Uncertainty Handling

This study introduces an advanced convolutional neural network (CNN)
model tailored for building occupancy detection that accommodates the
inherent uncertainties and contradictory information often encountered in
sensor data. By integrating neutrosophic layers into the CNN architecture,
we enable the model to effectively handle indeterminacy, vagueness, and
inconsistency present in real-world sensor readings. The approach employs
neutrosophic convolutional, max-pooling, and logic layers, providing a
comprehensive framework for feature extraction and decision-making.
Through a structured methodology encompassing data preprocessing, model initialization, training, evaluation, and optimization, we demonstrate the efficacy of the proposed model in accurately detecting occupancy status within residential environments. This enhanced CNN model offers improved accuracy, robustness, and interpretability, thereby facilitating its integration into smart building management systems and building automation applications, to enhance efficiency, comfort, and energy savings.

The establishment of fuzzy relations and the fuzzification of time series are the top priorities of the model for predicting fuzzy time series. A lot of literature studied these two aspects to ameliorate the capability of the forecasting model. In this paper, we proposed a new method(FTSOAX) to forecast fuzzy time series derived from the improved seagull optimization algorithm(ISOA) and XGBoost. For increasing the accurateness of the forecasting model in fuzzy time series, ISOA is applied to partition the domain of discourse to get more suitable intervals. We improved the seagull optimization algorithm(SOA) with the help of the Powell algorithm and a random curve action to make SOA have better convergence ability. Using XGBoost to forecast the change of fuzzy membership in order to overcome the disadvantage that fuzzy relation leads to low accuracy. We obtained daily confirmed COVID-19 cases in 7 countries as a dataset to demonstrate the performance of FTSOAX. The results show that FTSOAX is superior to other fuzzy forecasting models in the application of prediction of COVID-19 daily confirmed cases.

Programming by demonstration techniques facilitate the programming of robots. Some of them allow the generalization of tasks through parameters, although they require new training when trajectories different from the ones used to estimate the model need to be added. One of the ways to retrain a robot is by incremental learning, which supplies additional information of the task and does not require teaching the whole task again. The present study proposes three techniques to add trajectories to a previously estimated task-parameterized Gaussian mixture model. The first technique estimates a new model by accumulating the new trajectory and the set of trajectories generated using the previous model. The second technique permits adding to the parameters of the existent model those obtained for the new trajectories. The third one updates the model parameters by running a modified version of the Expectation-Maximization algorithm, with the information of the new trajectories. The techniques were evaluated in a simulated task and a real one, and they showed better performance than that of the existent model.

Reinforcement learning (RL) is a common paradigm for learning tasks in robotics. However, a lot of exploration is usually required, making RL too slow for high-level tasks. We present V-MIN, an algorithm that integrates teacher demonstrations with RL to learn complex tasks faster. The algorithm combines active demonstration requests and autonomous exploration to find policies yielding rewards higher than a given threshold Vmin. This threshold sets the degree of quality with which the robot is expected to complete the task, thus allowing the user to either opt for very good policies that require many learning experiences, or to be more permissive with sub-optimal policies that are easier to learn. The threshold can also be increased online to force the system to improve its policies until the desired behavior is obtained. Furthermore, the algorithm generalizes previously learned knowledge, adapting well to changes. The performance of V-MIN has been validated through experimentation, ...

This paper presents a new approach to plan high-level manipulation actions for cleaning surfaces in household environments, like removing dirt from a table using a rag. Dragging actions can change the distribution of dirt in an unpredictable manner, and thus the planning becomes challenging. We propose to define the problem using explicitly uncertain actions, and then plan the most effective sequence of actions in terms of time. However, some issues have to be tackled to plan efficiently with stochastic actions. States become hard to predict after executing a few actions, so replanning every few actions with newer perceptions gives the best results, and the trade-off between planning time and plan quality is also important. Finally a learner is integrated to provide adaptation to changes, such as different rag grasps, robots, or cleaning surfaces. We demonstrate experimentally, using two different robot platforms, that planning is more advantageous than simple reactive strategies for accomplishing complex tasks, while still providing a similar performance for easy tasks. We also performed experiments where the rag grasp was changed, and thus the behaviour of the dragging actions, showing that the learning capabilities allow the robot to double its performance with a new rag grasp after a few cleaning iterations.

This paper presents a new approach to plan high-level manipulation actions for cleaning surfaces in household environments, like removing dirt from a table using a rag. Dragging actions can change the distribution of dirt in an unpredictable manner, and thus the planning becomes challenging. We propose to define the problem using explicitly uncertain actions, and then plan the most effective sequence of actions in terms of time. However, some issues have to be tackled to plan efficiently with stochastic actions. States become hard to predict after executing a few actions, so replanning every few actions with newer perceptions gives the best results, and the trade-off between planning time and plan quality is also important. Finally a learner is integrated to provide adaptation to changes, such as different rag grasps, robots, or cleaning surfaces. We demonstrate experimentally, using two different robot platforms, that planning is more advantageous than simple reactive strategies for accomplishing complex tasks, while still providing a similar performance for easy tasks. We also performed experiments where the rag grasp was changed, and thus the behaviour of the dragging actions, showing that the learning capabilities allow the robot to double its performance with a new rag grasp after a few cleaning iterations.

This paper presents a new approach to plan high-level manipulation actions for cleaning surfaces in household environments, like removing dirt from a table using a rag. Dragging actions can change the distribution of dirt in an unpredictable manner, and thus the planning becomes challenging. We propose to define the problem using explicitly uncertain actions, and then plan the most effective sequence of actions in terms of time. However, some issues have to be tackled to plan efficiently with stochastic actions. States become hard to predict after executing a few actions, so replanning every few actions with newer perceptions gives the best results, and the trade-off between planning time and plan quality is also important. Finally a learner is integrated to provide adaptation to changes, such as different rag grasps, robots, or cleaning surfaces. We demonstrate experimentally, using two different robot platforms, that planning is more advantageous than simple reactive strategies for accomplishing complex tasks, while still providing a similar performance for easy tasks. We also performed experiments where the rag grasp was changed, and thus the behaviour of the dragging actions, showing that the learning capabilities allow the robot to double its performance with a new rag grasp after a few cleaning iterations.

Transactive energy as an emerging approach and sustainable technology can provide an exceptional opportunity for microgrids to exchange energy with each other for greater benefits in the cluster mode. In this mode of operation, some collective and individual interests can be realized for the microgrids based on transactive energy management. This paper proposes mathematical models for microgrid clusters using a transactive energy structure to manage energy exchange in the smart grid. In order to make an informed decision for the operation of microgrid clusters, chance-constrained programming is employed to consider the uncertainties in balancing collective and individual interests under the transactive energy management. In this research, sixteen commercial microgrids are considered in the process of evaluating the efficiency of the proposed models using the chance-constrained programming method. Simulation results prove the effectiveness of the transactive energy approach accompanying the implementation of chance-constrained programming in energy management of the microgrid clusters.

We describe a system allowing a robot to learn goal-directed manipulation sequences such as steps of an assembly task. Learning is based on a free mix of exploration and instruction by an external teacher, and may be active in the sense that the system tests actions to maximize learning progress and asks the teacher if needed. The main component is a symbolic planning engine that operates on learned rules, defined by actions and their pre-and postconditions. Learned by model-based reinforcement learning, rules are immediately available for planning. Thus, there are no distinct learning and application phases. We show how dynamic plans, replanned after every action if necessary, can be used for automatic execution of manipulation sequences, for monitoring of observed manipulation sequences, or a mix of the two, all while extending and refining the rule base on the fly. Quantitative results indicate fast convergence using few training examples, and highly effective teacher intervention at early stages of learning.

Accurate and reliable estimation of groundwater level is important for the development and management of water resources. In this study, models of adaptive neuro-fuzzy inference system (ANFIS) with multiple linear regression (MLR) method and its performance in predicting groundwater level were investigated. As a field of application, it was applied for General Directorate of State Hydraulic Works (DSİ) 5512 well of Dörtyol region of Hatay province. In the study, 147-month data sets between 2000 and 2015, including hydrological parameters such as Precipitation (P), average air temperature (T), relative humidity (RH), wind speed (W) groundwater level (GWL) time series, predict the groundwater level used. The determinant coefficient (R 2), mean square error (MSE) and mean absolute error (MAE) were used as the statistical performance evaluation criteria. As a result of this study, MLR and ANFIS models performed well for GWL estimation. In particular, the ANFIS model yielded better results than the MLR model.

Rainfall-Runoff relationship analyzes are essential for the protection of flood rooting, management of water resources and design of water structures. In this study, Neuro-Fuzzy (NF) and Support Vector Machines (SVM) methods are applied for Rainfall-Runoff prediction. Daily hydrological and seasonal data taken from Muskegon basin in USA were used for present study. 1397 daily data of rainfall, temperature and runoff from the study area were analyzed by NF and SVM methods. The results show that the SVM method lead to low errors and high determinations in the Rainfall-Runoff modeling. Models results are compared with daily observed data. SVM method can be used as an alternative to classical methods in Rainfall-Runoff prediction.

Monitoring human activities provides context data to be used by computational systems, aiming a better understanding of users and their surroundings. Uncertainty still is an obstacle to overcome when dealing with context-aware systems. The origin of it may be related to incomplete or outdated data. Attribute Grammars emerge as a consistent approach to deal with this problem due to their formal nature, allowing the definition of rules to validate context. In this paper, a model to validate human daily activities based on an Attribute Grammar is presented. Context data is analysed through the execution of rules that implement semantic statements. This processing, called semantic analysis, will highlight problems that can be raised up by uncertain situations. The main contribution of this paper is the proposal of a rigorous approach to deal with context-aware decisions (decisions that depend on the data collected from the sensors in the environment) in such a way that uncertainty can b...

We describe a system allowing a robot to learn goal-directed manipulation sequences such as steps of an assembly task. Learning is based on a free mix of exploration and instruction by an external teacher, and may be active in the sense that the system tests actions to maximize learning progress and asks the teacher if needed. The main component is a symbolic planning engine that operates on learned rules, defined by actions and their pre-and postconditions. Learned by model-based reinforcement learning, rules are immediately available for planning. Thus, there are no distinct learning and application phases. We show how dynamic plans, replanned after every action if necessary, can be used for automatic execution of manipulation sequences, for monitoring of observed manipulation sequences, or a mix of the two, all while extending and refining the rule base on the fly. Quantitative results indicate fast convergence using few training examples, and highly effective teacher intervention at early stages of learning.

We describe a system allowing a robot to learn goal-directed manipulation sequences such as steps of an assembly task. Learning is based on a free mix of exploration and instruction by an external teacher, and may be active in the sense that the system tests actions to maximize learning progress and asks the teacher if needed. The main component is a symbolic planning engine that operates on learned rules, defined by actions and their pre-and postconditions. Learned by model-based reinforcement learning, rules are immediately available for planning. Thus, there are no distinct learning and application phases. We show how dynamic plans, replanned after every action if necessary, can be used for automatic execution of manipulation sequences, for monitoring of observed manipulation sequences, or a mix of the two, all while extending and refining the rule base on the fly. Quantitative results indicate fast convergence using few training examples, and highly effective teacher intervention at early stages of learning.

We describe a system allowing a robot to learn goal-directed manipulation sequences such as steps of an assembly task. Learning is based on a free mix of exploration and instruction by an external teacher, and may be active in the sense that the system tests actions to maximize learning progress and asks the teacher if needed. The main component is a symbolic planning engine that operates on learned rules, defined by actions and their pre-and postconditions. Learned by model-based reinforcement learning, rules are immediately available for planning. Thus, there are no distinct learning and application phases. We show how dynamic plans, replanned after every action if necessary, can be used for automatic execution of manipulation sequences, for monitoring of observed manipulation sequences, or a mix of the two, all while extending and refining the rule base on the fly. Quantitative results indicate fast convergence using few training examples, and highly effective teacher intervention at early stages of learning.

Accurate multi-time scale prediction of groundwater level (GWL) is important for water resources planning and management. But it is difficult to achieve reliable and robust GWL predictive even by the use of soft-computing techniques which considers the uncontrollable error, indefinite input, and unneglectable uncertainty at the time of modeling process. Soft computing approaches integrated into the data pre-treatment approach, input selection model, or uncertainty analysis can be employed to resolve this issue. The design of extensive deterministic and uncertainty analysis of automated models for the prediction of GWL is yet to be extensively explored. In this aspect, this study focuses on the design of type-II fuzzy logic system with uncertainty handling (T2FLS-UH) for GWL prediction. The goal of the T2FLS-UH technique is for predicting the GWL with the consideration of uncertainty. The T2FLS-UH technique encompasses different stages of operations such as data preparation, feature selection, prediction, and membership function selection. Besides, an ensemble empirical mode decomposition (EEMD) technique is involved for the decomposition of the original signals as to various intrinsic mode functions (IMFs). In addition, the Boruta approach is used for the selection of appropriate input variables for the prediction process. Moreover, the T2FLS model is applied for the GWL prediction process to estimate the value for every IMF. To improvise the predictive performance, the membership function of the T2FLS technique is chosen optimally by seagull optimization (SGO) algorithm. A wide range of simulations was carried to highlight the enhanced predictive performance of the T2FLS-UH technique. The experimental values pointed out the supremacy of the T2FLS-UH technique over the recent state of art prediction techniques.

We have been engaged in research towards designing a software system called LOZ for learning the object oriented formal specification notation Object-Z. Initially, we conducted a survey to analyse the effectiveness of traditional methodologies for learning Z notation. In LOZ, the semiformal model UML is used in the intermediate phase between informal textual description and formal Object-Z description. We also employ a refinement unit that produces code from the specification. Based on the cognitive apprenticeship approach, we employ a four-phase instructional model in our system. Persuading the learners to be partially responsible for their own model and allowing them to decide their own levels of control over the learning process are key features of our system.

The student record (SR) is a major source of input for any decision making done by an Intelligent Tutoring System (ITS) and is a basis of the individualization in such systems. However, most ITSs still have "generalized" student models which represent a type of student rather than a particular one. Until the SR becomes truly representative of each individual student, the goal of providing individualized tutoring cannot be attained. In this paper we describe an Intelligent Fuzzy Temporal Relational Database (IFFReD), an intelligent system-independent SR which allows for almost any degree of individualization the designer wishes to incorporate. It is anticipated that this IFTReD will provide a significant improvement over standard AI storage techniques for the SR. These improvements will be realized in terms of: (1) intelligence; (2) greater storage efficiency; (3) greater speed in retrieval and query; (4) ability to handle linguistic codes, ranges, fuzzy possibilities, and incomplete data in student models; (5) friendliness of query language; (6) availability of temporal knowledge to give a history of past performance; and (7) a more holistic view of the student, permitting greater individualization of the tutor.

We have been engaged in research towards designing a software system called LOZ for learning the object oriented formal specification notation Object-Z. Initially, we conducted a survey to analyse the effectiveness of traditional methodologies for learning Z notation. In LOZ, the semiformal model UML is used in the intermediate phase between informal textual description and formal Object-Z description. We also employ a refinement unit that produces code from the specification. Based on the cognitive apprenticeship approach, we employ a four-phase instructional model in our system. Persuading the learners to be partially responsible for their own model and allowing them to decide their own levels of control over the learning process are key features of our system.