Under the Hood of Self-Driving Vehicles: The Technology Making them Work
Right from the development of the first automobile in the 18th century to this day, mankind has continued to search for ways to improve mobility through the automobile, and as such, we have seen countless innovations amounting to the improvement of the automobile. Mobility is the lifeblood of every economy, and it is essential that we keep finding ways to enhance it. The idea of self-driving vehicles transporting people from one place to another is not only economically viable, but also promises increased safety and a more connected city.
In this blog post, we will be looking at what self-driving vehicles are, what the Technology making them work looks like, and how we can benefit from them. This blog post will cover:
- The History of Self-Driving Vehicles
- Technical Approaches to Self-Driving Vehicles
- Levels of Self-Driving Technology
- Self-Driving Technology Achievements So Far
- The Future of Mobility
The History of Self-Driving Vehicles
Self-driving vehicles have been making headlines in the news and on the internet lately, and despite their impressive achievements and scale of operation, the technology isn’t new.
The first self-driving car ever developed was in 1995, by The Robotics Institute at Carnegie Mellon. It was called Navlab5 (see image below), and was successfully piloted from Pittsburgh to San Diego in the US.
Navlab5 was an autonomous car that was able to steer itself, and although the driver behind the wheel was responsible for accelerating and braking, the Navlab5 autonomous car made use of cameras mounted on the it, and a variety of other sensors to navigate over 2000 miles without human intervention, which was a huge milestone in the history of autonomous vehicles and robotics.
Prior to the development of the first ever self-driving car, Ernst Dickmanns, a German Computer Scientist, who is considered the pioneer of dynamic Computer Vision and Self-Driving cars, developed the world’s first self-driving vehicle in the 1980s, which was a Mercedes-Benz van called VaMoRs, retrofitted with tons of sensors and cameras.
The VaMoRs self-driving van was tested several times on the streets of Europe, and returned positive results. It was able to leverage advanced Computer Vision algorithms to perform driving techniques such as lane changing and vehicle tracking, which all today, play a key role in achieving true self-driving.
The VaMoRs van can be seen above on the left, and the inside of the vehicle together with onboard computers can be seen on the right. Here is a YouTube video showing the demonstration of the VaMoRs self-driving van on the streets of Europe back in the 1980s: link.
Technical Approaches to Self-Driving Vehicles
In order to achieve self-driving vehicle technology, highly sophisticated and computationally intensive algorithms are developed to process trillions of data points coming from the outside world, to give on-board computers the ability to make informed decisions on what driving techniques to adopt in order to move a vehicle from one point to the other.
In general, many of the industry players working on self-driving technology adopt one or both of the two techniques:
The first approach involves using a wide range of sensors such as Lidar, Radar, GPS, and Cameras in order to give the self-driving vehicle the ability to perceive the world around it, in computational form. This data is gathered and fed to an algorithm that is able to perform the actual perception of the environment. This usually comes in the form of using Artificial Intelligence techniques to detect lanes on roads, traffic signs, humans and a wide range of other objects that may be of interest to the algorithm.
Once the environment is perceived, data is passed to the planning layer (in red), where a set of trajectories for the vehicle to take is generated, considering factors like where free space is located or where pedestrians are located in order to either stop and wait for them to pass, or move around. The vehicle then uses a control algorithm to navigate from one trajectory to another, ensuring the vehicle moves according to the generated trajectory.
The second approach mainly relies heavily on a Reinforcement Learning algorithm, to perform self-driving. In this approach, driving data (in the form of videos) including their corresponding data points generated from sensors, is fed to a Reinforcement Learning algorithm.
This approach often relies less on sensors, and is often regarded as an ideal approach to self-driving; as the vehicle is considered to be taught to drive exactly how humans would drive. The main challenge with this approach is the scale of data collection and training required in order to provide a driving agent capable of driving close to how a human would. Even though this is quite a challenging approach, there are a few companies that have been successful in developing a highly reliable driving agent using this approach.
Levels of Self-Driving Technology
As research into the Technology of self-driving cars was heightening in the early 2000s , researchers and stakeholders decided to come up with a way to classify the levels of self-driving vehicles.
In order to achieve this, a scale was introduced. This scale starts from 0, all the way to 5, which represents no self-driving capabilities to complete self-driving capabilities respectively. The scale was put into the following levels:
Level 0 (No Automation): This level indicates a vehicle has no self-driving capabilities, and requires manual control by a driver behind the wheel, providing acceleration and braking controls.
Level 1 (Driver Assistance): This level of self-driving capability includes a single automated system for driver assistance, such as accelerating and braking. This is often a feature implemented as cruise control, and is common in modern day cars such as the 2020 Volvo S60 and the R2020 Hyundai Sonata. In level 1, the vehicle can perform lane centering and can maintain a reasonable distance between cars when in cruise control. This level also requires a human to do most of the work and be in control of the car at all times.
Level 2 (Partial Driving Automation): In this level of self-driving capability, the car is equipped with Advanced Driving Assistance Systems (ADAS), where the ADAS systems take control of steering, braking, and acceleration in very specific situations, including low-traffic environments like highways. The ADAS systems are able to perform both lane centering and adaptive cruise control simultaneously. In level 2, the driver must remain alert throughout and is required to intervene in many situations.
Level 3 (Conditional Driving Automation): In level 3, cars have “environmental detection” capabilities and are able to make informed decisions for themselves, such as accelerating past a slow-moving vehicle, and can run mainly on their own and require human intervention only in cases of extreme environments and failures. An impressive achievement is Tesla Motors’ Autopilot feature.
Level 4 (High Driving Automation): Level 4 self-driving capability makes it possible for cars to operate in real self-driving mode. At this level, the car is able to move from point A to point B without the need for human intervention. There have been quite a few industry players that have achieved this level of self-driving. Notable among them include Google’s Waymo self-driving cars, and that of Cruise. Both are currently offering self-driving ride hailing services in the US.
Level 5 (Full Driving Automation): Level 5 self-driving is the highest level of self-driving technology. Cars or vehicles with level 5 self-driving are capable of driving without human intervention in all conditions and environments. These cars have sensors and cameras among other technologies, that allow them to see and understand the road, making decisions and navigating without the need for a human driver. This is the pinnacle of true self-driving vehicles, and by far, no industry player has achieved this level.
Self-Driving Technology Achievements So Far
Despite the extreme level of difficulty involved in developing Self-Driving vehicles, we can boast of a few highly successful and operational self-driving vehicles that are legally driving on city streets and highways, transporting people from one location to another. Some of the notable achievements by far include:
General Motors’ Cruise
As described by Wikipedia, Cruise LLC is an American self-driving car company headquartered in San Francisco, California. Founded in 2013 by Kyle Vogt and Dan Kan, Cruise tests and develops autonomous car technology. The company is a largely-autonomous subsidiary of General Motors.
Cruise has two divisions of self-driving ride services: a ride hailing self-driving vehicle service currently operating in California, and a self-driving transportation robo-taxi still in development. The currently level 4 self-driving vehicle Cruise is using for its ride hailing services, has a ton of sensors, and a powerful onboard computer using high resolution 3D maps of cities for navigation. Below is the vehicle:
Cruise’s self-driving transportation robot-taxi still in development can be seen below, together with its sensors:
Google’s Waymo
As described on Wikipedia, Waymo LLC, formerly known as the Google Self-Driving Car Project, is an American autonomous driving technology company headquartered in Mountain View, California. It is a subsidiary of Alphabet Inc., the parent company of Google.
Waymo has been one of the leading self-driving Technology companies in the world, and just as Cruise has two lines of products, Waymo too has a ride hailing self-driving vehicle service that is currently operational in some states in the US, and a self-driving transportation van still in development, although some debuts have been made.
Waymo’s self-driving Technology is a level 4 one, where no human driver is required. The vehicle Waymo uses for its ride hailing service can be seen below:
Waymo’s transportation van which is still in development and is said to launch in the coming years can be seen below:
The Future of Mobility
All in all, the endgame of self-driving vehicles is to create a safer, more efficient, and convenient transportation system. With these vehicles, the burden of driving is taken away from the driver, allowing for a more relaxed and enjoyable travel experience. The use of self-driving vehicles also has the potential to significantly reduce traffic accidents, which are often caused by human error.
As we usher further into the digital age, there will continue to be a rising demand for more efficient ways of doing things, leveraging digital Technology. This blog post covered at a high level, what it takes to make self-driving cars work, and has touched on a few achievements so far worth showing to the world.
The future of mobility is one that is electric, and definitely one that will involve self-driving transportation. The Technology is here, and we have to embrace it in order to make transportation safer and more reliable.
Peace!
