Publications
Modelling And Control of Cable Driven Robotic Arm Using Maplesim
Sep 30, 2024Journal Advances in Electrical and Electronic Engineering
Publisher VSB-Technical University of Ostrava
DOI 10.15598/aeee.v22i3.5685
Issue 22
Volume 3
Cable-driven robotic arms (CDRA) are robots with novel structures, wherein flexible cables are used to drive rigid links identified to move the end effector according to a desired trajectory. Due to the complex and nonlinear characteristics of this type of robotic arm, it is challenging to derive the model, which requires critical analysis to be conducted. This paper presents the design, modeling, and Model Predictive Control (MPC) of a special 2D CDRA with four rigid links. Maplesim is employed as a tool to design and simulate the proposed robotic arm. First, the prototype model is constructed in Maplesim and simulated using random input signals, and the input and output data sets are collected. A data-driven scheme based on neural networks is used to learn the unknown kinematics of the CDRA and to solve the kinematic control issue. The Matlab-Simulink platform is used for this purpose, and the black box model is obtained using the neural network fitting tool. MPC is then used for the end effector trajectory tracking control and to validate the modeling processes. Furthermore, comparative simulations using two scenarios are applied to the controlled system to verify the effectiveness of the proposed modeling and control method with the aid of Mean Squared Error (MSE) as an optimality index. The result verified that CDRA is capable of following reference trajectories accurately with MSE of 10e-5 and 4.99e-5 for rhombus and circular trajectories respectively.
Real-Time End-to-End Self-Driving Car Navigation
Jan 27, 2023Journal International Journal of Intelligent Systems and Applications in Engineering
Publisher Ismail Saritas
DOI https://ijisae.org/index.php/IJISAE/article/view/2732
Issue 11
Volume 2s
In this study, a deep neural network (DNN)-based vision-based navigation for autonomous vehicles is proposed. This novel DNN-based system obtains the data from a single camera to provide vehicle control outputs that modify both the steering wheel angle and the vehicle’s velocity. In addition, it plays a major role in safely navigating the vehicle in a road traffic environment. Numerous autonomous driving algorithms use end-to-end DNN, where camera data is fed into complex machine learning algorithms exclusively to estimate the steering angle value, but this research proposes a light-novel network model that controls both steering and speed values with much more simplicity. Various neural blocks are organized with the ultimate objective of producing control actions to achieve the aim of the research. Experimental modifications are made to parameters such as the number of convolutional layers, the model size, padding, stride, and the number of neurons in fully-connected layers to make the model simpler and lighter to execute during inference. Using an embedded system called Jetson Nano 2GB, the designed model was trained and tested using the images collected along two different paths. Our DNN-based autonomous driving system successfully predicts speed and steering values with a mean error of 1.58% and maintains performance, allowing for highly efficient autonomous driving. Furthermore, the suggested DNN network maintains performance, attaining autonomous driving success with comparable efficacy to the other autonomous driving models. The lightweight end-to-end architecture with superb performance is especially suited for autonomous driving.
Comparative Transfer Learning Models for End-to-End Self-Driving Car
Dec 1, 2022Journal Al-Khwarizmi Engineering Journal
Publisher University of Baghdad
DOI https://doi.org/10.22153/kej.2022.09.003
Issue 18
Volume 4
elf-driving automobiles are prominent in science and technology, which affect social and economic development. Deep learning (DL) is the most common area of study in artificial intelligence (AI). In recent years, deep learning-based solutions have been presented in the field of self-driving cars and have achieved outstanding results. Different studies investigated a variety of significant technologies for autonomous vehicles, including car navigation systems, path planning, environmental perception, as well as car control. End-to-end learning control directly converts sensory data into control commands in autonomous driving. This research aims to identify the most accurate pre-trained Deep Neural Network (DNN) for predicting the steering angle of a self-driving vehicle that is suitable to be applied to embedded automotive technologies with limited performance. Three well-known pre-trained models were compared in this study: AlexNet, ResNet18, and DenseNet121. Transfer learning was utilized by modifying the final layer of pre-trained models in order to predict the steering angle of the vehicle. Experiments were conducted on the dataset collected from two different tracks. According to the study's findings, ResNet18 and DenseNet121 have the lowest error percentage for steering angle values. Furthermore, the performance of the modified models was evaluated on predetermined tracks. ResNet18 outperformed DenseNet121 in terms of accuracy, with less deviation from the center of the track, while DenseNet121 demonstrated greater adaptability across multiple tracks, resulting in better performance consistency.