The World of AI and Driverless Cars
I remember a couple years back when I was traveling around New York city with my uncle. He expressed how excited he was that he would soon receive his Tesla. He went on for hours about how Tesla was going to revolutionize the automobile industry and driving forever.
My interest in Tesla and driverless cars continued on, long after my New York trip. I remember reading Elon Musk’s biography and being fascinated with his vision for Tesla. A couple years since my New York trip and the world of EVs and driverless cars has had unimaginable progress. To everyone’s surprise, Tesla does not even lead the pack of self-driving cars anymore.
The 6 levels of autonomous driving
There are 6 levels of autonomous driving that define the amount of automation a car has control over without human supervision.
Below is a brief summary of the various autonomous driving levels
Level 0: (No Assistance)
No Automation - The driver has full control of the vehicle. This includes most cars that have been produced throughout most of last and this century.
Level 1: (Driver Assistance)
The vehicle can assist with steering or braking, but the driver is still responsible for most aspects of driving. Example:
Level 2: (Partial Automation)
The vehicle can take control of steering, acceleration, and braking in certain situations, but the driver must still be ready to take control at any time. Example: Cadillac Super Cruise and Tesla Autopilot
Level 3: (Conditional Automation)
The vehicle can take full control in certain situations and do not require human control in most circumstances, but the driver must remain alert and be ready to override when alerted by the system. Level 3 vehicles can make decisions e.g accelerating to overtake slow vehicles. Example: Audi A8.
Level 4: (High Automation)
The vehicle can take full control in most situations, but the driver may still have the option to take control if desired. These systems include traffic jam pilots. In most countries, Level 4 vehicles are only allowed to travel in limited areas. Example: Waymo, XPeng G9.
Level 5: (Full Automation)
The vehicle can take full control in all situations, and there is no need for a human driver. There are currently no Level 5 vehicles available and most advanced vehicles barely push past level 4.
Tesla recently received high level 2 autonomous driving certification. However, it is still a far cry from its other competitors that have already pushed past level 4 and are prepared to soon push for level 5.
The algorithm behind object detection
When I signed up for the Inspirit AI program, I was put into a group project aimed at creating an algorithm for basic object detection, used in nearly all driverless cars. While our algorithm pales in comparison to those used in driverless cars, the basis of our object detection algorithms are nearly identical.
YOLO and neural networks are commonly used for these algorithms. Below, we dive deeper into how these work and are applied.
Neural Networks:
Neural networks are a type of artificial intelligence that are used in a variety of applications, including driverless cars. Neural networks are designed to mimic the way the human brain processes information, using interconnected layers of nodes or neurons to analyze and interpret data.
Above is a diagram of a simple neural network. As seen above, an input is processed, analyzed and an output is released.
A Convolutional Neural Network (CNN) is a type of neural network that is specifically designed for image recognition and processing. CNNs are used to analyze images and identify patterns within them. The convolutional layers within a CNN are useful for image processing as they apply filters to the input image to extract specific features.
On the other hand, a neural network is a broader term that refers to any type of artificial neural network, including CNNs. Neural networks can be used for a wide range of applications, including speech recognition, natural language processing, and financial forecasting, in addition to image recognition.
In relation to driverless cars, neural networks are used to process data from various sensors and cameras to help the car understand its environment and make decisions about how to navigate safely. Here are some ways that neural networks are used in driverless cars:
Object detection: Neural networks are used to identify and classify objects in the car's environment, such as other vehicles, pedestrians, and obstacles. This information is then used to help the car plan its route and avoid collisions.
Lane detection: Neural networks can be used to detect lane markings on the road and determine the car's position relative to those markings. This information is used to keep the car in its lane and make adjustments as needed.
Traffic sign recognition: Neural networks can be trained to recognize traffic signs and signals, such as stop signs, speed limits, and traffic lights. This information is used to help the car navigate safely through intersections and obey traffic laws.
Path planning: Neural networks can be used to help the car plan its route, taking into account factors such as traffic conditions, road conditions, and obstacles in the environment.
Decision-making: Neural networks can be used to help the car make decisions about how to navigate safely through its environment, taking into account factors such as the car's speed, the distance to other objects, and the road conditions
YOLO(You Only Look Once)
We used YOLO to create our object detection algorithm. YOLO is a real-time object detection algorithm that was first introduced in 2016. It is a deep learning algorithm that uses a single neural network to detect objects in images and video streams.
The YOLO algorithm works by dividing an image into a grid of cells and predicts bounding boxes for each cell that contain objects. The algorithm then classifies each object within the bounding box and calculates confidence scores for each object detection.
One of the main advantages of YOLO is its speed. Unlike some other object detection algorithms e.g CNN(convolutional neural networks), that require multiple passes over an image, YOLO can process an entire image in a single pass, making it much faster and more efficient.
YOLO has been used in a variety of applications, including self-driving cars, security cameras, and robotics. It is also commonly used in computer vision research and has inspired the development of other object detection algorithms
Above is a brief summary of the process that YOLO undertakes for object detection.
Below is an example of a YOLO algorithm at work:
If you are interested in learning more about YOLO, CNNs and how our algorithm is formed, check out our slides that also explain the history behind them. YOLO Final Presentation
Case study: Xpeng’s level 4 driverless car technology
Xpeng Inc. is a Chinese electric vehicle (EV) company that specializes in developing and manufacturing intelligent EVs. The company was founded in 2015 by entrepreneur He Xiaopeng.
Xpeng's mission is to provide smart, sustainable mobility solutions that improve people's lives. The company focuses on developing high-quality, innovative EVs that incorporate advanced technologies such as artificial intelligence, autonomous driving, and connectivity.
Xpeng has also been investing heavily in research and development, particularly in the areas of autonomous driving and battery technology. The company has a team of over 3,000 engineers and researchers who are working on developing cutting-edge technologies to enhance the performance and safety of its EVs.
Before we dive into the specifics behind Xpeng’s technology, check out this fascinating post on how Xpeng cars are able to map out their geographical surroundings through the use of a series of cameras and neural networks. https://shorturl.at/djrsU
Xpeng's driverless cars, also known as Xpeng P7, use a combination of technologies to enable autonomous driving. Here are some of the key technologies that power Xpeng's driverless cars:
Lidar: (acronym for laser imaging, detection, and ranging)
Xpeng's driverless cars use multiple lidar sensors to create a 360-degree view of the vehicle's surroundings. Lidar sensors use laser beams to measure the distance to objects and create detailed 3D maps of the environment.
Cameras:
Xpeng's driverless cars are equipped with 14 cameras that provide high-resolution images of the environment. These cameras are used for object detection, traffic light recognition, and lane detection.
Ultrasonic sensors:
Xpeng's driverless cars use ultrasonic sensors to detect objects that are close to the vehicle. These sensors are also used for parking assist and collision avoidance.
Radar:
Xpeng's driverless cars use radar sensors to detect objects that are far away from the vehicle. Radar sensors are also used for adaptive cruise control and collision avoidance.
High-precision maps:
Xpeng's driverless cars use high-precision maps that are updated in real-time to provide accurate information about the environment. These maps include information about road conditions, traffic signals, and other objects in the environment.
AI algorithms:
Xpeng's driverless cars use AI algorithms to process the data collected by sensors and cameras. These algorithms are used for object detection, lane detection, and decision-making.
Together, these technologies enable Xpeng's driverless cars to navigate roads, avoid obstacles, and safely transport passengers to their destination. Xpeng is also working on developing a 5G-based communication system that will enable its driverless cars to communicate with other vehicles, infrastructure, and pedestrians in real-time.
Safety risks:
While there is a lot of promise with the future of autonomous vehicles, autonomous cars have also been riddled with problems that have resulted in mass recalls of vehicles.
Tesla, like other automotive companies, has had recalls for some of its vehicles due to safety concerns. Recalls are issued when a manufacturer or the National Highway Traffic Safety Administration (NHTSA) identifies a safety issue that could potentially harm drivers or passengers.
Below is a brief timeline of the various models that Tesla has recalled:
In February 2021, Tesla issued a recall for nearly 135,000 Model S and Model X vehicles due to a touchscreen failure issue that could affect the rearview camera display and other features.
In April 2021, Tesla issued a recall for over 5,500 Model 3 and Model Y vehicles due to loose brake caliper bolts that could affect braking performance.
In October 2021, Tesla issued a recall for nearly 12,000 Model X and Model Y vehicles due to potential seat belt issues that could result in failure during a crash.
It's worth noting that Tesla has had a history of controversy surrounding its approach to recalls. Some critics have accused the company of being slow to address safety concerns and not being transparent enough about the issues. However, Tesla has defended its approach, arguing that it prioritizes safety and compliance with regulations.
The benefits of driverless vehicles
While there are many critics towards the driverless car movement, we cannot understate how revolutionary and impactful the benefits of driverless vehicles are.
One of the most significant benefits of autonomous vehicles is improved safety. According to the National Highway Traffic Safety Administration (NHTSA), 94% of accidents are caused by human error. By removing the possibility of human error, autonomous vehicles could significantly reduce the number of accidents on the road, potentially saving countless lives and reducing the financial cost of accidents.
Additionally, these vehicles increase efficiency. These vehicles can communicate with each other and adapt to changing road conditions in real-time, which could lead to less congestion and more efficient traffic flow. Additionally, they could optimize routes and reduce fuel consumption, making the transportation system more efficient and environmentally friendly.
Autonomous vehicles could also provide greater accessibility for people who are unable to drive, such as the elderly and people with disabilities. This could help improve their quality of life and independence, allowing them to travel safely and easily without the need for a driver.
Driverless cars allow people to use their travel time more productively, such as working, studying, or relaxing, they could reduce stress and increase productivity. This could have a positive impact on people's mental health and work-life balance.
In addition, autonomous vehicles could potentially reduce transportation costs by eliminating the need for drivers and reducing insurance costs. This could make transportation more affordable for individuals and businesses, leading to increased economic opportunities.
Finally, autonomous vehicles could contribute to a cleaner environment. By reducing emissions and optimizing routes, they could help to reduce air pollution and contribute to a cleaner, healthier environment.
The future of driverless vehicles
Despite the magnitude of criticism towards autonomous vehicles, the future does look bright as more and more companies enter the fray of autonomous vehicles.
According to several projections, autonomous driving could create up to $400 billion in additional revenue by 2035, while the US autonomous vehicle market alone would balloon to $138 billion by 2030, up from $4 billion in 2021.
The greater the amount of competition in the autonomous vehicle industry, the quicker innovation takes place. However, it is of the utmost importance that autonomous vehicles are properly regulated to ensure that ‘rogue car’ situations are kept to a minimum.
In conclusion, the world of driverless cars and the technology behind them is fascinating and will bring tremendous economic benefits in the long run.
About Inspirit AI
AI Scholars Live Online is a 10 session (25-hour) program that exposes high school students to fundamental AI concepts and guides them to build a socially impactful project. Taught by our team of graduate students from Stanford, MIT, and more, students receive a personalized learning experience in small groups with a student-teacher ratio of 5:1.
