Journal of Modeling and Simulation
Volume:55 Issue: 1, Winter-Spring 2023

The purpose of automated negotiations, as a novel field of study in Artificial Intelligence, is focused on autonomous agents that can appear as humans' intelligent representatives, attend negotiations with other agents, and attain acceptable outcomes. The so-called automated negotiating agents are implemented such that they can beat as many opponents as possible in different kinds of domains. Like what happens in our daily negotiations, agents in automated negotiations do not reveal their preferences explicitly. Numerous research studies have heretofore accentuated that an opponent model would be a great salvation to reduce this uncertainty, since it can be of much assistance in making wiser decisions in the next steps, reaching ideal eventual utility, and more satisfaction, accordingly. Although most opponents in our world have single-peaked preferences, the functionality of negotiating agents in modeling single-peaked opponents has not been studied. Gaussian agents are one important sort of single-peaked agents that utilize the Gaussian function to ascribe the ranking of each negotiation item. The Gaussian opponent's bliss point estimation is of high importance during a negotiation. Therefore, we first proposed a variety of Gaussian bidding agents and then focused on how accurately Automated Negotiating Agents Competition (ANAC) attendees during 2010-2019 would model these bidder agents. The results of our experiments revealed that existing ANAC agents are performing well regarding individual utility and social welfare on average, but they are poor in modeling Gaussian negotiating bidding agents.

As the number of features in practical applications grows, the feature selection process becomes increasingly important. Since the feature selection problem is NP-hard, no exact algorithm can determine the optimal subset over a reasonable period of time. However, traditional feature selection methods are time-consuming and tend to get stuck in local optima. Unlike traditional search techniques, metaheuristics are more effective at exploring and exploiting the search domain because they use several operators. Besides these behaviors of metaheuristics, we present an improved Equilibrium Optimizer algorithm using the density of population and entropy operator, which has proven a good exploration ability to provide a promising candidate solution. After that, a new feature selection model is developed based on the improved Equilibrium Optimizer algorithm. K-nearest neighbors is used as an evaluator for the new solutions. In order to test the performance of the proposed algorithm, simulation experiments are conducted on a set of 14 standard test functions containing both unimodal and multimodal functions. To evaluate the effectiveness of the proposed algorithm, 15 UCI benchmark datasets and five metaheuristics, GA, CS, GSA, RDA, and BBA  are applied. The experimental results revealed the effectiveness of our approach in terms of accuracy performance for the feature selection process.

Translating face sketches to photo-realistic faces is an interesting and essential task in many applications like law enforcement and the digital entertainment industry. One of the most important challenges of this task is the inherent differences between the sketch and the real image such as the lack of color and details of the skin tissue in the sketch. With the advent of adversarial generative models, an increasing number of methods have been proposed for sketch-to-image synthesis. However, these models still suffer from limitations such as the large number of paired data required for training, the low resolution of the produced images, or the unrealistic appearance of the generated images. In this paper, we propose a method for converting an input facial sketch to a colorful photo without the need for any paired dataset. To do so, we use a pre-trained face photo generating model to synthesize high-quality natural face photos and employ an optimization procedure to keep high fidelity to the input sketch. We train a network to map the facial features extracted from the input sketch to a vector in the latent space of the face-generating model. Also, we study different optimization criteria and compare the results of the proposed model with those of the state-of-the-art models quantitatively and qualitatively. The proposed model achieved 0.655 in the SSIM index and 97.59% rank-1 face recognition rate with a higher quality of the produced images.

Image registration is a fundamental issue in medical image analysis. It refers to the matching process between two or more images using the optimization of a similarity metric to find an optimal transformation function. In recent decades, many studies have been done on the medical image registration topic. Therefore, this paper has investigated four main methodologies to solve the registration problem in medical applications. One of the most important topics in this area is the registration of multi-modal images. In this paper, we have reviewed various multimodal image registration techniques based on deep learning and proposed a classification for these methods. Also one of the essential components of the medical image registration framework is the similarity measure function. There are different similarity metrics in this area and choosing an appropriate measure according to the application is a challenging problem. This paper is to present a review of different similarity measures in medical applications and a classification of these methods. Based on this classification, techniques are investigated and each subclass is evaluated using performance criteria. Therefore, the main goals of this article are as follows: 1) Investigating the most significant image registration approaches. 2) Systematic review of deep learning-based multimodal medical image registration and classify them. 3) Providing classification for various similarity measure techniques according to registration applications. 4) Creating an appropriate platform for evaluating these approaches and introducing the main challenges.

The issue of position/force control of collaborative robotic systems moving a payload is proposed in this paper. The proposed approach must be able to maintain the orientation/position of the payload on the reference trajectory while applying a limited force to the object through the robot's end-effector. With this in mind, linear/nonlinear PID control schemes have been proposed to achieve accurate and robust tracking performance. Lyapunov's stability analysis is utilized to confirm the stability of the controlled system. It proves that the controlled system is stable, while the object’s orientation/position tracking errors are uniformly ultimately bounded (UUB) in any bounded region of state space. It also presents some conditions for proper selection of the linear/nonlinear PID controllers’ gains in the form of two theorems. The proposed controllers apply to two coordinated 3DOF robotic arms that carry a payload. The simulation results tested two types of trajectories, including simple and complex paths. The results are also compared to those of a strong state-of-the-art approximator, the Chebyshev Neural Network (CNN).

In recent years, Convolutional Neural Networks (CNNs) have made significant strides in the field of segmentation, particularly in semantic segmentation where both accuracy and efficiency are crucial. However, despite their high accuracy, these deep networks are not practical for real-time use due to their low inference speed. This issue has prompted researchers to explore various techniques to improve the efficiency of CNNs. One such technique is knowledge distillation, which involves transferring knowledge from a larger, cumbersome (teacher) model to a smaller, more compact (student) model. This paper proposes a simple yet efficient approach to address the issue of low inference speed in CNNs using knowledge distillation. The proposed method involves distilling knowledge from the feature maps of the teacher model to guide the learning of the student model. The approach uses a straightforward technique known as pixel-wise distillation to transfer the feature maps of the last convolution layer of the teacher model to the student model. Additionally, a pair-wise distillation technique is used to transfer pair-wise similarities of the intermediate layers. To validate the effectiveness of the proposed method, extensive experiments were conducted on the PascalVoc 2012 dataset using a state-of-the-art DeepLabV3+ segmentation network with different backbone architectures. The results showed that the proposed method achieved a balanced mean Intersection over Union (mIoU) and training time.

Human activity recognition system (HARS) is critical for monitoring individuals' health and movements, especially the elderly and the disabled, in different environments. Wearable sensors are valuable tools for information acquisition due to the mobility of people. Researchers have proposed different methods for sensor-based HAR, which face challenges. This paper uses deep learning (DL), which combines the deep convolution neural network (CNN) and long short-term memory (DeepConvLSTM) to extract and select features to recognize high-performance activity. Then, the Softmax-Support Vector Machine (SofSVM) performs the classification and recognition operations. This paper uses a comprehensive nested pipe and filter (NPF) architecture. Filters are usually independent, but some filters can be bidirectional to improve the performance of complex activity recognition. There is two-way communication between convolutional layers and long short-term memory (LSTM). The Opportunity dataset is public and includes complex activities. The results with this dataset display that the proposed work improves the weighting F-measure of HAR. This paper has also shown other experiments; that involve comparing the number of sensors, max-pooling size, and convolutional filter size. According to this dataset, the proposed method weighting F-measure is 0.929. The proposed approach is indeed effective in activity recognition, and the NPF architecture covers all the components of the activity recognition process.

The complexity and dynamic order of large-scale systems is continuously increasing. Considering the many challenges that exist for these systems, it is very important to provide a robust distributed controller that performs well against uncertainties, computation volume, and interaction between subsystems. A robust-distributed ℒasso-MPC (RD-LMPC) approach is suggested in this study for multi-robot systems in the presence of polytopic uncertainty. In addition, a distributed Kalman filter is used to capture interactions between subsystems. To evaluate and perform the effectiveness of the suggested approach, the results obtained on the multi-robot system are compared with the results of the predictive control methods of the centralized, distributed model, and L1 adaptive control}.

In this article, we manage the energy consumption of numerous intelligent homes and charging stations including several electric vehicles in real-time using a computationally efficient predictive controller. The studied scenario is made up of a couple of primary layers. At the low level of the hierarchical framework, users are clustered into different groups based on their vehicles’ departure times. Meanwhile, the energy consumption of subordinate users is controlled by multiple aggregators, interaction among which is modeled as an aggregative game. The high-level interactive and distributed control problem can be solved by a predictive controller, wherein the terminal constraints related to the reference energy of each cluster’s storage capacity are transferred to the end of the prediction horizon. Additionally, each aggregator can only exchange local data with some neighboring aggregators through an untrustable communication network. As a result of denial of service attacks on the aggregators' network, the strong connectivity of the communication graph may be directly destroyed, leading to performance degradation. To address such an issue, each aggregator reconstructs the attacked information of its neighbors using a linear combination of received data in the last two iterations. Furthermore, a time-of-use pricing tariff whose value is small for faithful households is investigated so that convergence time remains unaltered. Practical examples are simulated to assess the usefulness of the proposed iterative algorithm.

This paper presents a Track-Before-Detect (TBD) approach to search and localize a radio-emitted target in a wide marine environment using just the received signal strength (RSS) measurements. In this problem, the lost target transmits radio signals, and the unmanned aerial vehicle (UAV), guided on a search path, receives the target transmitted signal strength by its mounted antenna. The guidance law directs the UAV to the best detection points where the probability of target detection is maximum. At the same time, the estimation module evolves the posterior distribution of the radio target states, including the target position, heading, and transmitter power. The best detection points are calculated based on this evolved target’s states’ posterior. The superiority of the proposed method is due to the consideration of the antenna radiation pattern, which is accurately modeled in this paper and ensures the strength of the filter against the uncertainties of the measurement model and the target model. The simulation results validate the performance of the proposed method in the autonomous localization of a lost moving target.

This paper develops a neural network-based fault tolerant LQR control for orbital maneuvering of satellites in low altitude orbit that are subjected to disturbances like earth gravity, atmospheric drag, third body, and solar radiation. The controller is also developed such that it tolerates the effects of additive and effective loss of actuator faults. A neural network is used in the controller to estimate disturbances and faults and decrease their effects. Due to the high efficiency of electric thrusters, they are used widely in LEO (Low Earth Orbit) missions. Therefore, in this paper Hall effect thrusters are used as the actuator. For the maneuvering purpose, the reference orbit parameters are derived from a reference orbital dynamics which is the Kepler dynamics subjected to only the disturbance of gravity. To avoid singularity, the orbital dynamics of six modified elements are used in the control design besides the six classical elements. Then, by the desired orbital parameters from the reference orbit, the relative motion elements are calculated to apply in control laws. By Lyapunov analysis, the updating laws of the weights of the control and neural network are derived and also ultimately boundedness of the error between the nominal orbital elements and the faulty orbital elements is proved. To show the effectiveness of the proposed control, it is applied in the nonlinear Kepler dynamics which is affected by an accurate model of natural disturbances.

One of the main issues in inertial navigation systems is attitude determination, which means estimating the level angles (i.e., roll and pitch). This paper investigates the attitude estimation problem for an accelerated rigid body using three gyros and three accelerometers. The most critical challenges in attitude determination systems are external accelerations and gyroscope drift errors. Thus, a novel method based on the adaptive filter-Kalman algorithm is proposed to estimate and compensate for these errors. Linearization was performed around a general work point, and the covariance matrix's adaptive values were obtained so that leveling angles were accurately determined despite external accelerations. The simulation results, along with the car test, which was performed in different dynamic conditions with external accelerations, showed that the introduced algorithm has a high capability in accurately estimating leveling angles. This approach can be used for GPS-less navigation Algorithms.