This research introduces an innovative approach for End-to-End steering angle prediction and its control in electric power steering (EPS) systems. The methodology integrates transfer learning-based computer vision techniques for prediction and control with fuzzy signatures-enhanced fuzzy systems. Fuzzy signatures are unique multidimensional data structures that represent data symbolically. This enhancement enables the fuzzy systems to effectively manage the inherent imprecision and uncertainty in various driving scenarios. The ultimate goal of this work is to assess the efficiency and performance of this combined approach by highlighting the pivotal role of steering angle prediction and control in the field of autonomous driving systems. Specifically, within EPS systems, the control of the motor directly influences the vehicle’s path and maneuverability. A significant breakthrough of this study is the successful application of transfer learning-based computer vision techniques to extract respective visual data without the need for large datasets. This represents an advancement in reducing the extensive data collection and computational load typically required. The findings of this research reveal the potential of this approach within EPS systems, with an MSE score of 0.0386 against 0.0476, by outperforming the existing NVIDIA model. This result provides a 22.63% better Mean Squared Error (MSE) score than NVIDIA’s model. The proposed model also showed better performance compared with all other three references found in the literature. Furthermore, we identify potential areas for refinement, such as decreasing model loss and simplifying the complex decision model of fuzzy systems, which can represent the symmetry and asymmetry of human decision-making systems. This study, therefore, contributes significantly to the ongoing evolution of autonomous driving systems.

Developing and constantly changing technologies, efforts to achieve maximum efficiency with minimum fuel consumption, as well as the development of comfort and safety systems, have become very essential topic in car manufacturing and design. Whereas comfort and security were not given a high importance in the first produced cars, they are indispensable elements of today's automobiles. Since public transportation uses road in large scale, the need for safety and repose is also increasing. Nowadays, vehicles have better security and comfort systems, which react very quickly to all kinds of loads and different cases of driving (braking, acceleration, high speed, cornering), where the tires can keep the road at its best, utilizing an advanced suspension system. In this study, a quarter-car model was fulfilled using LabVIEW (Laboratory Virtual Instrumentation Engineering Workbench) software. The control of this model has been realized by applying two different controllers. PID (proportional, integral, derivative) controller which is a common and conventional control method and the Fuzzy Logic controller which is considered as an expert system that is becoming more and more widely used. In both control approaches, controlling the suspension system was achieved successfully. However; It has been determined that controlling the system using Fuzzy Logic controller gave better dynamic response than applying the PID controller for the quarter car suspension model that has been used in the direction of this study.

The field of autonomous driving has seen remarkable progress in recent years, with the prediction of steering angles based on varying road conditions emerging as a critical area of focus. While previous efforts have concentrated on lane detection, road object identification, and 3-D reconstruction, our research centers on a vision-based model that leverages deep networks to translate raw camera images into steering angles without the need for predefined feature learning. In our paper, we introduce an end-to-end model that employs deep transfer learning to predict steering angles from image sequences captured by onboard cameras. This model merges two deep learning architectures: a convolutional neural network (CNN) and a long short-term memory (LSTM) network. We utilize the VGG16 network, pre-trained on ImageNet and renowned for its performance, to extract spatial features from the images. Concurrently, the LSTM network processes the temporal information embedded within the image sequences. Our proposed model comprehensively processes spatial-temporal data and adeptly models the nonlinear relationship between the input images and the steering angles. We conducted an experimental study using a publicly available dataset to evaluate the model's effectiveness. The outcomes of our experimental analysis reveal that our model delivers highly efficient and accurate steering angle predictions, effectively emulating human driving patterns. Moreover, it surpasses current models in both performance and training efficiency, achieving a Mean Squared Error (MSE) score of 0.0728 as its loss function.

This study introduces a hybrid sensor fusion approach that integrates Schreiber’s nonlinear filter with traditional filtering methods to enhance the performance of IMU-based systems in autonomous vehicles. As autonomous vehicles grow more dependent on Inertial Measurement Unit (IMU) data for real-time stability and control, the need for resilient and accurate sensor fusion becomes critical. This research addresses that need by introducing a method capable of maintaining robustness under highly dynamic and uncertain conditions. Accelerometer and gyroscope data from an IMU are first fused using a complementary filter. The fused signals are then refined by phase-space reconstruction and local manifold projection, improving noise resilience and maintaining system dynamics. Two datasets are used to assess the methodology: one was collected indoors with a smartphone, and another was captured outdoors using a Bosch sensor in various environmental settings. The proposed method demonstrates superior noise reduction, greater resistance to outliers, and improved signal consistency compared to conventional complementary and Kalman filters. The findings demonstrate how chaos-based nonlinear filtering may improve the reliability of sensor fusion on a variety of sensing platforms in highly dynamic environments. Given the importance of IMU data for maintaining vehicle stability, this study seeks to support the development of more stable autonomous transportation systems.

This paper introduces a pioneering approach to robot environment representation by integrating a novel hypergraph-based method for modeling fuzzy signatures with a quadtree-like structure for obstacle detection. By structuring fuzzy signatures through hypergraphs, we establish a robust framework that not only streamlines information representation but also simplifies the aggregation-based decision-making process. This synergy is applied to the domain of mobile robotics, where accurate and efficient environment representation is essential. Utilizing a quadtree-like structure for data organization, our technique systematically evaluates feature points against a set of fuzzy operations, determining the significance of obstacles and reconstructing the environmental model through the traversal of the quadtree-like structure. Furthermore, the hypergraph-based formalism sets the stage for a future transition to a tensor-based representation of fuzzy signatures, as envisioned in future work.

This study investigates the impact of different aggregation operators (commonly referred to as fuzzy operators) on the application of fuzzy signatures. Fuzzy signatures are specialized multidimensional data structures that symbolically represent data. As a use case, the study focuses on robot environment representation and path planning, presenting the results obtained by applying various aggregation operators including $minimum$, $maximum$, $algebraic$ $product$ and $arithmetic$ $mean$ on the normalized values obtained from the robot sensors. The comparison highlights their effects on the computational load and path lengths of the path planning task. The findings reveal that the most efficient aggregation operator, in terms of computational load and the path length, is the $algebraic$ $product$ aggregation operator. Specifically, the $algebraic$ $product$ consistently yielded the shortest paths (as low as 22 nodes) and the lowest execution times (down to 0.0913 s), demonstrating superior efficiency compared to the $maximum$ operator, which resulted in path lengths up to 34 nodes and execution times reaching 0.1923 s. This represents an improvement of up to 35.3% reduction in path length and 52.5% reduction in execution time when comparing the $algebraic$ $product$ to the $maximum$ operator based on observed extreme values. Furthermore, this work provides the first empirical comparison of fuzzy aggregation operators specifically for fuzzy signature-based mobile robot path planning.