Search

This study aims to find the best artificial neural network weight values to be applied to the autonomous vehicle system with ultrasonic multisensor. The implementation of neural network in the system required long time process due to its training process. Therefore, this research is using offline training before implementing to online training by embedding the best network weight values to obtain the outputs faster according to desired targets. Simulink were used to train the system offline. Eight ultrasonic sensors are used on all sides of the vehicle and arranged in a serial multisensory configuration as inputs of neural network. With eight inputs, one sixteen-depth hidden layer, and five outputs, it was trained using the back-propagation algorithm of artificial neural network. By 100000 iterations, the output values and the target values are almost the same, indicating its convergency with minimum of errors. The result of this training is the best weights of the networks. These weight values can be implemented as fixed-weight in online training. [ABSTRACT FROM AUTHOR]

A good dataset was required for attaining good accuracy in machine learning, especially in prediction, so that prediction accuracy was high. The imbalanced or too small dataset was a common problem in machine learning. This study proposed a method for determining the dataset's quality. If the dataset is not feasible, preprocessing can be performed to improve the dataset's quality before making predictions. Adaptive Least Mean Square (LMS) was merged with Min-max Normalization and Fuzzy Intuitive Sets (FIS) algorithms to create the proposed technique. This method might assess the value of uncertainty and information, which will influence the dataset's feasibility. If the dataset has an uncertainty value closed 1.5 and an information value of less than 0.5, it is usable. The method has been tested on both public and private datasets. According to all experiments conducted, the uncertainty value and information value on the stated threshold can have prediction accuracy above 70% with various methodologies. [ABSTRACT FROM AUTHOR]

The steering system in autonomous vehicles is an essential issue that must be addressed. Appropriate control will result in a smooth and risk-free steering system. Compared to other types of controls, type-2 fuzzy logic control has the advantage of dealing with uncertain inputs, which are common in autonomous vehicles. This paper proposes a novel method for the steering control of autonomous vehicles based on type-2 fuzzy logic control combined with PI control. The primary control, type-2 fuzzy logic control, has three inputs—distance, navigation, and speed. The fuzzy system's output is the steering angle value. This was used as input for the secondary control, PI control. This control is in charge of adjusting the motor's position as a manifestation of the steering angle. The study results applied to the EPS system of autonomous vehicles revealed that type-2 fuzzy logic control and PI control produced better and smoother control than type-1 fuzzy logic control and PI. The slightest disturbance in the type-1 fuzzy logic control showed a significant change in steering, while this did not occur in the type-2 fuzzy logic control. The results indicate that type-2 fuzzy logic control and PI control could be used for autonomous vehicles by maintaining the comfort and safety of the users. [ABSTRACT FROM AUTHOR]

Noise is a problem often found in daily life. Noise also make people could not concentrate to do their work. Efforts to reduce noise have been proposed, but, due to variety of the noise's characteristics, every noise problem requires different solution. This research aim to cancel the vehicle's noise while maintaining the information heard. These conditions happened in the hospitals classrooms, or work room near the roadway. The vehicle's noise changes very fast, so the adaptive system is the good solution candidate for solving this problem. On the beginning, the simulation process had the trouble with the iterations. Matlab software only can execute the certain range of iteration. It could not cancel the noise, even the information becomes criptic. The problem is how to cancell the vehicle's noise with the restriction software and still manage the important information. This research will modify the LMS adaptive algorithm so that the iteration could be done by the system and the main goal of the research could be reached. The modification of the algorithm is based on the filter length (L) used to adapt with the noise. Therefore, this research conducted simulation of the adaptive noise cancelling with two process steps. The output of the first adaptive process have the.same characteristics with the noise that would be cancelled, thus the first adaptive process have the error near to zero. The second adaptive process changes the input by the output of the first process and mix the information into the noise. Error occured in the final process is the information heard as the dominant output. [ABSTRACT FROM AUTHOR]