Deep Learning Implementation of Autonomous Driving using Ensemble-M in Simulated Environment

Abstract

Making Autonomous Driving a safe, feasible and better alternative is one the core problems the world is facing today. The horizon of the applications of AI and Deep Learning has changed the perspective of the human mind. Initially, what used to be thought as subtle impossible task is applicable today and that too in the feasibly efficient way. Computer vision tasks powered with highly tuned CNNs are outperforming humans in many fields. Introductory implementations of autonomous vehicle were merely achieved using raw image processing and hard programmed rule-based logic systems along with machine/deep Learning used as secondary objective handlers. With the autonomous driving method proposed by NVIDIA, the usability of CNNs is more adequate, adaptable and applicable. In this paper, we propose, the ensemble implementation of CNN based regression models for autonomous-driving. We have taken simulator generated driving view image dataset along with mapped file of steering angle in radians. After applying image pre-processing and augmentation, we have used two CNN models along with their ensemble and compared their performance as to minimize the risks of unsafe driving. We have compared Nvidia proposed CNN, MobileNet-V2 as regression model and Ensemble-M results for comparison their respective performance, MSE scores and compute time to process. In result analysis, the MobileNet-V2 model performs better in densely-featured roads and Nvidia model performs better in sparsely-featured roads whereas Ensemble-M normalizes the performance of both models and efficiently result in the least MSE score (0.0201) with highest computation time utilization.

Steering Angle Prediction for Autonomous Vehicles Using Deep Transfer Learning

“…Recently, many deep learning models have been born and applied to replace the basic CNN to increase the accuracy of the predicted angle. To minimize the mean square error between the true and predicted steering angles, Gupta et al [18] employed the MobileNetV2 model for lane-keeping. Du and Gao et al [19] utilized transfer learning for dataset learning and implemented convolution layer modeling within the Long Short-Term Memory (LSTM) recurrent layer to develop their model.…”

“…Gupta et al [18] The approach taken in the study is to utilize the MobileNetV2 model for lane-keeping, which involves predicting steering angles based on the input data. This approach aims to minimize the mean square error between the predicted and true steering angles.…”

Deep Learning-Based Lane-Keeping Assist System for Self-Driving Cars Using Transfer Learning and Fine Tuning

“…The LeNet network architecture, comprising activation functions, convolutional layers, and fully connected layers, was employed in other research papers [16,17] to maximize performance and achieve optimal results. Gupta et al [18] employed the MobileNetV2 model for lane-keeping, aiming to minimize the mean square error between the true and predicted steering angles. Babiker et al [19] conducted an experiment combining the VGG-19 model for classification with an end-to-end steering control system to simultaneously detect traffic signals and steering angles, achieving a signal accuracy score of 86%.…”

Performance of post phthalate Ziegler Natta catalysts with activity limiting agents for propylene gas phase polymerization

“…Several other patents disclose novel mixtures of external donors and processes where an ALA is used, including in the production of impact propylene copolymers [6] , in the production of high melt flow propylene polymers [7] , in the production of ethylene based polymers [8] , the discontinuous addition of thermal runaway control agents in commercial production of polypropylene [9] , the use of monoether compounds as ALAs [10] , utilizing fluorinated fatty acids as ALAs [11] , and an improved multi-zone olefins polymerization process operating in deep condensed modes employing several thermal runaway control agents [12] .…”