**8. Challenges and Opportunities**

The most recent research obstacles for intelligent, autonomous, and connected electric vehicle technologies are examined in this section. The following details are provided [164–173]. Better decision-making capabilities for driving are provided by autonomous vehicles, which eliminate intoxication, distraction, exhaustion, and the inability to make quick decisions. Many of these elements contribute to the technologies' capacity to outperform human decision-making abilities when it comes to driving [169]. Hence, real-time responses and error avoidance represent key hurdles for AI-integrated autonomous cars. The significance of autonomous vehicle safety and performance measures has been covered in numerous research studies. These measurements ought to take into account sensor error, programming errors, unforeseen events and entities, likelihoods of cyberattacks and threats, and hardware failures. In the future, it will be crucial to develop these indicators and analyze them in a real-time setting. The comparative evaluation of autonomous driving systems is highlighted in Table 2.


**Table 2.** Challenges and future directions of modern intelligent vehicle technologies.

Cyberattacks fall under a number of areas, including those that target control systems, driving system components, communications across vehicle-to-everything networks, and risk assessment and survey systems. Sensor attacks, mobile-application-based vehicle information system assaults, IoT-infrastructure-based attacks, physical attacks, and side-channel attacks are the main threat types that need to be investigated and examined. Moreover, cybersecurity uses artificial intelligence for attack identification. Another intriguing feature is autonomy architecture. Autonomous systems that integrate sensors, actuators, control mechanisms, a vehicle's environment for monitoring, external control variables, speed, visibility, and object identification are crucial subsystems to pay attention to and investigate in architecture.

The cost of communication will rise as the number of autonomous vehicles rises. This results in packet delay or loss, which indirectly reduces performance or increases communication error. Human life depends on autonomous vehicles and their implementation. The drawbacks of previous efforts include the lack of in-depth research of current trends such as the use of deep learning and IoT. Furthermore, it is crucial to discuss intelligent tools and software, which are not covered in the works that have already been published. Moreover, improvements in effective simulation are needed. To create autonomous vehicles, object identification, path planning, sensors, and cloud computing should all be enhanced.

Path planning and motion control for autonomous vehicles can be determined using a predictive model. A more advanced AI-based model for AVs is required. Each element of real-time architecture must be taken care of. For instance, object detection and object tracking are necessary for scene recognition [179]. Current AV architectures do not provide a start-to-end representation [180]. System errors and scalability management should be able to be handled by the AVs' architecture. As AVs must communicate with other cars in real-time while also perceiving their environment, real-time architecture is necessary. AI-based methods can accomplish this. Infrastructure and devices act as the primary agents in AVs, and they must cooperate for accuracy [181]. The SAE categorizes automation levels on a scale of 0 to 5, where 0 denotes no automation and 5 denotes complete performance. To reach level 5, businesses and researchers are working very hard [81]. According to SAEJ 3016, the following component classes are necessary for architecture:


Design, development, validation, and real-time monitoring of AVs have all considerably benefited from AI. AI is a useful tool for perception, path planning, and decision making. AVs employ AI in the following ways:


old template-matching techniques with those of more current deep learning methods in order to achieve greater efficiency. A U-shaped fully connected convolutional neural network is used to train the segmentation of vessels and backgrounds in pixels of images (Unet). Likewise, other advanced technologies such as blockchain and quantum can be explored for AV mobile information systems [182–185]. The wireless sensor network is used in autonomous vehicles for information communication [186–192].
