertainty

Flood is one of the major challenges facing human societies. Adapting to future flood risks involves deep uncertainty, especially when long-term projections of climate change are considered. This study proposed a Two-stage Robust Decision Making (2S-RDM) framework to help devise flexible and robust strategies capable of addressing the inherent deep uncertainty associated with managing flood risks. Taking the Yangtze River Basin in China as a case study, we simulated flood risks across ∼0.6 million scenarios until 2050. This analysis considered four types of uncertain factors, i.e., future climate change, socio-economic growth, industrial structure transformation, and population aging. We then examined the effectiveness of four adaptation measures and their combinations, i.e. building elevation, tunnel construction, people relocation, and river basin conservation. Our projections show that without immediate adaptation, an estimated 0.9 to 27.3 million people will be impacted by floods until 2050, accompanied with $33.8 to $198.5 billion economic losses in the entire basin. When defining the goal as limiting the affected population <0.05% and ensuring economic losses <0.02%, we identified 24 global robust strategies capable of meeting this criterion in >80% of scenarios. Then, we compared the 24 global robust strategies regarding their relative costs and performances in each of the future scenario pools. The final recommended solutions are hybrid strategies that integrate engineering-based measures with ‘soft’ adaptation options (e.g. Elevation++, Tunnel++, and Relocation). This study provides tools to design flood adaptation strategies not only robust across diverse scenarios but also flexible for decision-makers to customize and refine their strategies based on specific needs.

Symbol error rate minimization using deep learning approaches for short-reach optical communication networks

Short-reach optical communication networks have various applications in areas where high-speed connectivity is needed, for example, inter- and intra-data center links, optical access networks, and indoor and in-building communication systems. Machine learning (ML) approaches provide a key solution for numerous challenging situations due to their robust decision-making, problem-solving, and pattern-recognition abilities. In this work, our focus is on utilizing deep learning models to minimize symbol error rates in short-reach optical communication setups. Various channel impairments, such as nonlinearity, chromatic dispersion (CD), and attenuation, are accurately modeled. Initially, we address the challenge of modeling a nonlinear channel. Consequently, we harness a deep learning model called autoencoders (AEs) to facilitate communication over nonlinear channels. Furthermore, we investigate how the inclusion of a nonlinear channel within an autoencoder influences the received constellation as the optical fiber length increases. Another facet of our work involves the deployment of a deep neural network-based receiver utilizing a channel influenced by chromatic dispersion. By gradually extending the optical length, we explore its impact on the received constellation and, consequently, the symbol error rate. Finally, we incorporate the split-step Fourier method (SSFM) to emulate the combined effects of nonlinearities, chromatic dispersion, and attenuation in the optical channel. This is accomplished through a neural network-based receiver. The outcome of this work is an evaluation and reduction of the symbol error rate as the length of the optical fiber is varied.

A robust optimization approach for a two-player force-design game

We present a new approach to force design that relies on robust decision making with a min–max objective, rather than assumptions about the goals and strategy of an opponent. This idea is explored mathematically in the framework of a round-based two-player Stackelberg game representing an arms race, which features the acquisition of assets by both players and an evaluation of the defensive capability against attack from an opponent using a portfolio of possible tactics. Mathematical analysis has been carried out to determine the structure of optimal strategies for this type of game. This allows the strategy of the first player to be represented as a decision tree with possible moves by the second player consisting of convex combinations of extreme points. Using this insight into the structure of solutions, the optimal strategy for the game can be computed using a large linear program. The effectiveness of this approach is demonstrated using numerical examples.

Handover management procedures for future generations mobile heterogeneous networks

Handover (HO) management in Heterogeneous Networks (HetNets) poses challenges arising from network densification and dynamic environmental behaviors. Existing HO decision algorithms struggle to efficiently utilize network resources and ensure a high-quality user experience amidst the complexity of HetNets and the burgeoning growth of mobile users. This paper introduces a robust and data-driven HO decision model designed to enhance HO performance in HetNets. Initially, a conventional HO decision algorithm is developed based on users' Reference Signal Received Power (RSRP) values in MATLAB. Various simulation cases explore different HO parameters to observe their impact on handover performance. To address these challenges, a data-driven HO decision model leveraging Long Short-Term Memory (LSTM), a deep learning technique, is proposed for the regression task. The LSTM model is trained and tested using obtained RSRP values, and the future RSRP values predicted by the model are employed to trigger HO decisions in the proposed algorithm. Results from the traditional HO decision algorithm are compared with those of the proposed machine learning-based approach across various simulation runs, considering average Signal-to-Interference-plus-Noise Ratio (SINR), RSRP, user throughput values, the number of HOs and the Radio Link Failure (RLF) ratio. Different user speeds are also considered to establish a relationship between HO frequency and mobile user speed. The proposed model achieved reducing the rate of radio link failure to levels that are deemed acceptable in order to ensure a continuous connection.

Identifying and Interpreting Hydrological Model Structural Nonstationarity Using the Bayesian Model Averaging Method

Understanding hydrological nonstationarity under climate change is important for runoff prediction and it enables more robust decisions. Regarding the multiple structural hypotheses, this study aims to identify and interpret hydrological structural nonstationarity using the Bayesian Model Averaging (BMA) method by (i) constructing a nonstationary model through the Bayesian weighted averaging of two lumped conceptual rainfall–runoff (RR) models (the Xinanjiang and GR4J model) with time-varying weights; and (ii) detecting the temporal variation in the optimized Bayesian weights under climate change conditions. By combining the BMA method with period partition and time sliding windows, the efficacy of adopting time-varying model structures is investigated over three basins located in the U.S. and Australia. The results show that (i) the nonstationary ensemble-averaged model with time-varying weights surpasses both individual models and the ensemble-averaged model with time-invariant weights, improving NSE[Q] from 0.04 to 0.15; (ii) the optimized weights of Xinanjiang model increase and that of GR4J declines with larger precipitation, and vice versa; (iii) the change in the optimized weights is proportional to that of precipitation under monotonic climate change, as otherwise the mechanism changes significantly. Overall, it is recommended to adopt nonstationary structures in hydrological modeling.

Assessing Water Management Strategies in Data-Scarce Mountain Regions under Uncertain Climate and Socio-Economic Changes

Water management in mountainous regions faces significant challenges due to deep uncertainties arising from data scarcity, knowledge gaps, and the complex interplay of climate and socio-economic changes. While existing approaches focused on uncertainty reduction and water system optimization contribute to managing uncertainties, they often require probability distributions that can be difficult to obtain in data-scarce mountain regions. To address these challenges, we demonstrate the effectiveness of Exploratory Modeling and Analysis (EMA) in assessing water management strategies and identifying operational ranges that avoid future water scarcity. Through a case study in the complex and data-scarce Peruvian Andes, we employed EMA to run 12,000 simulations by 2050, incorporating deep uncertainties from climate and socio-economic scenarios, and hydrological modeling parameters. This analysis identified specific policy combinations demonstrating greater robustness across diverse scenarios and uncertainties. EMA explicitly identifies operational ranges of policies to avoid water scarcity but also highlights the conditions that might trigger policy failure. We also delve into the roles of the different factors used in EMA and their significance in water management applications. Our research illustrates that an exploratory hydrological modeling approach based on robust decision-making can foster a more informed decision-making process for long-term water adaptation in rapidly changing mountain regions under data scarcity and deep uncertainties.

A dynamical neural network approach for distributionally robust chance-constrained Markov decision process

A dynamical neural network approach for distributionally robust chance-constrained Markov decision process

Programmatic Assessment: New Approach of Student Assessment in Medical Education of the Current Century

Over the past few decades, education programs in health professions have moved toward a competency-based paradigm throughout the world. Competency-based education (CBE) relies on the fundamental principle that predetermined outcomes (competencies) guide teaching, learning, and assessment to ensure that graduates demonstrate proficiency in key competency domains and are able to provide care for patients with high quality throughout their careers. The purpose of CBE is to transform learners into healthcare workers who have lifelong learning and are committed to professional excellence, and the necessary competencies have been created in them. In CBE, learners are usually the center of attention and are actively involved in the learning and assessment process. Providing frequent and meaningful performance feedback enables learners to shape their learning paths by identifying appropriate learning opportunities for further development. Assessment is critical in achieving the goals of CBE. However, traditional approaches to assessment, which generally focus exclusively on the summative function of assessment, are not appropriate in CBE. Assessment programs should not only ensure robust decision-making about the progress and development of learners' competence but also facilitate the production of high-quality feedback for learning and support reflective practice and the use of feedback. Therefore, to continuously improve performance in CBE, assessment programs and assessment functions should be combined and integrated for learning. Programmatic assessment is a comprehensive approach in the assessment system that is theoretically aligned with key principles of CBE, as it aims to optimize assessment for learning while ensuring informed decision-making about learners' achievement of intended outcomes. In programmatic assessment, low-stakes assessments are continuously and purposefully considered to provide meaningful feedback on the longitudinal performance of learners to develop their competencies, while high-stakes assessments are based on the meaningful aggregation of data from multiple low-stakes assessments, which are collected over longer periods of time, at different periods in different contexts, and with multiple raters. However, while programmatic assessment approaches are increasingly being accepted as essential tools in the effective implementation of CBE, various research findings have indicated that it is difficult to merge different assessment performances in programmatic assessment. Performing programmatic assessment requires a change not only in assessment design but also in instructors and learners' views and data about objective and high-quality evaluation performance. Therefore, changing the assessment culture and fundamental data is necessary. Capacity building is a key element in any change process. All stakeholders involved in programmatic assessment must develop the capabilities needed to bring about the desired change in assessment culture. However, research findings suggest that stakeholders may not always be willing or able to perform the new roles as intended for them; instead, old behaviors "simply" adapt to new environments.

Bayesian variational time-lapse full waveform inversion

SUMMARY Time-lapse seismic full-waveform inversion (FWI) provides estimates of dynamic changes in the Earth’s subsurface by performing multiple seismic surveys at different times. Since FWI problems are highly non-linear and non-unique, it is important to quantify uncertainties in such estimates to allow robust decision making based on the results. Markov chain Monte Carlo (McMC) methods have been used for this purpose, but due to their high computational cost, those studies often require a pre-existing accurate baseline model and estimates of the locations of potential velocity changes, and neglect uncertainty in the baseline velocity model. Such detailed and accurate prior information is not always available in practice. In this study we use an efficient optimization method called stochastic Stein variational gradient descent (sSVGD) to solve time-lapse FWI problems without assuming such prior knowledge, and to estimate uncertainty both in the baseline velocity model and the velocity change over time. We test two Bayesian strategies: separate Bayesian inversions for each seismic survey, and a single joint inversion for baseline and repeat surveys, and compare the methods with standard linearized double difference inversion. The results demonstrate that all three methods can produce accurate velocity change estimates in the case of having fixed (exactly repeatable) acquisition geometries. However, the two Bayesian methods generate significantly more accurate results when acquisition geometries changes between surveys. Furthermore, joint inversion provides the most accurate velocity change and uncertainty estimates in all cases tested. We therefore conclude that Bayesian time-lapse inversion using a joint inversion strategy may be useful to image and monitor subsurface changes, in particular where variations in the results would lead to different consequent decisions.

Illusion or reality? Building a metaverse community focused on value creation in the agricultural sector

Obsolete models, lack of professionalisation and an environment characterised by continuous changes have turned out to be a dangerous mix that has ended up stopping or slowing down the development of the agricultural sector. It is now necessary to design and carry out initiatives that counteract this trend, especially initiatives based on innovation and the use of technology. This article aims to analyse the implications of the new ways of working to be used in spaces created under the umbrella of a new technological order known as metaverse. Through the creation of a community of metaverses, capable of promoting interaction, collaboration and communication between the different agents involved in the value chain of the agricultural sector, the aim of this research is to create an integrated and collaborative trusting environment that allows consensual, well-founded and robust business decisions focused on cost savings and the optimisation of quality and productivity. This research seeks to benefit all agents involved in the agricultural sector value chain. By placing producers at the centre of the system, the approach described in this study aims to avoid the current imbalances between the interests of different agents, while promoting an innovative and sustainable agricultural sector committed to meeting the challenges of today’s society.

A robust decision-making framework to improve reservoir water quality using optimized selective withdrawal strategies

In arid locations such as Oman, improving the quality of water released from a reservoir can be daunting. This is particularly true for Wadi Dayqah Dam, which struggles to provide sufficient, high-quality water for downstream needs. The study's objective was to establish more reliable operational guidelines, as well as identify vulnerable outlets and release ratios. To do so, a multi-objective optimization algorithm (NSGA-II) was combined with a selective withdrawal approach, using CE-QUAL-W2 and an Artificial Neural Network to simulate water quality in a reservoir. Then, a Robust Decision Making (RDM) approach was utilized to diagnose solutions that maintain acceptable water quality according to users’ requirements. Finally, scenario discovery models (Patient Rule Induction Method (PRIM) and Classification and Regression Tree (CART)) were used to calculate vulnerable outlets and release ratios based on the established threshold. The study adopted an XLRM framework to manage a proposed model under climate change scenarios and specify the effects of involved factors. XLRM stands for eXogenous factors, Levers used by the policy-makers, Relationships between L and X, and Measurement matrix to evaluate scenarios. This study found that a higher-quality reservoir release correlated with more activated outlets and decreased pollutant concentration. Evaluating robust solutions showed that vulnerability increased under scenarios with lower annual rainfall, and water released from low and mid-level outlets was more vulnerable than that from the surface. To enhance operational stability and reduce vulnerability, the suggestions advocate for a strategic approach in releasing varying water ratios from outlets situated at different levels. Specifically, the proposal emphasizes the controlled release of both smaller and larger volumes of water from outlets positioned at lower and higher levels within the dam structure, respectively. This deliberate variation in water release aims to bolster the overall resilience of the operational framework, contributing to a more robust and reliable system.

A fair bed allocation during COVID-19 pandemic using TOPSIS technique based on correlation coefficient for interval-valued pythagorean fuzzy hypersoft set

The relationship between two variables is an essential factor in statistics, and the accuracy of the results depends on the data collected. However, the data collected for statistical analysis can be unclear and difficult to interpret. One way to predict how one variable will change about another is by using the correlation coefficient (CC), but this method is not commonly used in interval-valued Pythagorean fuzzy hypersoft set (IVPFHSS). The IVPFHSS is a more advanced and generalized form of the Pythagorean fuzzy hypersoft set (PFHSS), which allows for more precise and accurate analysis. In this research, we introduce the correlation coefficient (CC) and weighted correlation coefficient (WCC) for IVPFHSS and their essential properties. To demonstrate the applicability of these measures, we use the COVID-19 pandemic as an example and establish a prioritization technique for order preference by similarity to the ideal solution (TOPSIS) model. The technique is used to study the problem of optimizing the allocation of hospital beds during the pandemic. This study provides insights into the importance of utilizing correlation measures for decision-making in uncertain and complex situations like the COVID-19 pandemic. It is a robust multi-attribute decision-making (MADM) methodology with significant importance. Subsequently, it is planned to increase a dynamic bed allocation algorithm based on biogeography to accomplish the superlative decision-making system. Moreover, numerical investigations deliberate the best decision structures and deliver sensitivity analyses. The efficiency of our encouraged algorithm is more consistent than prevalent models, and it can effectively control and determine the optimal configurations for the study.

Habitual consumption of anthocyanin is associated with improvements in mood and cognitive performance in older people with memory complaints

The role of flavonoids on cognitive performance in older adults has been intensively studied, with the subclass of anthocyanins showing promising outcomes(1). However, there is conflicting evidence in the case of individuals at high risk of developing dementia, namely those with mild cognitive impairment (MCI). A recent study has suggested that cognitive scores in people with MCI were higher in those who had higher anthocyanin intake (&gt;10 mg)(2). Baseline data from 65 participants of an ongoing clinical trial that had an MIS (Memory Index Score) score ≤ 13 with self-reported subjective memory complaints, (mean age 69.1y ± 6.2) were used to investigate the relationship between dietary anthocyanin intake and indices of mood and cognitive performance. Repeated 24-hour dietary intake was recorded through Intake24 (a computer-based program) for three days (2Xweekday and 1Xweekend day) and anthocyanin intake was quantified using the PhenolExplorer food composition database. The primary outcome of interest was the Buschke and Grober Free and Cued Selective Reminding Test-Immediate Recall (FCSRT+IR) (assesses auditory anterograde memory functioning), while other cognitive functions assessed included: Spot the Word-2 (assessing premorbid estimate); Oral Symbol Digit Test (speed of processing); List Sorting (working memory); Trail Making Test A &amp; B (speed of processing/executive function); and Verbal Fluency (language/semantic memory). Subjective memory complaints were assessed using the Memory Assessment Clinic-Q (MAC-Q) and mood was assessed using the Geriatric Depression Scale (GDS). Independent t-tests were used to compare differences in cognitive tasks and mood scores between high (&gt;10 mg/d) and low consumers of anthocyanins (&lt;10 mg/d). There was a trend for high anthocyanin consumers (n = 35, median = 44.87 (10.01, 177.83)) to score better on FCSRT-Delayed Free Recall scores (16.57 ± 3.74) compared to lower consumers (n = 30, median = 0.01 (0, 9.51)), (15.97 ± 0.18) with a mean difference (SE) of −1.06 (0.58) (p = 0.06) suggesting recall of 1.06 more words after a 20-30 minute delay. Higher consumers had a lower GDS score (1.77 ± 3.73) compared to lower consumers (3.73 ± 1.77), p = 0.01. Spot the word test scores (assesses premorbid verbal abilities using a robust lexical decision task) were higher for high anthocyanin consumers (53.06 ± 4.38) compared to lower consumers (50.40 ± 4.38), mean difference (SE) = −2.66 (1.10), p = 0.01, meaning participants with higher intake of dietary anthocyanin were able to point at 2.7 more real words than low consumers. Dietary consumption of anthocyanin in older adults with MIS is associated with beneficial effects on depressive scores and the ability to retrieve words. Further research is warranted to identify optimal dosage for recommended intake. This cross-sectional study used baseline data from a randomised controlled trial registered with the Australian New Zealand Clinical Trials Registry (ANZCTR):12622000065796.

Modeling of Gamma Index for Prediction of Pretreatment Quality Assurance in Stereotactic Body Radiation Therapy of the Liver.

The purpose of this study was to develop a predictive model to evaluate pretreatment patient-specific quality assurance (QA) based on treatment planning parameters for stereotactic body radiation therapy (SBRT) for liver carcinoma. We retrospectively selected 180 cases of liver SBRT treated using the volumetric modulated arc therapy technique. Numerous parameters defining the plan complexity were calculated from the DICOM-RP (Radiotherapy Plan) file using an in-house program developed in MATLAB. Patient-specific QA was performed with global gamma evaluation criteria of 2%/2 mm and 3%/3 mm in a relative mode using the Octavius two-dimensional detector array. Various statistical tests and multivariate predictive models were evaluated. The leaf speed (MILS) and planning target volume size showed the highest correlation with the gamma criteria of 2%/2 mm and 3%/3 mm (P < 0.05). Degree of modulation (DoM), MCSSPORT, leaf speed (MILS), and gantry speed (MIGS) were predictors of global gamma pass rate (GPR) for 2%/2 mm (G22), whereas DoM, MCSSPORT, leaf speed (MILS) and robust decision making were predictors of the global GPR criterion of 3%/3 mm (G33). The variance inflation factor values of all predictors were <2, indicating that the data were not associated with each other. For the G22 prediction, the sensitivity and specificity of the model were 75.0% and 75.0%, respectively, whereas, for G33 prediction, the sensitivity and specificity of the model were 74.9% and 85.7%%, respectively. The model was potentially beneficial as an easy alternative to pretreatment QA in predicting the uncertainty in plan deliverability at the planning stage and could help reduce resources in busy clinics.

Joint flexibility-risk managed distributed energy trading considering network constraints and uncertainty

This paper addresses the operational challenges in distribution systems caused by the integration of fluctuating renewable energy sources and rising peak loads from widespread electrification. We propose a new distributed energy coordination framework that facilitates seamless interoperation between peer-to-peer (P2P) energy trading and the operation of distributed flexibility resources. This framework ensures network constraints are respected while promoting Energy Trading Consistency (ETC). Our approach utilizes a two-stage dynamic energy and flexibility sharing mechanism. It achieves energy efficiency through demand-management and P2P energy trading in the day-ahead (DA) stage, followed by adjustments of flexible resources and flexibility sharing through a receding-horizon feedback mechanism inspired by model predictive control (MPC) in the real-time (RT) stage. High-risk scenarios and computational overhead are managed by employing conditional value-at-risk (CVaR) measures, as well as approximation of constraints through dual robust reformulation and linear decision rules. To uphold the autonomy and privacy of prosumers, we propose distributed algorithms based on the fast consensus alternating direction method of multipliers (ADMM). Numerical results demonstrate that our framework achieves a 20% reduction in social cost while providing a coherent and interactive P2P energy trading environment, facilitating reliable network operation under high penetration of renewable energy sources without third-party intervention.

Planning for wastewater infrastructure adaptation under deep uncertainty

Infrastructure in low-lying coastal areas faces challenges from climate change, sea level rise, and the impact of compound hazards. Dynamic adaptive pathways planning (DAPP) is increasingly being applied as a way of planning under deep uncertainty. Stress testing for robustness is an integral part of DAPP which provides decision-makers with confidence. We outline a seven-step approach—combining scoping workshops, systems mapping, DAPP, exploratory modelling, robust decision-making, real options analysis and validation workshops—to support decision-making for infrastructure in low-lying coastal areas. We apply the seven steps to two wastewater treatment plant (WWTP) case studies in New Zealand to quantify indicators, signals, triggers and adaptation thresholds within DAPP plans and to identify adaptation pathways that are robust against future uncertainty. Case study one focuses on the implementation of an existing DAPP at Helensville WWTP. Our modelling enabled the challenge of quantifying indicators for adaptation thresholds and triggers to be overcome. We show that an adaptation threshold occurs at 31 cm of RSLR, the trigger point is sufficient lead time to enable relocation, and the indicator is the rate of observed RSLR. Case study one demonstrates in a quantitative way how an existing DAPP can be functionally implemented by a water management agency. Modelling for case study two, the Seaview WWTP, showed that 26 cm and 56 cm of RSLR are key thresholds. Nuisance flooding may occur after 26 cm of RSLR, which could happen as early as 2040 under a high emissions scenario. Inundation of plant assets may occur after 56 cm of RSLR, which could occur as early as 2060. Modelling showed that implementing changes to plant layout would allow the plant to remain on site for its design life (until 2080). Five adaptation archetypes were developed—sequences of adaptive actions that achieve the performance objective of continuing levels of service and avoid inundation of WWTPs. The seven-step approach is a way to stress-test a DAPP, to quantify signals, triggers and adaptation thresholds and to simulate implementation of a DAPP under a range of scenarios. This can facilitate more robust decision-making for wastewater infrastructure assets under future uncertainty.

Robust Decision-Making for the Reactive Collision Avoidance of Autonomous Ships against Various Perception Sensor Noise Levels

Autonomous ships are expected to extensively rely on perception sensors for situation awareness and safety during challenging operations, such as reactive collision avoidance. However, sensor noise is inevitable and its impact on end-to-end decision-making has not been addressed yet. This study aims to develop a methodology to enhance the robustness of decision-making for the reactive collision avoidance of autonomous ships against various perception sensor noise levels. A Gaussian-based noisy perception sensor is employed, where its noisy measurements and noise variance are incorporated into the decision-making as observations. A deep reinforcement learning agent is employed, which is trained in different noise variances. Robustness metrics that quantify the robustness of the agent’s decision-making are defined. A case study of a container ship using a LIDAR in a single static obstacle environment is investigated. Simulation results indicate sophisticated decision-making of the trained agent prioritising safety over efficiency when the noise variance is higher by conducting larger evasive manoeuvres. Sensitivity analysis indicates the criticality of the noise variance observation on the agent’s decision-making. Robustness is verified against noise variance up to 132% from its maximum trained value. Robustness is verified only up to 76% when the agent is trained without the noise variance observation with lack of its prior sophisticated decision-making. This study contributes towards the development of autonomous systems that can make safe and robust decisions under uncertainty.

Fuzzy Einstein-based decision-making model for the evaluation of site selection criteria of floating photovoltaic system

The site selection step is one of the milestones required to ensure the success of a renewable energy project. The present study proposes a novel framework for the suitable site selection of floating photovoltaic (FPV) systems by applying a robust Multi-Criteria Decision-Making (MCDM) method. A comprehensive literature review was performed to identify the exclusion and decision criteria. A questionnaire was prepared, and 30 experts from academia, industry, and non-governmental organizations answered it. The Fuzzy Einstein-based Logarithmic Additive Estimation of Weight Coefficients (LAAW) model was applied to assess the expert estimates and to determine the weights of 4 main and 41 sub-criteria. Based on the fuzzy model results, the main criteria were ranked in the following order of importance: economical, technical, environmental, and social and regulations. Although the number of decision criteria and the MCDM method significantly differ, the results of the present study are consistent with the findings of previous studies in the literature.

Solving Non-rectangular Reward-Robust MDPs via Frequency Regularization

In robust Markov decision processes (RMDPs), it is assumed that the reward and the transition dynamics lie in a given uncertainty set. By targeting maximal return under the most adversarial model from that set, RMDPs address performance sensitivity to misspecified environments. Yet, to preserve computational tractability, the uncertainty set is traditionally independently structured for each state. This so-called rectangularity condition is solely motivated by computational concerns. As a result, it lacks a practical incentive and may lead to overly conservative behavior. In this work, we study coupled reward RMDPs where the transition kernel is fixed, but the reward function lies within an alpha-radius from a nominal one. We draw a direct connection between this type of non-rectangular reward-RMDPs and applying policy visitation frequency regularization. We introduce a policy-gradient method, and prove its convergence. Numerical experiments illustrate the learned policy's robustness and its less conservative behavior when compared to rectangular uncertainty.