Frequently Asked Questions About Autonomous Vehicles
In this article, we answer some of the FAQs on autonomous vehicles and self-driving cars. From levels to timelines, we're here to give you the answers you need.
Created on September 5|Last edited on January 16
Comment
We're on the precipice of a future where transportation is completely autonomous, watching self-driving vehicles transition from science fiction to reality.
In 2020, the worldwide autonomous vehicle market was valued at $76.13 billion–it's projected to reach $2,161.79 billion by 2030. Considering the investments and the rapid pace of development, we know the rise of fully-autonomous vehicles is coming, but how close are we? How long will it take before we can enjoy the perks of owning a truly autonomous vehicle? What kind of other complexities will it bring along with it? And what are the pros and cons of living in a world with autonomous vehicles?
In this article, we will cover some frequently asked questions on autonomous vehicles and understand them on a deeper level.
Here's what we'll be covering:
Table of Contents
What Is an Autonomous Vehicle?What Are the 6 Levels of Autonomous Vehicles?What Are the Advantages of Autonomous Vehicles?What Are the Disadvantages of Autonomous Vehicles?When Will Truly Autonomous Vehicles Be Available?How Do Self-Driving Cars Work?Are Autonomous Vehicles Safer?What AI Is Used in Autonomous Vehicles?What Are the Top Companies Producing Autonomous Vehicles?What Are Some of the Datasets Used by Engineers Making Autonomous Vehicles?Can an Autonomous Vehicle Be Hacked?
Let's start with the most obvious question:
What Is an Autonomous Vehicle?
An autonomous vehicle is a vehicle capable of sensing its environment and moving safely with little or no human input. Autonomous vehicles combine a variety of sensors to perceive their surroundings, such as cameras, radar, lidar, sonar, GPS, and inertial measurement units.
Based on a number of sensors positioned throughout the vehicle, autonomous automobiles build and update a map of their surroundings. Radar sensors keep track of the whereabouts of adjacent automobiles. Traffic lights, road signs, other vehicles, and pedestrians are all detected by video cameras. Lidar (light detection and ranging) sensors estimate distances, find road boundaries, and recognize lane markers by reflecting light pulses off the environment around the car.
When parking, ultrasonic sensors on the wheels pick up on curbs and other cars.
There are 6 levels of autonomous vehicles that are classified according to their level of automation. These 6 levels can be further classified into Driver Support and Automated Driving. The driver support vehicles are meant to only assist the driver, whereas, in automated driving vehicles, the human is considered a passenger.
According to the SMMT, "There are two clear states - a vehicle is either assisted with a driver being supported by technology or automated where the technology is effectively and safely replacing the driver."
What Are the 6 Levels of Autonomous Vehicles?
The Society of Automotive Engineers (SAE) created a classification system that specifies the level of driving automation that a vehicle and its equipment may offer in order to establish agreed-upon criteria early in the transition to autonomous vehicles.
There are 6 levels of driving automation ranging from 0 (fully manual) to 5 (fully autonomous). These levels can be further classified into driver support (levels 0 to 2) and automated driving (levels 3 to 5).
Level 0: No Driving Automation
The human driver is completely responsible for all actions required for controlling the car. Cars at this level may have a warning system that can warn us to take action, but the car itself can't take action on its own. The features are limited to providing warnings and momentary assistance. Example features include blind spot warnings, and lane departure warnings.
Level 1: Driver Assistance
The advanced driver assistance system (ADAS) on the vehicle can provide steering OR brake/acceleration support to the driver. It must be noted that the ADAS can only provide any one of these supports at a time. Example features include lane centering OR adaptive cruise control.
Level 2: Partial Automation
The advanced driver assistance system (ADAS) on the vehicle can manage both steering AND brake/acceleration at the same time, instead of just one or the other. Tesla's autopilot is classified as level 2. Example features include simultaneous lane centering AND adaptive cruise control.
Level 3: Conditional Driving Automation
The autonomous driving system (ADS) on the vehicle can perform all actions required for controlling the car in certain situations only. The driver must be alert at all times so that when the ADS requests, the driver must take control of the vehicle. Example feature is a traffic jam chauffeur.
Level 4: High Driving Automation
Under limited conditions, the autonomous driving system (ADS) on the vehicle can monitor the driving environment and perform all driving functions required and will not operate unless all required conditions are met. Google's Waymo falls under this level, as it is limited by geofencing (limited geographic conditions).
Level 5: Full Driving Automation
The autonomous driving system (ADS) will not require the driver to take over driving and it can drive the vehicle under all conditions. Vehicles in this level can be called truly autonomous.
You'll find more information on the autonomous driving levels in the article, "The 6 Autonomous Driving Levels Explained".
💡
What Are the Advantages of Autonomous Vehicles?
The advantages of owning an autonomous vehicle are multifaceted and range from improved safety, traffic efficiency, environmental friendliness, convenience, and more.
Here's what self-driving vehicles can bring to society:
Improve safety: According to World Health Organization's Road Traffic Injuries Report, approximately 1.3 million people die each year as a result of road traffic accidents. It is said that one of the major factors for road crashes is human errors such as drunk or distracted driving. With the help of AVs, we can eliminate these costly human errors.
Traffic efficiency: A CAV energy impacts study indicated that adaptive cruise control if used by a majority of vehicles, could increase lane capacity by up to 80%. Even decreasing the number of accidents could reduce congestion, because up to 25% of congestion is caused by traffic incidents. Source.
Environmental gains: Reduced traffic congestion and maintaining a constant speed could significantly reduce fuel consumption and in turn reduce carbon emissions. A report from Ohio University states: "The transition to the new-age cars is expected to contribute to a 60% fall in emissions".
Independence: Many seniors and people with disabilities are dependent on others for mobility. With the help of self-driving cars, tasks like going to the doctor, and visiting family could become easier for seniors and those with disabilities.
What Are the Disadvantages of Autonomous Vehicles?
Like how every great technology has its cons, autonomous vehicles have their fair share of downsides too:
High upfront cost: Industry experts estimate that a self-driving car could cost hundreds of thousands of dollars in its initial stages. Due to this sky-high price, self-driving cars would be unaffordable for most people. As time goes by and technology improves, the price will drop significantly. Self-driving cars would, therefore, initially be out of reach for the majority of people, but eventually, they would probably be within reach for the majority of middle-class families.
Technical errors: Even though autonomous cars are built taking into account multiple safety factors, and the chances of failure are quite low, there have been accidents with self-driving cars in the past. This is because autonomous vehicles are still in the development stage and are not perfect. In the due course of time, there will be fewer accidents due to autonomous cars, and there will be more features where the user can take control in case of an emergency or failure.
Job displacement: Autonomous vehicles could potentially displace some jobs, such as truck and taxi drivers, as they become more prevalent. It is important to consider the potential impact on these workers and how to mitigate any negative effects.
Moral machine dilemma: Despite their potential to improve road safety, they still lack the ability to make judgments between multiple unfavorable outcomes. For instance, in case of an emergency, a self-driving car has to take any one of two actions: A or B.
- A: Hitting a tree, which could damage the car and injure the passengers inside.
- B: Striking a pedestrian.
The car now has to decide whether it wants to save the pedestrian or the passenger(s). To learn more about the moral machine dilemma, you can visit MIT's Moral Machine Platform.
When Will Truly Autonomous Vehicles Be Available?
Some estimates predict that level 4 autonomous vehicles will be on the road by 2025 though level 5 fully-autonomous vehicles likely won't be available until at least 2030. Even if that target is hit, however, they'll likely be years away from wide adoption.
Despite the heavy investments in the autonomous vehicles industry and the promising results, we still have a long way to go to enjoy a ride in a truly autonomous vehicle. It must be noted that there is a huge gap between a level 2 (partially automated) vehicle and a level 5 (fully automated vehicle), and there are a lot of factors that limit us from reaching level 5 AV. Limits like weather conditions, traffic conditions, and road conditions could vary a lot from region to region.
The highest level we have reached so far in this race for truly autonomous vehicles is in cars equipped with Waymo, a division of Alphabet, which has reached geofenced level 4 (Phoenix, Arizona). Waymo's CEO, John Krafcik, believes that a self-driving car that can drive in any condition, on any road, without ever needing a human to take control - usually called a "level five" autonomous vehicle - will basically never exist.
How Do Self-Driving Cars Work?
Self-driving cars work using sensors, actuators, machine learning systems, complex algorithms, and highly optimized processors to execute software. Based on this system, autonomous automobiles build and update a map of their surroundings and determine how to navigate it.
Radar sensors keep track of the whereabouts of adjacent automobiles. Traffic lights, road signs, other vehicles, and pedestrians are all detected by video cameras. Lidar (light detection and ranging) sensors estimate distances, find road boundaries, and recognize lane markers by reflecting light pulses off the environment around the car.
When parking, ultrasonic sensors on the wheels pick up on curbs and other cars. Then, sophisticated software analyses all of this sensory data, draws a path, and issues command to the actuators in the automobile that manage acceleration, braking, and steering. The software aids in adhering to traffic regulations and avoiding obstructions through the use of hard-coded rules, obstacle avoidance algorithms, predictive modeling, and object identification.
Are Autonomous Vehicles Safer?
Autonomous vehicles are programmed with safety as a primary consideration. Fully automated vehicles will be able to communicate with each other and avoid most collisions.
Additionally, these systems will also be able to analyze traffic and detect bystanders, cyclists, and wildlife by using various technologies like LiDAR, Radar, and Vehicle-to-Vehicle Communication.
The World Health Organisation's 2022 Road Traffic Injuries Report shows that approximately 1.3 million people die each year as a result of road crashes. Autonomous vehicles can reduce this number to a significant amount depending on the adaption rate.
So will they be safer than a human driver? Certainly.
This said, with current Level 2 assistance systems like ADAS, these systems can only be used in very specific situations like highways, but human intervention and attention are always necessary.
Carrying this into the future, these will always be nuances of concern.
Autonomous vehicles will reduce basic accidents like backing out of a parking space into someone or reduced-attention accidents, but there will always pose the risk of unintentional accidents. This can be observed in the 2018 Tesla Crash, in which an Apple Engineer died in a crash against a barrier.
Overall, it is likely that autonomous vehicles will be able to reduce the number of crashes caused by human error, but they are not yet able to completely eliminate the risk of accidents. It will take more time and research to fully understand the safety implications of autonomous vehicles and determine how they compare to human-driven vehicles.
What AI Is Used in Autonomous Vehicles?
There are three leading AI technologies being used in Autonomous vehicles: Computer Vision, Machine Learning, and Neural Networks.
Computer Vision, the technology which enables computers to see, allows our vehicle to detect objects of various kinds, including but not limited to bystanders, cyclists, wildlife, traffic signs, and other vehicles.
Machine Learning, the technology which allows machines to learn by themselves, is a major part of our autonomous vehicle system. ML is used for Driver Monitoring, Sensor Fusion, Security, etc.
- Driver Monitoring systems are used to detect if the driver is awake or sleepy.
- Sensor Fusion is the ability of a system to combine various sensors to get a deeper understanding of the environment.
- Sensor Fusion is used in Vision, Cameras and other sensors (radar, LiDAR) concurrently process information to produce a clear image.
- Security concerns involving Hacking are carefully managed by Machine Learning technologies to detect and increase the security of the system.
Neural Networks play a major role in Navigation systems. Navigation in current Autonomous vehicles is accomplished through the usage of Deep Neural Networks (DNNs).
What Are the Top Companies Producing Autonomous Vehicles?
Several companies around the world are working on the development of autonomous vehicles. Some of the top companies in this field include:
- Waymo: Waymo is a self-driving technology company that was spun off from Google in 2016. It is widely considered to be a leader in the development of autonomous vehicles, and it has conducted extensive testing of its technology on public roads.
- Cruise: Cruise is a self-driving technology company that was acquired by General Motors in 2016. It is focused on the development of autonomous taxis and has been testing its technology on public roads in a number of cities.
- Tesla: Tesla is a well-known manufacturer of electric vehicles that have also been working on the development of autonomous driving technology. Its Autopilot feature, which is available on its electric vehicles, allows for partial automation of the driving task.
- Aurora: Aurora is a self-driving technology company founded by former executives from Google, Tesla, and Uber. It is focused on the development of autonomous driving technology for use in a variety of applications, including delivery and transportation.
- Apple: Apple has been rumored to be working on the development of autonomous vehicles for several years, and it has recently been granted permission to test its technology on public roads in California.
Many other companies around the world are also working on the development of autonomous vehicles, including traditional automotive manufacturers, technology companies, and startups.
What Are Some of the Datasets Used by Engineers Making Autonomous Vehicles?
Data is a critical ingredient for machine learning. The quality ( size and coverage, diverse driving environments, resolution, dense labeling ) of the data used has a great impact on the results. Therefore, engineers use a rich and diverse set of real-world high-quality multimodal sensor datasets to train their models and algorithms. Engineers working on autonomous vehicles typically use a variety of datasets to train and evaluate their models. Some common types of data that may be used include:
Lidar point clouds: Lidar sensors are used to create 3D models of the environment by emitting lasers and measuring the reflected signals. These point clouds can be used to detect and classify objects in the environment.
RGB camera images: Cameras mounted on the vehicle can be used to capture images of the environment, which can be used to detect and classify objects in the scene.
Radar data: Radar sensors can be used to detect objects and measure their distance and velocity. This data can be used to detect and track objects in the environment.
GPS and IMU data: Global positioning system (GPS) data can be used to determine the location and orientation of the vehicle, while inertial measurement unit (IMU) data can be used to measure the vehicle's acceleration and angular velocity.
Vehicle control and sensor data: Data from the vehicle's control systems, such as steering and acceleration, can be used to understand how the vehicle is interacting with the environment.
These datasets are often combined and annotated with labels indicating the presence and location of objects in the environment, such as pedestrians, vehicles, and traffic signs. This annotated data is used to train machine learning models that can detect and classify objects in real-time as the vehicle is driving.
Companies like Waymo and Audi have open-sourced their datasets for researchers and students. They also provide a getting started/tutorial notebook. You may find the links to these datasets and notebooks below.
Can an Autonomous Vehicle Be Hacked?
The European Union Agency for Cybersecurity's (ENISA) report states that self-driving vehicles are vulnerable to hacking due to the advanced compute they contain. The ENISA report stated that automakers should take precautions against a variety of attacks, including adversarial machine learning, sensor attacks using laser beams, overburdening object detection systems, and malicious activities on the backend.
According to the report, artificial intelligence systems may target autonomous cars and cause harm that would be difficult for humans to notice. Automakers will need to regularly check the software in self-driving cars to make sure it hasn't been tweaked in order to stop such attacks. Even with today's semi-autonomous vehicles on the road, experts claim that the possibility of attacks is real.
Feel Free To Add Your Question!
We always want to make sure we're answering the questions you want to know.
Feel free to ask anything about autonomous vehicles in the comments below and we'll do our best to answer them.
Add a comment
Iterate on AI agents and models faster. Try Weights & Biases today.