Abstract:

An approach to driving decisions (DDS) A concept for driving decision-making for an autonomous car is based on machine learning, and it uses internal vehicle data, like steering and RPM level, to forecast different types of behaviour, like speed (steering), changing lanes, etc. All methods at the time were designed to focus on exterior data, such as the state of the roads and the number of people, but not on internal variables. So, the author is analysing internal data to determine the steering state and lane changes effectively. All internal data will be gathered from sensors, saved on the cloud, read by the application, and then subjected to ML algorithms to ascertain steering angle or changing lanes. The DDS algorithm is based on a genetic algorithm to select the ideal gene values that aid in making better decisions or predictions, and is used to implement this. The genetic algorithm was used by the DDS algorithm to select the ideal value, enabling quicker and more accurate prediction. Performance of the proposed DDS with genetic algorithm is compared to that of currently used machine learning techniques like Random Forest and MLP (multilayer perceptron algorithm.). Compared to random forest and MLP, the proposed DDS displays higher prediction accuracy.

Description:

RF, MLP, DDS, genetic algorithm