Review on Artificial Intelligence-Based Sustainable System for Autonomous Vehicles

Introduction

In the late 19th century, with the advent of electricity brought about by the Second Industrial Revolution, the first electric cars came into being. Because of the advantages it offered over gasoline-powered cars in terms of noise reduction, passenger comfort, and ease of operation, electric vehicles were a popular choice for motor vehicle propulsion in the early 20th century. However, concerns about having insufficient Energy stored in the batteries used at the time prevented widespread use of these vehicles. In the late 90s, hybrid electric cars, which combine internal combustion engines with electric motors, started to gain popularity. Mass manufacturing of plug-in hybrid electric vehicles did not begin until the late 2000s, and battery electric cars were not commercially viable choices for buyers until the 2010s.

Electric motors are used as the primary propulsion source in these vehicles rather than an afterthought. Vehicle sales increased in the 2010s thanks to government incentives implemented in 1990 in Norway and later in the 2000s in other big markets, such as the US and EU. The electric vehicle industry is anticipated to significantly increase due to rising public interest, awareness, and structural incentives, such as those included in the green recovery efforts after the COVID-19 epidemic. As a result of quarantines imposed during the COVID-19 epidemic, less greenhouse gases were released by diesel and gasoline automobiles. Policies about large electric cars are among the many things the International Energy Agency has called on nations to undertake to achieve climate targets. In 2022, electric vehicles accounted for 14% of all new vehicle sales, up from 9% in 2020 and 5% in 2021. The percentage of sales of electric cars may rise from 1% in 2016 to over 35% by 2030. The worldwide electric vehicle industry was valued at $280 billion in July 2022 and was projected to reach $1 trillion by 2026. American, European, and Chinese markets are anticipated to account for a significant portion of this expansion. Electric two-wheelers and three-wheelers are expected to increase in emerging nations, according to a literature study in 2020. However, four-wheeled electric vehicles seem economically unlikely to have such growth. More than 20% of all EVs are two- or three-wheelers, making them the most electrified road transport category now and in the future.

The fuel and technology used to generate Energy impact the carbon footprint and other emissions of electric cars. A battery, flywheel, or supercapacitors may all be used to store the power in the vehicle. Most vehicles' ability comes from a limited number of sources, primarily fossil fuels that aren't replenishable when it comes to internal combustion engines. Electric cars benefit greatly from regenerative braking, which returns the kinetic Energy usually wasted as heat during friction braking to the onboard battery as electricity. Figure 1 shows the overall layout of the Electric vehicle.

Figure 1.

Shows the overall layout of the electric vehicle

Battery Performance Prediction Algorithm

While several computational approaches may be used to estimate stability, this paper's Free Energy Perturbation method explicitly uses a rigorous molecular dynamics simulation methodology to account for solvent effects and sample conformational dynamics. (Scarabelli et al., 2022)The model's prediction performance is greatly enhanced using appropriate training data, as AE is kept below 2.5% and MRE approaches 0.7%. The model's superiority in SOC prediction is shown by comparing simulation results and the determination of the ideal hyperparameters. Accurate state of charge prediction for real-world battery systems using a novel dual-dropout-based neural network(R. Li et al., 2022). Compared to battery systems, this one has clear benefits. An exponential forgetting-like time-weighting strategy is used to exponentially reduce the impact of earlier data points on the regression analysis (Verbrugge et al., 2005). Using empirical mode decomposition (EMD) to reduce local fluctuation, It can be accurately predict SOH in a unified framework that includes one-step, multistep, and long-term forecasts. Then, the use the decoupled residual SOH series as the training set (Cai et al., 2022). Online optimum scheduling for building energy management is an everyday use of reinforcement learning. Regarding renewable energy sources, machine learning is often used in solar and wind power systems to predict solar irradiance, wind resources, PV power, intelligent control, fault diagnostics, and maximum power point tracking (MPPT) (Y. Zhou, 2022).