The dream of fleets of self-driving cars efficiently taking us to our destinations has captured consumers’ imagination and fueled billions of dollars in investment. However, a fully autonomous vehicle needs to overcome significant challenges before it can hit the open road.
To achieve full autonomy, the sensors that “see” the road must be cost-effective and highly reliable. This means focusing on radar, cameras and LiDAR (using lasers to judge distances by measuring the time it takes for a signal to bounce back).
Autonomous cars use multiple sensors, onboard computing and advanced software to detect their surroundings and make decisions. This information is transmitted to other vehicles and infrastructure using low-latency connectivity.
AVs must be able to understand and interpret all of the data they receive in order to safely operate in unpredictable scenarios. They must also be able to communicate with each other and exchange this data.
In order to make this process as fast and economical as possible, ADAS developers will need to optimize their sensors, reduce the number of sensors and sensors’ failure modes, and provide high-performance computing capability. This will enable AVs to operate in all weather conditions and across a wide range of terrains.
It’s easy to imagine a future with breezy commutes where robot navigators make deadly crashes a thing of the past. But, a self-driving utopia is still years away. According to Google’s car project director Chris Urmson, it could be as many as 30 years before fully autonomous cars are the norm.
Autonomous vehicles rely on complex sensors that are situated throughout the vehicle to create and maintain a detailed map of their surroundings. Radar sensors track other vehicles, video cameras read road signs and traffic lights, and Lidar (light detection and ranging) sensors bounce pulses of light off the car’s surrounding environment to measure distances, detect lane markings and even sense uneven surfaces in roads and sidewalks.
The challenge for autonomous cars lies in figuring out how to respond to unexpected situations that might occur on the road. From navigating four-way stops to dealing with a cop waving cars around an accident scene, humans use a generalized common sense that is hard for robots to grasp.
While full societal benefits are still unknown, automated vehicles promise to offer numerous economic and efficiency gains. Eliminating human error — which accounts for 94% of car crashes — could reduce road costs significantly.
Software-defined autonomous systems will allow cars to adapt quickly to changing conditions. They can use machine learning to improve their performance and optimize sensor outputs. And they can communicate with one another to avoid accidents or to reroute around congestion.
Ride-sharing companies, such as Uber and Lyft, are already exploring driverless technology. But AVs will also have applications in freight, logistics and “last mile” delivery services.
Some of these AVs may be at SAE Level 2 autonomy, which allows the vehicle to manage functions like steering and acceleration but requires a human in the driver’s seat to take over at any time. However, there are many challenges ahead: regulation, rethinking highway infrastructure and the question of who is ultimately responsible for a crash when a driverless car is in control.
A major challenge will be ensuring that the technology can function properly in all environments and conditions, such as bad weather and traffic congestion. This requires a heavy reliance on sensors that can detect objects and determine the vehicle’s position with high accuracy.
Another critical issue is how self-driving cars will handle life-or-death decisions. Consider a situation in which a child darts into the street in front of a car. The vehicle must decide whether to swerve and risk passenger safety, or continue on and strike the child, possibly killing it.
Autonomous vehicles could also threaten the service-based economy, particularly in urban centres. People who depend on jobs such as taxi driving, food delivery and home health care could see their livelihoods threatened by a shift to autonomous vehicles. However, government officials can help accelerate the transition to autonomous vehicles by establishing clear guidelines for their testing and production. They can also encourage competition among states to find the best approaches to regulation.