Autonomous vehicles use a combination of sensors to monitor their surroundings. Radar and high-powered cameras detect road infrastructure, other vehicles, pedestrians and traffic signals. Lidar, which bounces pulses of laser light off the surrounding environment, paints a 3D picture of the vehicle’s surroundings.
Powerful computer systems process the sensory input and instantaneously decide how to navigate the car’s path, adjusting acceleration, braking and steering accordingly. Hard-coded rules, obstacle avoidance algorithms and predictive modeling help the software follow traffic laws and navigate obstacles.
Sensors
AVs need multiple types of sensors to detect and recognize objects around them. They include image, lidar, radar and ultrasonic sensors, each with different functions. Each sensor has its own limitations. Therefore, they are used in combination through sensor fusion to achieve safe autonomous driving.
Cameras are the most familiar sensor type in our cars today. They capture images or video of the environment and can be used to identify pedestrians, vehicles, traffic lights and signs. However, regular single-lens cameras only provide a 2D image, and cannot distinguish distances between objects.
Radar sensors are able to detect objects through electromagnetic waves and determine their speed and location. They can also help with blind spot monitoring, lane keeping and parking. However, radar systems can be obstructed by rain or snow. Therefore, it is important to understand the effects of weather on AV sensor performance so that they can be properly simulated in any hardware-in-the-loop or software-in-the-loop tests.
AI
AVs use AI to analyze and interpret data collected by sensors. This data is used to create detailed environmental maps that guide the vehicle’s actuators, which control acceleration and braking. The software also uses hard-coded rules, predictive modeling, and object recognition to follow traffic laws and avoid obstacles.
This information is compared to databases of known objects, such as pedestrians, vehicles, signs, and other road markings. AI systems are then able to recognize these objects instantly and respond accordingly.
Using advanced algorithms, these systems make instantaneous decisions on whether to slow down, swerve or accelerate as needed. They can even adjust the sensitivity of sensor systems to manage challenging conditions like poor weather or bumper-to-bumper traffic. They may also log relevant snapshots of sensor data after an incident to aid in accident investigations and insurance claims. They can also contribute to eco-friendliness by reducing fuel consumption through optimal driving behaviors and energy-saving acceleration. These features are important for lowering emissions of greenhouse gases and other pollutants.
Map Building
A vehicle’s localization capabilities are vital to autonomous driving, and the technology is more complex than many people realize. AVs need high-definition maps to understand their surroundings and interpret scenarios.
These maps are not the same as navigation maps that most drivers use in their cars, which rely on satellite signals. Instead, AVs need real-time HD maps with accurate information about road geometry and lane markings, as well as parking signs, fire hydrants, traffic lights and other landmarks.
Swift Navigation’s DRIVE Map uses data from millions of passenger vehicles to deliver high-precision positioning with two to four centimeters of accuracy. The solution also includes a six-axis inertial measurement unit, which helps overcome temporary loss of satellite signals caused by canyons and tunnels. It’s these technologies that will enable AVs to safely and efficiently operate on public roads. AVs are capable of dramatically improving safety, reducing ride-hailing and freight trucking costs and making our cities less congested.
Communication
Many of the technologies needed to develop autonomous vehicles require sophisticated communication systems to operate efficiently. This is because AVs cannot function in isolation; they need to exchange data with other vehicles, infrastructure and pedestrians to make informed decisions on the road.
In fact, a lack of efficient communication is largely to blame for some of the recent incidents involving self-driving cars, such as the fatal crash of a Tesla vehicle in March 2018. The driver failed to take action when the car detected a pedestrian crossing the roadway and instead kept accelerating, hitting the woman.
However, advanced communication technology can mitigate this issue, and it is rapidly evolving to bring us closer to a future where AVs can communicate seamlessly on the roads. From DSRC to C-V2X and 5G, these advances are improving the safety and efficiency of our future road transport system. By eliminating human error and reducing transportation costs, AVs promise to transform the way we move.