TO PURCHASE OUR PROJECTS IN ONLINE (OR) OFFLINE
CONTACT:VENKAT INNOVATIVE PROJECTS
NAME: VENKATARAO GANIPISETTY
Mobile & WhatsApp :+91 9966499110
Mobile & WhatsApp :+91 9573201550
Email :venkatjavaprojects@gmail.com
Email :venkatinnovativeprojects@gmail.com
Website:venkatprojects.com
website:venkatinnovativeprojects.com/
About Project:
Now modern world enter into the world of self driving vehicles where tiny sensors will sense environment (road condition) or driving behaviour data and then input those data to Artificial Intelligence model which will predict vehicles next motion or will predict driver behaviour.
In past many existing algorithm was introduced such as Question Based Driver behaviour identification where software will ask question to driver for behaviour recognition, image based behaviour identification which will heavily dependent on image lighting and many more but those techniques were not accurate enough.
Later machine learning models was introduced which will take input from sensors and then predict driver behaviour but those techniques prediction accuracy is good but can be further enhance. So author of this paper employing combination of deep learning algorithms called CNN and LSTM where CNN will be used to extract features from dataset and then this extracted features will be input to LSTM to train time series based temporal data and this trained model can be used to classify driving style of the driver.
In propose work author defining two layers of CNN for features extraction and one layer of LSTM for temporal based driver style classification. In propose work author has trained another algorithm called CNN + LSTM + I by inducing sensor signals (where I represents inducing of signals) but we don’t have such signal so we cannot implement this model but we have implemented existing CNN and propose CNN + LSTM.
To train above models author has used simulated and real world driving dataset but not publish this dataset so we have used Driver Style Classification dataset from KAGGLE and below are the dataset details
dataset screen first row represents dataset column names and remaining rows represents dataset values and this dataset has 3 different class labels called ‘Normal, Aggressive and Conservative’. So by using this dataset we will train and test all algorithms.
Extension Concept
In propose work to manage temporal time series dataset we have another advanced algorithm called BI-Directional LSTM but in propose work author has not used this model so as extension we have added this model with LSTM layer and this model help us in further improving thus this model is called as CNN + LSTM + BI-LSTM.