The Future of Self-Driving Cars: Can They Truly Become Autonomous Vehicles?

Remember how Tom Cruise’s character, Chief John Anderton, uses his Level 5 autonomous vehicle Lexus CS 2054 in the movie Minority Report? Or maybe you saw Detective Del Spooner (Will Smith’s character) ride an Audi RSQ concept car in self-driving mode in I, Robot. Well, driverless cars are no longer just a concept or part of science fiction movies. Hundreds of thousands of these vehicles are driving the roads of multiple countries worldwide, and the future of these self-driving cars looks more autonomous than ever.

Be it Waymo’s robotaxi or Tesla’s Full Self-Driving (Supervised) vehicle, you can find an increasing adoption of these automobiles by regular users. Unlike in the beginning, people are now becoming a little more comfortable with riding in these cars. For instance, a 2018 Brookings survey found that only 21% of adult internet users were inclined to ride a self-driving vehicle, while 79% feared it. However, when the American Automobile Association (AAA) conducted a survey last year, that number came down to 60%.

With this increasing adoption, the future of autonomous vehicles (AVs) looks safe, but there’s still a long path for the automotive and self-driving technology fields to navigate.

What Is an Autonomous Vehicle?

An autonomous vehicle is any car, truck, or other mode of transportation that uses technologies like artificial intelligence (AI), Internet of Things (IoT), and more to study its surroundings to operate with little to no human input. The formal references to AVs come from the Society of Automotive Engineers (SAE) International’s J3016 standard.

At its core, there are three pillars that allow AVs to run themselves:

• Perception: These vehicles assess and understand their surroundings, from objects to pedestrians.
• Decision-making: They predict what will happen next, such as whether a pedestrian stops or continues to walk, and choose a safe course of action accordingly.
• Control: Have complete control of the automobile to steer, brake, or accelerate when needed.

But how do driverless cars execute these three interconnected pillars?

How Do Self-Driving Cars Work?

A self-driving car operates based on two core environments. The first of them is the onboard stack, which focuses on sensing and understanding the environment the AV is in. The other one is the backend or cloud stack. The primary responsibility of this stack is to collect and process data to continuously train the autonomous vehicle and improve decision-making accuracy.

This all starts with the sensor layer. An AV is equipped with multiple sensor systems facilitated by the Internet of Things to make the car smart. One of the most important systems to capture the surroundings of a car is LiDAR (Light Detection and Ranging). It is a high-tech scanning method used to create 3D maps of the surroundings. The system shoots rapid light pulses towards a target and notes the time they take to bounce back to create the map. However, Elon Musk believes that cameras and artificial intelligence (AI) are all a self-driving car needs.

In an X post he replied to in 2025, he said that “LiDAR is a fool’s errand.”

But regardless of what he believes, most autonomous vehicles running on the road today use LiDAR technology, including Waymo.

Besides LiDAR, there are cameras, ultrasonic sensors, and an inertial measurement unit. All these devices and sensors have the same goal of helping self-driving cars understand their surroundings.

Once the vehicle has the data of the environment, computing begins. Many cars include a kind of specialized data center shrunk to fit in the trunks of the vehicles or under the seats. NVIDIA’s Drive Thor/Orin platforms or Tesla’s custom AI4 chip are some examples. The purpose of this compute platform is to process neural network inference with low latency for real-time response.

These computing systems are usually convolutional neural networks (CNNs) or transformer-based architectures that classify and track objects for localization. Localization allows AVs to answer “Where am I?” and “Who or what is around me?”

The next stage after localization is perception, which tells self-driving cars what the nearby objects or people would do next. With the help of artificial intelligence, the system generates insights like the pedestrian has a 70% chance to stop or a rolling ball has an 80% chance to come in the car’s path. The system considers these insights to make decisions whether to accelerate, decelerate, stop, steer, etc.

All this data and decision-making is constantly sent to the cloud for further analysis and future reference. Engineers use this data to train better algorithms to further reduce the chances of mistakes and accidents.

Based on how accurately these AI systems work and how much humans need to intervene, the SAE has classified self-driving vehicles into different autonomy levels.

Autonomous Vehicle Levels

The SAE J3016 framework, which remains the industry standard for classification, has defined six levels of autonomy, which are:

Benefits of Driverless Cars

Ever since the first car with some level of autonomy hit the road, industry experts are pushing towards getting to a standard where human involvement is minimal to none. And this push is not just a tech industry flex but because there are genuine benefits of autonomous vehicles, including:

Enhanced Road Safety

As per estimates by the World Health Organization (WHO), around 1.16 million people die annually in road traffic crashes, and more often than not, these casualties are a result of very basic human errors. The United Nations General Assembly has set a target to halve that number by 2030. Autonomous vehicles come with a promise to help achieve that target.

The Insurance Institute for Highway Safety conducted a study on autonomous vehicle safety and released the results last month. It concluded that Waymo’s Level 4 vehicles had a 91% lower rate of rear-ending another vehicle. Similarly, its rate of being rear-ended was also 40% lower. These self-driving cars were involved in far fewer crashes compared to human-driven automobiles.

Use of AI in self-driving cars minimizes errors that humans could make. Moreover, AI algorithms can achieve 95% accuracy in predicting how the people or objects around a car would act. Thus, AVs can make the right decisions to prevent clashes.

Increased Accessibility

Be it the elderly or someone who is physically challenged, many individuals worldwide cannot ride a vehicle. In the end, they have to rely on someone else for transportation, which erodes the sense of independence. Driverless cars let them travel from point A to point B without needing anyone else’s support. They can go to their hospital or clinic visits or simply go out and meet different people to curb loneliness, which is a growing issue among these populations.

Reduced Congestion

TomTom Traffic Index released a list of the most congested cities worldwide. Mexico City, Mexico, topped the list with 75.9% congestion; followed by Bengaluru, India (74.4%); Dublin, Ireland (72.9%); and Lodz, Poland (72.8%). While this congestion is largely because of poor infrastructure, traffic management also plays a key role. And that’s something driverless cars can help with.

For instance, these AI-enabled automobiles can maintain consistent speeds and safe following distances. As the need for unnecessary braking and acceleration reduces, the traffic flow becomes easier to manage. It is these stop-and-go waves that spread across traffic and result in congestion. Besides that, these vehicles also travel with consistent spacing between different vehicles to increase the total number of cars or trucks that can use a particular road at any given time.

Additionally, AVs can communicate with each other and traffic infrastructure for coordinated driving. This results in more efficient route selection and intersection approach.

Minimized Emissions

This is associated with reduced congestion, too, because when traffic flow is optimal, it results in less time spent by the vehicle on the road, which influences the total emissions. AI-enabled AVs can even optimize braking and acceleration. With better use of regenerative braking, the need for energy consumption goes down, which again affects the carbon footprint. Similarly, optimized routes, reduced idling, and smoother driving patterns.

The Future of Self-Driving Cars

Experts believe that the future of autonomous vehicles is bright.

“I think if you go as far as 2040 or ’50, AV technology will be a commodity. I think everybody will expect that the new cars that you buy have level-four autonomous capability. AV technology is going to be like the shower in your hotel room: You just expect it.”

Says Philipp Kampshoff, a McKinsey & Company expert.

We are already witnessing a shift in this as the number of AVs on the roads keeps increasing. Currently, around 20 cities worldwide, primarily in the US and China, have commercially operated Level 4 robotaxis on roads. In fact, Goldman Sachs estimates that the number of self-driving cars on US roads could increase from just 1,500 currently to approximately 35,000 by 2030.

Thanks to this increasing adoption, plans for infrastructural changes are also taking place. United Nations Economic Commission for Europe (UNECE) has recently announced a global regulatory framework for the safe introduction of these vehicles on public roads worldwide. This could help avoid fragmented national approaches.

Some technologies that could play a vital role in this future of autonomous vehicles are:

Agentic AI

Whether you call it agentic AI or AI agents, this methodology, which lets artificial intelligence systems plan, decide, and act, could be the breakthrough engineers are looking for to attain Level 5 autonomy. These systems are capable of multi-step reasoning, adaptive planning, and handling novel situations rather than just pattern-matching against training data. Hence, could self-learn and operate in new areas even without heavy training from scratch.

Blending agentic AI solutions within autonomous vehicles would let them make better decisions in complex traffic. Similarly, agentic AI will optimize route planning. For instance, instead of simply following the fastest route, AI agent-powered autonomous vehicles will take into account additional factors, such as traffic congestion, road construction, weather conditions, accident reports, battery or fuel levels, etc., before selecting the most optimal route.

However, both organizations and governments need to first assess agentic AI readiness before they can successfully implement these solutions at scale.

5G and V2X Connectivity

Low-latency 5G networks are improving connectivity among self-driving vehicles and between these autonomous automobiles and other IoT devices. This enables V2X (Vehicle to Everything) communication. This means that vehicles will communicate with other vehicles to share their speed, position, and intent data. Additionally, they will connect with smart traffic signals and road sensors to better understand their surroundings.

This connectivity allows vehicles to “see” beyond their own sensor range. For example, they could anticipate a red light change or a vehicle running a stop sign around a blind corner. This is something no onboard sensor suite alone can achieve. Access to all this data in real-time will further reduce the chances of accidents.

Edge Computing and AI Chips

As onboard AI chips become more powerful, even inference could happen on board without relying too much on round-trips to the cloud. This will reduce latency and improve split-second decision accuracy. It is a critical requirement on the roads because even milliseconds matter there.

What’s Keeping Autonomous Vehicles From Reaching Level 5?

Although SAE Level 4 is reached and could become mainstream in a few years, attaining Level 5 autonomy is still a challenge.

To begin with, there are many edge cases that AVs can never fully understand. Unpredictable human behavior, extreme weather, unmarked construction zones, and unusual road debris are difficult to comprehend for even the most advanced models currently available. While the risk can never be completely eliminated, driverless vehicles will have to improve the accuracy of their predictive and decision-making models to reduce these edge cases significantly.

Apart from that, regulations surrounding AVs and their accountability differ across states and countries. The same goes for infrastructure gaps because smarter cars need smarter roads, and not every region is capable of facilitating these self-driving automobiles yet. Only a few cities in a handful of countries are ready even for Level 4 deployment yet. The Conversation tested running a Tesla Model Y with FSD for 100 days on Queensland, Australia roads. The self-driving car struggled on bridges, in school zones, boom gates, and railway crossings. All of these are part of infrastructural gaps that need to be identified and narrowed before the arrival of Level 5.

Another challenge is cybersecurity risks. Anything that connects with the internet can be hacked. That’s exactly what happened with Tom Cruise’s character in the movie Minority Report. In a sequence where Cruise’s character, John Anderton, is trying to escape, the police take control of his car remotely and redirect it. The same can happen in the real world, and it can have severe consequences, such as harm to someone’s life.

What Industries Can Benefit From a Successful Future of Self-Driving Cars?

The advent of complete autonomy in self-driving cars is not just beneficial for the automotive industry, as many more can reap the benefits.

• Public transit and ride-hailing: Without any doubt, public transport would benefit the most, as robotaxis are already proving. The World Economic Forum believes that AVs and public transport can play complementary roles. On one hand, driverless vehicles offer direct and flexible service, and on the other, public transport has high-capacity, scalable solutions. Thanks to that, AVs can extend the reach of public transport and even replace underperforming bus routes through on-demand, microtransit options, especially in rural areas.
• Freight: The freight industry plays a role in multiple sectors, such as transportation, hospitality, retail, etc. Consider the example of the retail industry, where last-mile delivery is a costly challenge. This final part of the larger supply chain suffers from high labor costs, heavy fuel consumption, and failed delivery attempts. The retail industry has long been using artificial intelligence for purposes like personalized experiences, dynamic pricing, demand forecasting, and much more. AI-powered autonomous vehicles in retail can add to that by making deliveries easier, less costly, and always on time. PepsiCo and Walmart are already leveraging this technology for deliveries in coordination with driverless vehicle operators like Gatik.
• Healthcare: AVs can bridge the access gap in the healthcare sector. For instance, they can help the elderly maintain their regular checkup schedules. Similarly, they can allow transporting medical supplies to remote areas or expand non-emergency patient mobility. Healthcare organizations can even use them to power mobile clinics.
• Waste management: Many municipalities worldwide are already using artificial intelligence and IoT sensors to sort trash to separate recyclable and hazardous materials from regular waste. However, the collection of that trash and moving it from residents’ houses to landfills or waste management facilities still happens with the help of human transporters driving the trucks. The use of autonomous vehicles can free these drivers to focus on other strategic tasks, such as handling bins or checking collection points.

Conclusion

Autonomy is a long and uneven climb up the SAE ladder, and many automobile or technology giants are yet to reach the highest point. While some companies like Waymo have been able to produce and operate Level 4 autonomous vehicles, reaching Level 5 still remains a long-term horizon.

It is mostly first-time passengers who are more afraid because once someone experiences the ride, the technology is good enough to offer them a comfortable ride, which increases their trust. Advances in technology and agentic AI will further make autonomous vehicles more capable of operating in diverse locations and conditions without human intervention. Therefore, the future of self-driving cars still looks bright because of growing adoption and increasing trust by users.

Frequently Asked Questions