Unsolvable Challenges of Complete Safety in Self-driving cars
Self-driving cars may follow the rules of the road, but what about the language of driving? Computers and sensors in the vehicle can stop the process of subsequent car-based crashes.
Image: In cities and countries worldwide, drivers use a wide range of hand signals to communicate with other drivers.
Car buyers may not buy cars that can ignore their decisions.
But as a community linguist studying human-computer interaction. I began to think about how self-driving cars interact with the human drivers they encounter on the road.
Driving can involve a series of social signals and unspoken rules, some of which vary from country to country – even region or city. How are driverless cars able to navigate this complexity? Can they be programmed to do this?
What can Self-driving cars do?
Self-driving cars are equipped with LIDAR technology that creates a 360-degree view of the car’s surroundings.
We know that Self-driving cars are equipped with a technology called LIDAR that creates a 360-degree view of the car’s surroundings. Image sensors can interpret signs, lights, and lines. A separate radar detects objects, while a computer has all this information and mapping data to guide the machine.
Although self-propelled vehicles will ideally be able to “talk” to each other for smoother navigation and reduced crashes, the technology is still in its infancy.
When the car and the driver have different intentions, which one should ultimately control?
But any self-propelled vehicle must also interact with classic cars and their drivers. As well as pedestrians, cyclists, and emergencies such as lane closures, passive stop lights, emergency vehicles, and accidents.
Sophisticated driving language
This is where things get tricky.
For example, if you are driving and passing a speedometer, you may use your lights to signal drivers coming in the other direction that they are aware of the speed limiter. But flashing lights in situations where it is not clear what the choice is can also mean “your beams are too bright,” “forgot to turn on your headlights,” or “go ahead.”
To interpret meaning, one considers the context: time of day, type of road, weather. But how does an autonomous vehicle react?
Other forms of communication help us navigate, from loud noises and sirens to hand signals and even bumper stickers.
It is relatively easy for a malicious driver to set aside safety features.
Of course, humans use a variety of hand movements – showing the shaking of the hand in front of a moving car, which indicates that the driver of the rear car should slow down and even point a finger angrily.
Voices can express love, anger, arrival, departure, warnings, and more. Drivers can express their complete dissatisfaction with a hard and continuous blow on the horn. And, of course, emergency sirens encourage drivers to clear the way.
But the specific meaning can vary depending on the region or country. For example, a few years ago, Public Radio International aired a story about the horned language in Cairo, Egypt, originally “spoken” by men. These horns can have complex structures. For example, four beeps followed by a long beep means “open your eyes” to warn someone who is not paying attention.
Image: Traffic travels from Tahrir Square in Cairo, Egypt.
What if the connection is down?
Vehicles can be taught to “read” road signs, so they can probably recognize common warnings on bumpers.
However, navigating construction sites or crash scenes may require instructions from a human not to be planned. This creates a great opportunity for error. Because hand signals vary greatly from region to region (and even from person to person), self-propelled vehicles may not detect the motion signal or, more catastrophically. May mistakenly point a hand gesture toward Follow an obstacle or another car.
This requires reflection: How much knowledge about our social and linguistic values is embedded in the system? How can driverless cars interpret manual and audible signals?
Warning systems can warn the driver that the car is leaving the lane. Now either towards the oncoming traffic or perhaps towards the off-road.
Other forms of communication help us navigate, from loud noises and sirens to hand signals and even bumper stickers.
Image: Existing cars can be stopped when identifying pedestrians
Cities are trying to protect crowded pedestrian areas and public events following car and truck attacks worldwide, most recently in New York City. It is very difficult to prevent vehicles as weapons, but technology can help you in this regard.
It is very difficult to prevent vehicles as weapons, but technology can help you in this regard.
Currently, cities are trying to determine where and how to place statues, barbed wire nets, and other barriers to protect the population. Police departments are trying to gather better information about potential threats. And train officers to respond – while ordinary people seek advice to survive vehicle attacks.
These solutions are not enough: it is unusual to place physical barriers everywhere. And it is impossible to prevent a vehicle from being picked up by potential attackers.
As researchers in self-driving vehicle technologies, we see potential solutions exist and are embedded in many road vehicles today. However, there are ethical questions about who should control the vehicle – the driver behind the wheel or the computer system that understands the potential danger of human actions.
How do self-propelled cars communicate with the human drivers they encounter on the road?
A computer solution
About three-quarters of the cars and trucks surveyed by Consumer Reports in 2017 have forward collision detection as a standard or optional feature. These vehicles can detect obstacles – including pedestrians – and avoid or stop colliding with them. By 2022, emergency braking will be mandatory on all vehicles sold in the United States.
Safety features of today’s cars include lane departure warnings, adaptive cruise control, and various types of collision avoidance. These systems include various sensors such as radars and cameras that track the car’s going on.
They work passively, not communicating with the driver or controlling the car most of the time. But when certain events occur – such as approaching a pedestrian or an obstacle.
Cities (and countries) have different social, linguistic signs.
Warning systems can sound to warn the driver that the car is derailing. Now either towards the oncoming traffic or perhaps toward the road itself.
They can even control the car and adjust the speed to maintain a safe distance from the car in front. And collision avoidance systems have a variety of capabilities, including audible alerts that require the driver to respond, automatic emergency braking, and even steering the car out of danger.
Existing systems can detect the danger and whether it is horned by the car or not (or whether the car is horned by it). Reinforcement of these systems can prevent various driving behaviors that are commonly used during attacks but are not used in safe vehicle operations.
Sensors and systems in current cars can detect obstacles and prevent collisions.
Avoid collisions
An ordinary driver seeks to avoid obstacles, especially pedestrians. A driver who uses a car as a weapon does the opposite and targets people. Conventional car collision avoidance systems usually operate by warning the driver. And then only at the last minute, they take control and apply the brakes.
A person trying to attack a car may try to disable the electronics associated with that system. It is difficult to defend against the physical change of car safety equipment. Still, if it is revealed that the car has been tampered with, the manufacturers can prevent the car from starting or limit the speed and distance it can travel.
However, it is relatively easy for a destructive driver to put aside the safety features. It is assumed that the car is working properly for many vehicles if the driver is actively driving or using the brake and accelerator pedals. Is controlled. In such cases, the safety systems do not act to hit the brakes.
These sensors and systems can detect what is in front of them, which helps make better-informed decisions. The system can be programmed to ignore the driver when humans are on the road to protect pedestrians from vehicle attacks. Existing technology can do this, but it is not currently used.
A question about control
This begs the difficult question: when the car and the driver have different intentions. Which one should ultimately be in control? A system designed to prevent vehicles from hitting the crowd can cause drivers in parades if they mistakenly feel that spectators or other marchers are in danger.
It can also prevent the release of a car surrounded by protesters or attackers. And military vehicles, police, and emergency response should usually operate in or near the population.
Vehicles can be taught to “read” road signs.
Conclusion
Balancing car and human control involve more than public policy and corporate planning. Car buyers may not buy cars that can ignore their decisions. Many AI developers are also concerned about malfunctions. ٍEspecially in systems that work in the real physical world and can ignore human instructions.