Emerging pedestrian wearable technology will enhance human driving cars and... [+] It also applies to the "see me" aspect of self-driving cars.

Pedestrians lead a difficult life.

When you try to cross the road, on the surface, you are taking your own life in your own hands, because you dare to travel before the raging car traffic. Even when you use the crosswalk legally and a large number of cars stand still before a red light, the driver may still step on the accelerator by mistake and suddenly accelerate forward toward you.

Always pay attention to those cars and their drivers that are temporarily stationary, just in case you can miraculously discern the impending sudden movement and possibly give way.

The most frightening situation is that there are no traffic signals at the crosswalk. These uncontrolled wildcard situations are a waiting disaster.

Will the upcoming car stop?

Once they stop, will they stay still until I cross the street?

We teach children to observe in both directions. This includes wise advice to cross only in safe places. Some young people seem to think that if there is a crosswalk, it must be a safe place to cross the road. Perhaps these children think that the faded paint lines on the streets can form a veritable protective barrier that will not be broken by moving cars.

Of course, not all pedestrians must cross the road in a legal or cautious manner.

Jaywalkers make a rather risky decision by crossing the road at any intersection on the street, at their own risk, and believe that they know best where to cross the road correctly. Sometimes people who cross the road indiscriminately do not notice the oncoming traffic, putting themselves in a terrible situation. In other cases, people who cross the road will observe the traffic a lot, but rely on their skillful dancing and twisting skills to try to escape like a frog. Sadly, no matter which method they choose to use, they are likely to lose the game.

And losing the game of being a pedestrian is very expensive.

Some pedestrians seem to think that they will win the race against the car and its driver, which seems strange. A car driver can use a multi-ton vehicle, and if you compare the fragile human body with the impact and destructive power of a moving car, he will undoubtedly be the winner.

Without arguing, humans will lose and cars will win.

I am not saying that this car is right. The point is, if you weigh the physics and forces involved and only consider what happens when a person is hit by a car, the car will exert a huge adverse force on pedestrians. Unless you believe in comic books and superpowers, there is the possibility of humans reaching out to stop the car, while humans are still standing and happy, well, keep on dreaming.

However, every day, many pedestrians dare to let cars come to pick them up. A pedestrian walked into the street and stared at a car driver. The car driver must decide whether to slow down and possibly stop to let pedestrians cross the road. Similar to the two gunners, they entered a staring game.

In terms of armed methods, they are unequal gunmen.

Pedestrians think they are flexible, and if the situation requires, they can probably escape death by rotating. Perhaps the more important tool that pedestrians have at their fingertips is that if a car drives forward and hits a pedestrian, the car driver will encounter a lot of trouble.

This is bad news for drivers. Drivers will have to repair the damage caused to their cars. There is no doubt that car insurance rates for drivers will rise. Drivers may be revoked or their driving licenses cancelled. The driver may face criminal charges.

The problem is that all those special factors that support pedestrians seem to be conceptual rather than particularly tangible at this critical moment. The reality on the ground in the middle of the street shows that when a person collides with a car, the person will be the most injured person. Of course, you can argue that the car driver may be in serious trouble, but that is long after the moment the pedestrian loses his life due to injury or destruction.

The result of pedestrians is either gnashing their teeth, or crossing the street unscathed, or they may be hit and severely damaged or killed.

The preferred method of these eye-to-eye matches may vary from region to region. Some people believe that in New York City, pedestrians must try to make eye contact with the driver, otherwise the pedestrian is basically sending a signal that the driver has the right of way and the pedestrian has left. At the same time, some people suggest that the best way in Boston is to deliberately avoid eye contact with the driver. If the pedestrian does make eye contact, it means that the pedestrian has now lost the strategy of crossing the road and the driver can move on.

Or maybe the opposite.

These rules about pedestrians are not very comforting. Suppose the driver is from a foreign country and does not know the expected appointment? Another possibility is that local experienced drivers decide to violate conventions at that specific date and time, so they will never "act in accordance with customary rules that pedestrians call sacred and inviolable in the area."

Many towns and cities have been taking active measures to try to stop the seemingly rising wave of collisions between cars and pedestrians. These municipal activities are often referred to as "Vision Zero" or similar catchy names. The idea is to remind everyone to pay attention to the precautions of pedestrians, hoping that the number of deaths from car collisions involving pedestrians will be reduced to zero.

It is easy to see that the problem has two components.

There are pedestrians and their desire to cross the road, and there are vehicles that use the street to travel and hope not to be disturbed or delayed. When the two meet, there is a high potential for fireworks.

You can try to make pedestrians pay more attention and move carefully when crossing the road. This is convenient, but not foolproof. People may not remember the thoughtful warnings posted on billboards and social media. Posting signs on the street is certainly useful, although the problem again needs to make pedestrians aware that the signs are there, let them read the signs, and then they need to pay attention to the signs in the throes of trying to cross the road.

Don't bet all your chips on this matter.

For car drivers, they may see countless signs and billboards about pedestrians and keep these warnings in mind. On the other hand, when you rush to work or try to go home to visit your children, the reality is that many drivers do not pay special attention to the needs of pedestrians. We all know that car traffic itself is a dog-biting world, and this feeling extends to the ongoing war between cars and pedestrians.

We seem to be caught in a dispute that we have not found a solution to this problem. Despite extensive attempts to inform and motivate pedestrians to be safer, and even though the same measures have been taken for car drivers, the sad truth is that the problem remains fundamentally. To clarify, we still need to try this kind of outreach, and it does make a difference, obviously, but the point here is that the chance of reaching the boasted zero is at best a long-term goal.

You may be wondering what other solutions are available.

Part of the problem involves ensuring that car drivers are aware of the presence of pedestrians at the moment when pedestrians and cars are about to intersect in the so-called space-time continuum.

It may be difficult for pedestrians to see.

They may wear clothes that blend with popular driving scenes. As far as children are concerned, they are smaller and less visible. At night, pedestrians may be covered by darkness. There are countless ways for car drivers to be unable to spot pedestrians.

In theory, if the car driver knows that a pedestrian is there in advance, this will at least reduce the number of incidents involving the driver without his knowledge and therefore failing to react in time to avoid hitting the pedestrian. This is not a panacea, because the driver can choose not to avoid pedestrians, deliberately hit pedestrians and fully know they are going to do so. This is the other part of the problem, and it is also a variant that I will introduce in a later column.

In any case, there have been various methods trying to make pedestrians stand out so that car drivers can notice.

Passing guards wore brightly colored jackets. Some parents wear the same clothes when sending their children to school. Another possibility is to hold and wave a flag. You may have seen those triangular flags that fluoresce in orange and are mounted on long poles. Pedestrians holding the pole in an upright position will help to be found from a greater distance.

Some pedestrians avoid fairly obvious choices, especially because they don’t look cool or otherwise arouse the anger of cynics. A more prudent method is to install reflectors on your clothes, or display them on the hats or caps you wear. As a form of attracting the attention of car drivers, these seem to be more acceptable.

These are all helpful and of course encouraged as a means for car drivers to wake up from their zombie trance and become aware of the pedestrian in front of their car. Having said that, we may all agree that, unfortunately, this still does not guarantee that pedestrians will be seen (it increases the probability, but it does not increase the probability to 100%).

Shift gears and think about the future of automobiles that require the emergence of autonomous vehicles. For a truly self-driving car based on artificial intelligence, there will be no one on the steering wheel. The driving system based on artificial intelligence will be responsible for driving.

This raises the question of whether self-driving cars are better at detecting pedestrians than human drivers. Some people worry that self-driving cars may be worse at pedestrian detection. That would definitely be disturbing. We seem to hope and assume that self-driving cars will be at least as good as humans, and we are indeed happy that self-driving cars do better on the task of pedestrian detection.

Human drivers are easily distracted while driving. The emergence of ubiquitous smartphones has become a major interference factor when driving. Some authorities have pointed out that human drivers watching cat videos or attempting to text messages while driving may have contributed to the increase in pedestrian and car-related deaths. Humans also drive after drinking.

The driving system based on artificial intelligence will not drink and drive. The artificial intelligence driving system will not be as distracted to drive as humans lose their attention or choose to look away while driving. In this sense, it seems obvious and natural that self-driving cars are expected to do better at detecting pedestrians.

If self-driving cars can better detect pedestrians, this indicates that the number of fatal accidents and crashes related to pedestrian cars will decrease.

Some people wonder whether it is still necessary to wear brightly colored jackets and reflectors in the age of self-driving cars.

Can we put aside these designs and just assume that the artificial intelligence driving system can detect us correctly?

The short answer is no, you cannot give up those useful preventive measures. For security purposes, they are still useful and recommended.

Some people go further, claiming that you might want to add some new and additional methods so that self-driving cars are easy to spot. In other words, there are means to be discovered by human drivers, and this also applies to being discovered by self-driving cars. In addition, there are some methods that can be discovered by self-driving cars. These methods are specific to the nature of self-driving cars (may or may not be applicable to human drivers themselves).

This is an interesting question today: Will the market for pedestrian attire and devices that warn human car drivers be spurred to develop new products in response to the incorporation of truly self-driving cars on our roads?

Before we solve this problem, let us first clarify what I mean by truly self-driving cars.

Understand the level of self-driving cars

As a clarification, a true self-driving car is a car in which artificial intelligence drives the car completely on its own and without any human assistance during the driving task.

These driverless vehicles are considered to be level 4 and 5 (see my explanation in this link), while cars that require human drivers to share the driving effort are usually considered to be level 2 or 3. Shared driving tasks are described as semi-autonomous and usually contain various automatic add-ons called ADAS (Advanced Driver Assistance Systems).

There is no real level 5 self-driving car yet, and we don’t even know if this is possible or how long it will take to get there.

At the same time, Level 4 efforts are gradually gaining some traction through very narrow and selective public road tests, although there is controversy as to whether this test itself should be allowed (we are all experimental guinea pigs that happen on our highways. Highways and trails, some people think, please refer to my report in this link).

Since semi-autonomous cars require human drivers, the adoption of these types of cars will not be significantly different from driving traditional cars, so there is not much new content on this topic to introduce them (however, as you will see a Yes, the points raised next are generally applicable).

For semi-autonomous vehicles, it is important to remind the public of a disturbing aspect that has recently emerged, that is, even though human drivers continue to post videos of themselves falling asleep on the steering wheel of level 2 or level 3 cars, we all need to avoid being misled. , It is believed that the driver can divert attention from the driving task when driving a semi-autonomous car.

No matter how much automation is invested in Level 2 or Level 3, you are the party responsible for the driving behavior of the vehicle.

For level 4 and level 5 truly self-driving cars, there will be no human drivers involved in driving tasks.

All occupants will be passengers.

Artificial intelligence is driving.

For the artificial intelligence driving system, it is somewhat simple to detect other cars near the self-driving car. Usually other cars are relatively large. Autonomous cars can use sensor devices such as on-board cameras, radars, lidars, and ultrasonic units to almost determine that the car is in a driving scene. interest.

Pedestrians are another beast altogether.

Trying to detect pedestrians can be tricky.

As I discussed before, for example, some AI driving systems sometimes classify snowmen as human pedestrians (see my analysis in this link). The point is that pedestrians are usually not as easily spotted as cars, but we would definitely expect that artificial intelligence driving systems will treat pedestrians with the same care, respect, and concentration as they do with nearby cars (maybe that’s a somewhat controversial proposition, but you get the idea Essentials).

The main means of detecting pedestrians involves cameras in self-driving cars.

Streaming video is being collected while the vehicle is driving. The algorithm examines the video image and tries to identify whether there are any spots in the image that look like pedestrians. Pedestrians may be standing behind light poles or other obstacles, so the image may only partially indicate the presence of pedestrians. For computer-based image processing, it can be a daunting task to perform a Where's Waldo inspection on the video to determine where pedestrians may be in cluttered objects in a busy street scene.

A child may stand behind an adult, and the vantage point of the oncoming self-driving car and its camera may not be visible at all.

Imagine that a child chooses to suddenly and unexpectedly dash from behind an adult and directly into the middle of the street. Since the artificial intelligence driving system does not detect the presence of the child earlier (hidden behind the adult), this means that the presence of the child will be detected later in the game, so to speak. This in turn may reduce the viable options of trying to stop instead of having to turn away from the child who is now at an alarming distance.

Would wearing a traditional fluorescent vest or jacket help (the kind used to remind human drivers)?

How about the bright orange flag hanging on a tall pole?

Yes, these are still good.

Of course, it is assumed that the image processing algorithm being used takes this into account.

Anything that can visually attract the eye of a human driver may be beneficial to the use of autonomous vehicles. The key is that automakers and self-driving car developers need to ensure that they integrate these cues into the image analysis and pedestrian detection programs they are using.

This includes the use of machine learning (ML) and deep learning (DL). To take advantage of today's ML/DL, you need to train computational pattern matching algorithms through large amounts of data. Therefore, it is important to include images of pedestrians wearing "see me" clothing, which is used to attract the attention of human drivers. ML/DL pattern matching is expected to be effective in identifying those convenient clues that indicate the presence of pedestrians.

One difficulty with these pattern matching methods is that they can be fooled intentionally or unintentionally.

Suppose a fluorescent jacket is placed on the mannequin. The mannequin is placed on the sidewalk, possibly outside a shop selling clothes. It is conceivable that self-driving cars passing through that street might misunderstand the mannequin as a human pedestrian. why? One possibility is that pattern matching "learns" that as long as there is a fluorescent jacket on a character with the overall shape of the chest and arms, it means that someone is there and is therefore classified as a pedestrian.

Humans are unlikely to make such a classification error.

Of course, as mentioned earlier, humans make many other types of driving mistakes, so they won’t get a free pass just because they can usually tell the difference between a mannequin and a real person.

Someone suggested that maybe we should come up with a new symbol that can be standardized as the universal "I am a pedestrian" symbol, and then make sure that all autonomous vehicles are programmed to find that symbol. Then people will buy clothes sewn with the symbol, or carry a flag with the symbol. On the surface, this may seem like a brilliant idea, but it has many shortcomings and may not receive much attention (for a discussion of this issue, see my column).

Another big controversial topic concerns what happens when the AI ​​driving system does detect a pedestrian. As pointed out earlier, humans will struggle over whether cars continue to drive or pedestrians drive. Currently, the default setting of most artificial intelligence driving systems is for pedestrians to move on. Unfortunately, this has led some people to decide to use this default setting and cross the road at will (see my report again).

In any case, remember that artificial intelligence driving systems usually use more than just video to perceive driving scenes. In general, everyone is using radar, and almost everyone is using LIDAR (for an analysis of Tesla's decision not to use LIDAR, please refer to the link here).

Various research laboratories and entrepreneurs have been experimenting with specially adapted clothing and electronic equipment to assist pedestrian radar detection. Pedestrians will wear such clothing, such as modified jackets or coats, lined with materials and electronics to enhance the ongoing radar detection of any autonomous vehicles nearby.

In short, the radar sensor on the self-driving car is transmitting radar signals. These radar signals reach the objects in the driving scene and are reflected back to the radar unit. Based on flight time and other factors, radar tries to find the location of objects, their distance from the autonomous car, and the shape of the object (and the movement or speed of the object).

Radar does not always get good readings. There are many reasons why radar signals are blocked or distorted. In this case, if the pedestrian wears or holds something that can enhance or assist the radar signal, the chance of the radar detecting the presence of the pedestrian can be increased.

Sometimes, during driving, artificial intelligence driving systems often rely more or less on cameras and more or less on radar. All in all, artificial intelligence driving systems usually have a multi-sensor data fusion (MSDF) function, which involves trying to combine and coordinate the detection from each sensor on the vehicle.

Another method requires the use of a beacon device, which emits a signal to indicate the presence of a pedestrian. The beacon can be worn by pedestrians or hand-held. Clothing manufacturers may choose to sew such a beacon into the lining of a jacket, or attach the beacon to a hat or cap. Self-driving cars may try to detect the beacon signal, and then consider there may be pedestrians nearby accordingly.

Various issues regarding the use of pedestrian beacons remain unresolved. For example, the beacon just reminds pedestrians that they are within a certain range, or will they provide details such as where the pedestrian is and other parameters of the pedestrian? Will pedestrians inadvertently become bold enough to cross the road because they have a lighthouse? Are self-driving cars expected to provide confirmation responses to beacon wearers? Suppose someone places a beacon on a fire hydrant or light pole. Would this confuse self-driving cars and diminish the value of owning the beacon? etc.

In addition, a related question is whether the sensor kits of autonomous vehicles already have the facilities needed to detect beacons. If not, then there will be the problem of having to add additional sensory devices to the self-driving car, which may increase the costs associated with the self-driving car and create other accompanying difficulties.

Some people think that they may need to buy and wear some special clothes or have special equipment to warn in advance that there are pedestrians near the self-driving car.

Shockingly, the titles of these clickbait are shocking.

This is a typical response when you don’t look closely at what is actually proposed. Responsible researchers and entrepreneurs do not recommend that you have to use these products, just like wearing a fluorescent jacket or owning a bright orange flag. These are just useful ways to increase the chance of being detected.

Doomsdayists want to immediately condemn this showing that self-driving cars have inherent flaws and should not be driven on the road unless they can perfectly detect pedestrians at any time, any weather conditions, various driving scenarios, etc. Unrealistic and short-sighted expectations.

Some people still insist on using V2P (vehicle-to-pedestrian) electronic communication (this is a bit similar to the earlier discussion about the use of electronic beacons). The concept is that pedestrians holding smartphones can electronically communicate with nearby self-driving cars, so self-driving cars can also convey information to pedestrians. In this way, approaching self-driving cars may send messages to pedestrians on street corners, asking them to stay still and not enter the street. This may help, but it certainly won't be a panacea for some people to bet (as I described in my column).

The reality is that pedestrians and cars will still be mixed on our streets. Human-driven cars have the weaknesses of human drivers. Self-driving cars will be limited by the nature of the sensors used, the programmed AI functions, and driving scene conditions, as well as the nature of where pedestrians are, what they are doing, and what they intend to do.

My prediction is that these pedestrian enhancement products designed to help self-driving cars further detect will gradually and ruthlessly realize to provide value.

This is still a fairly nascent market.

There are not many autonomous vehicles involved in today's trials. Until there are a sufficient number of autonomous vehicles on the road, selling and buying such equipment is of little use. That being the case, we can also turn our attention to Level 2 and Level 3 cars. They may be equipped with the same types of sensor kits as self-driving cars. Therefore, the market for these pedestrian-oriented products may appear earlier than the long-awaited emergence of truly self-driving cars.

Doting parents would consider buying a fluorescent jacket with a special radar reflector array stitched inside. Will this help protect their children from the unpredictability of human drivers and the emergence of autonomous vehicles?

Some people assume that if these parents are buying smart socks for their babies (monitoring heart rate and oxygen levels, with Bluetooth connectivity) and smart diapers (I won’t say what they do, you might guess), they will definitely Want anything that can give their beloved toddler or child a clear advantage when crossing the road.

Remember my words, you will eventually see such walking wearables in large retailers and even those nearby mom-and-pop stores.