Autonomous vehicles (AVs) will usher in a new era for transportation. They will make road trips safer, more convenient and less expensive.
But before autonomous cars are fully realized, companies must overcome a number of challenges. These include perception, mapping, localization and fail-safe mechanisms.
Waymo
Autonomous cars are a major advance over traditional vehicles, with many potential benefits. These include less traffic accidents, fewer injuries, and a lower cost of ownership. However, a number of challenges still remain.
One of the biggest hurdles is a vehicle’s ability to perceive the world around it. This is a complex task that involves multiple sensors, including cameras, LiDARs, and radar. The main purpose of these sensors is to create a digital map of the surrounding environment to guide the vehicle in its path.
To do this, Waymo uses a set of highly detailed custom maps that give the car important context in every location where it will drive. This map also includes real-time sensor data, and it allows the vehicle to accurately determine its position in the world at all times.
The vehicle’s localization module combines mapping with cameras, GPS, and algorithms to get the best position possible on the road. This module has accumulated over two decades of research and experience, and is considered to be one of the most advanced modules in the self-driving industry today.
Another important part of the localization module is the vehicle’s ability to anticipate the actions of other drivers and road users. This is achieved through the use of machine learning models and real-time information about the world’s roads.
While these models can be trained using TensorFlow, the company also has its own TPUs (tensor processing units) that can train them up to 15 times with higher efficiency. These TPUs can be placed in the platform’s data centers, and they provide a faster way to update the self-driving software.
Having a large database of data is also vital to the success of self-driving tech. This data can help the system understand how a particular situation might play out in order to avoid it.
It can also use the data it has to predict the paths that other drivers and road users might take, which can save time and improve safety. The algorithm can use the data it has collected over its 20+ million miles of real-world driving and its billions of simulation miles to anticipate the behavior of other road users.
Cruise Control with Automatic Distance Control
Cruise control with automatic distance control (ACC) is a popular feature that helps drivers stay safe on long journeys by maintaining a set following distance to the car in front of them. ACC works by monitoring traffic ahead and using radar or laser sensors to detect cars in your path. It can also increase or decrease the following distance as needed, and can be adjusted with a button on your steering wheel.
The technology is available on a wide range of vehicles and is used by most new cars sold in the US. It’s a useful technology that can save drivers time and money, especially on long trips.
It also makes driving safer and more comfortable, reducing the need to use the accelerator pedal in high-speed situations. It can reduce braking in heavy traffic and help keep you on the right side of the road during lane changes.
Adaptive cruise control (ACC) systems have been around for years, but have only recently gained the ability to control the throttle and brakes of your car. Early systems, such as lidar-based Distance Warning and laser-based Preview Distance Control, only warned of slower traffic and didn’t actually control the car’s speed.
Later, ACC systems were introduced with radar-based technologies that could control the car’s acceleration or braking by sensing changes in traffic. The technology became a staple on American cars in the 2000s and now is almost always found in combination with pre-crash systems that can alert you to hazards or begin braking before a collision.
The system uses an in-car computer to monitor the speed and direction of the car in front of it, and can slow or accelerate the vehicle to maintain a pre-set distance from the car in front of you. It can also increase or decrease the distance in response to changes in road conditions and other factors that affect traffic.
ACC is one of the most advanced self-driving car techs, and is an essential component for fully autonomous driving. It can be paired with other driver assist features, such as lane centering, to ensure the vehicle is where it should be at all times.
Integrated GPS
Integrated GPS (sometimes called an autopilot system) is a feature that is built into many cars. This technology relies on the global positioning system, which is a satellite-based navigation system that uses 24 satellites to localize a car’s position.
Compared to standalone navigation systems, integrated GPS is more accurate and doesn’t rely on data. It’s also harder to steal and is covered under your vehicle warranty.
Its accuracy isn’t just a selling point; it can improve the resale value of your car. It’s also a great option if you need GPS while driving a lot.
In addition to GPS, autonomous cars use radar sensors, cameras and lidar (light detection and ranging) to localize their surroundings. These are used to find traffic lights, pedestrians and other vehicles that can help the driver navigate safely.
This information is then used to create a map of the environment. Most of these maps are sourced from Google, and can be augmented with photographs provided by the people who own or know about the area.
Research into how to ensure that the self-driving car’s GPS can stay connected with the various GPS and GNSS satellites is ongoing. This includes assessing the different inertial effects that affect integrity of the navigation system, such as coasting and Schuler feedback.
One technique for minimizing the effect of these inertial effects is to use multiple reference stations, which can be a cost-effective solution for improving accuracy. Several studies have shown that a single-station system can achieve centimeter-level accuracy when the baseline length is greater than 100 kilometers.
The accuracy of GPS can be improved by combining it with inertial sensor information, such as accelerometers and gyroscopes. INS/GNSS integration is commonly based on a Kalman filtering scheme to preserve the GPS integrity information.
Integrating a GPS and inertial sensor system requires an effective algorithm to calculate the optimal integration. This is a difficult problem because inertial systems are sensitive to the movement of their surrounding objects, such as other cars and pedestrians.
Research is focusing on how to improve this problem using the most advanced inertial sensors available, such as inertial navigation systems (INS). These systems are designed to predict the motion of a vehicle based on its own internal measurements, and integrate these into the GPS/GNSS navigation system to maximize accuracy.
ADAS
Advances in self-driving car tech are helping to reduce the number of road accidents and injuries. These safety systems use a combination of sensors, cameras and computer processors to provide information about the driving environment and act as an early warning system to alert drivers of potential dangers.
Advanced driver assistance systems (ADAS) can include features such as automatic emergency braking, pedestrian identification, surround vision, parking assist, lane-keeping and driver drowsiness detection. They also help drivers to avoid traffic jams, road closures and blockages, and give important traffic information to make driving more safe.
ADAS consists of a suite of sensors, interfaces, and a powerful computer processor that integrates all the data to make decisions in real time. The sensors include camera-based technology, as well as lidar, a scanning laser that creates a 3D image of the vehicle’s surroundings and can be used to determine whether objects are in a driver’s field of vision or not.
In addition, sensors collect data on traffic flow and road conditions and send it to the ADAS computer, which prioritizes the information. This allows the ADAS to issue alerts that warn of hazards such as vehicles in the wrong lane or traffic that is entering a driver’s path.
While ADAS can be a great safety feature, it still has many challenges. For example, drivers must have a clear understanding of what options are available and how to implement them properly. Additionally, these systems are expensive, with scalability concerns preventing them from being deployed on more factory-built vehicles.
The ADAS market represents a major opportunity for semiconductor companies, especially those that can improve system-level capabilities. These companies may need to change their business model, for example, by expanding into software and integration capabilities and by developing new strategies for working with OEMs and other players throughout the value chain.
Ultimately, ADAS offers manufacturers the chance to build safer cars and save lives by eliminating the majority of road accidents that could have been avoided by human drivers. This makes ADAS a critical part of the automotive industry’s future, particularly when autonomous vehicles are in development. The key is to find ways to increase ADAS adoption, which will require semiconductor companies to continue investing in new technologies and enhancing their existing ones.