A Comprehensive Introduction to Self-Driving Cars — Part 1: Introduction
Introduction to self-driving car, including definition, applications, technologies, levels of driving automation.
Self-driving cars have been a hot topic for years and have received significant attentions from both academy and car industry. Once fully developed and matured, they will bring great convenience and safety to various aspects of our lives.
As a high-tech application, many cutting-edge technologies have been involved in self-driving cars, including hardware, software, and algorithms, etc., which are widely used in different industries too. Accordingly, learning and understanding how self-driving cars wok and grasping key technologies help you not only get familiar with self-driving cars per se, but also equip you with a rich set of skills that are crucial for your careers in various industries.
So, I am planning a series of articles that comprehensively guide you through different aspects of self-driving cars, including levels of driving automation and functionalities, top players, hardware configuration, software architecture, hardware/software provider, and key technologies and algorithms.
The first article answers the following questions.
- What are self-driving cars? Benefits for us?
- Commercial applications.
- What technologies are involved?
- Levels of driving automation and classification of self-driving cars.
What are self-driving cars? Benefits for us?
Self-driving cars are the vehicles with autonomous driving capability. They are able to understand their surrounding environment, interact with other traffics, and make reasonable decisions to navigate to their destinations. The autonomous driving capability provides us many benefits. Below are some examples.
- Driver’s fatigue due to a long driving can be relieved;
- The challenging driving operations, like parallel parking, reverse parking, etc., are made easier;
- Car parking problems can be effectively addressed because self-driving cars do not need to be parked nearby the destination;
- Self-driving cars are very beneficial for those who have challenges related to driving, like the elderly and the disabled.
Commercial applications
Self-driving cars have many commercial applications, such as local driverless taxi, good delivery, autonomous trucking, autonomous mining, and intelligent port operations. At present, many well-known companies worldwide, such as Waymo, Cruise, Aurora, TuSimple, etc., have been developing and deploying autonomous driving technologies for driverless taxi and autonomous trucking. At the same time, many new electric vehicle manufactures and traditional automakers have integrated different levels of Advanced Driver Assistance Systems (ADAS) on consumer-oriented vehicles, such as, Tesla’s Full Self-Driving (FSD) system, and Xiaopeng’s Navigation Guided Pilot (NGP) system.
Technologies involved
There are a variety of disciplines and technical fields involved in self-driving cars. Examples of technical fields are: computer science, artificial intelligence, robotics, sensors, data science, communications, automatic control, vehicle engineering, etc.
These technologies are often used in many other applications. For example, vehicle and pedestrian detection, lane marker detection, drivable space detection, etc., in the perception module of self-driving cars require extensive use of deep learning technology and computer vision technology. For another example, the localization module of self-driving cars is achieved by Simultaneous Localization and Mapping (SLAM) technology and vision-based or Lidar-based odometry. The sensor fusion module may be achieved by leveraging the state estimation technology, like Kalman Filter, or Particle Filter, etc. All these technologies are widely used in Augmented Reality (AR), robotics, and smart drones, etc. In addition, sensors used by self-driving cars, like camera, Lidar, Radar, etc., have also been used in the above applications.
Accordingly, skills and knowledge obtained by learning self-driving cars are essential not only for a career in the industry of self-driving cars but also careers in many other industries, like AR, robotics, and smart drones, etc.
Levels of driving automation and classification
The Society of Automotive Engineers (SAE) defines six levels of driving automation ranging from Level 0 (full manual) to Level 5 (fully autonomous) [1].
At Level 0, the driver needs to operate the vehicle all the way, and the automated driving system only provides warnings and momentary assistance to the driver. Examples include Autonomous Emergency Braking (AEB), Blind Spot Detection (BSD), and Lane Departure Warning (LDW).
For AEB, the automated driving system detects the presence of pedestrians or vehicles ahead using forward-looking radars or cameras, and sends a warning to the driver. The BSD function sends a warning to the driver if there exist vehicles in the blind spot of the ego-vehicle. The LDW function detects whether the ego-vehicle deviates from the current driving lane, and if the deviation occurs, it sends an alarm to the driver.
At Level 1, the automated driving system controls the ego-vehicle in only one direction (either longitudinal or lateral). In the longitudinal direction, the driving system provides acceleration/brake support to the driver. In the lateral direction, the driving system provides steering support to the driver. Examples of features at Level 1 are Lane Centering (LC)and Adaptive Cruise Control (ACC). For the LC feature, the automated driving system steers the ego-vehicle to drive along the center line of the current lane. The ACC feature, once receiving detections about the preceding vehicle in the current lane of the ego-vehicle using the equipped sensors like camera, Lidar, radar, etc., automatically adjusts the speed of the ego-vehicle and maintains the ego-vehicle at a constant distance from the preceding vehicle.
At Level 2, the automated driving system controls the ego-vehicle in both longitudinal and lateral directions simultaneously.
Note that for Levels 0–2, the automated driving system only provides driver assist features and the driver must constantly supervise the features for safety. Starting from Level 3, the driving autonomy changes to the automated driving features from driver assist features.
Level 3 provides the automated driving features under limited driving conditions. At Level 3, the automated driving system fulfills most of driving features without the need of supervision from the driver. In case some automated driving feature fails, the system sends warnings to the driver who needs to take over the vehicle within limited time.
Unlike the Level 2 features, which have been unified and developed over the car makers worldwide, there is a lack of unified definition about the driving conditions for the Level 3 features. Also, worldwide policies and regulations about the Level 3 driving have yet been formulated. Car makers define their own Level 3 driving scenarios. For example, Audi defined the Traffic Jam Assist (TJA) as a Level 3 feature in its A8 vehicle.
Based on the SAE definition, on one hand, the Level 3 automated driving system allows the driver’s hands off the steering wheel. Without the need of supervising the driving system, the driver is able to do other activities, like playing game, working, etc. On the other hand, the Level 3 definition requires the driver to take over the vehicle within limited time. The two aspects of the definition seem conflicting to each other. It is hard to answer: (i) to what extent can the driver be released from driving? (ii) does the driver need to constantly get ready to take over? (iii) is the driver able to take over the vehicle within limited time if the driving system fails? (iv) does the driver need to take special trainings in order to take over in time? All aforementioned uncertainties affect the evolution of the Level 3 automated driving system.
Beyond Level 3, here come the two highest levels of automated driving: Level 4 and Level 5. For the features of these two levels, the driver needs to neither supervise the features nor take over the vehicle. All the driving tasks are fulfilled by the automated driving system. Accordingly, the steering wheel and the foot pedal may not be installed in the vehicle. The difference between Level 4 and Level 5 is that Level 4 provides the features under specified conditions, like harbors and ports, mining zones, and parks, etc. The variations of these specified conditions are more predictable and lead to lower technical challenges than Level 5. Compared to Level 4, Level 5 achieves the highest level of autonomy and allows for automated driving under all conditions.
Summary
In this article, we first introduce the definition of automated driving, commercial applications, and technologies involved. We also cover in detail the SAE levels of driving automation.
In the next article, we will describe the top players in the industry of self-driving cars and their technology and product development routes.