October 21 news, according to foreign media reports, the next time you drive or walk on the street, stop and think about how you understand the surrounding environment. You should pay special attention to the children playing soccer on the lawn of the front yard and the cyclist who looks a bit unstable and nervous. You can't be distracted by women walking on the streets, although she is walking towards a group of friends waving to her on the sidewalk.
You need a lot of social and cultural experience to do these things, but you almost inconvenience you can instinctively draw conclusions. But imagine that if you are a driverless car and want to do the same thing, but there is no accumulated knowledge or shared humanity, they can allow you to interpret clues from the subtle behavior of others. Treating each pedestrian, cyclist, and vehicle as obstacles allows you to avoid hitting anything, but it is also easy for you to be isolated.
Anca Dragan, an automation expert in the department of electrical engineering and computer science at the University of California, Berkeley, said: “We call it the problem of frozen robots. Anything the car can do is too dangerous, because only the most Bad human behavior can lead to collisions."
Researchers such as Dragan are addressing the challenge of interpreting and predicting human behavior in order to make driverless cars safer, more efficient, and more confident. After all, if every machine had to stop for everyone with unpredictable behavior, we would soon be blocked by millions of terrified robots.
To prevent traffic jams, these researchers are relying on artificial intelligence (AI) and the ability to teach the driving system by modeling and repeatedly observing what these actions mean and how the system should respond to them. In fact, driverless cars should first realize that humans are not an obstacle. Dragan said: “Different from the way the tumbleweed moves under wind power, human movement depends on their decision. When they want to do something, they will take action to realize. The first thing we think about is based on them. The actions taken so far to infer what people want to do, so from this perspective, their behavior is rational."
For example, a driver accelerated in the right lane of the highway. The computer knows that people should slow down as they approach the exit, and it can be inferred that this person is likely to continue to move forward instead of exiting at the forthcoming exit. This is a basic example: When computers can estimate what people want to do and how they can achieve their goals, they can reasonably predict what people will do next and respond accordingly.
The key is that even machine learning cannot be limited by the individual elements of the scene. Melissa Cefkin, an anthropologist at Nissan’s Silicon Valley R&D Center, said: “Making progress in this area is very important, but the machine only sees the situation on the road. As humans, we are better at identifying There are specific behaviors that seem to be the same thing on the machine, but from the perspective of our society, it is another matter."
Imagine that when you drive along a city block, you see a person walking toward the curb. The robot driver may calculate his speed and trajectory, make sure he may cross the road, and then stop to avoid hitting him. But you see him holding the car key, realizing that he is walking to the side of the street and coming to the car door he stopped at the side of the road. You will certainly slow down, but there is no need to stop. Saifkin said: "The way people move in a particular environment has been coded culturally and socially. This is not always a human-human interaction, but a human-object interaction."
This is still a very simple example. Serfogin pointed out what she called the “multi-factor problemâ€, that is, the reaction of pedestrians and other drivers to people around them. She said: "If a pedestrian wants to pass in front of me, but does not look at me, they are likely to cause traffic jams. So now I want to know is whether it is safe to continue driving based on other traffic conditions. â€
It seems that the world seems to be moving in a direction of "driving hell," but you don't need to worry. Sevkin said that it is difficult to teach AI-based unmanned systems to adapt to human eccentric behavior, but it is by no means impossible. Dariu Gavrila is studying smart cars at Delft University of Technology, training computers to deal with various challenges, including through complex intersections (there are various moving targets, road debris, traffic police) and other unusual Things, such as someone in the middle of the street cart. Gavrila said that his goal is to develop a more adaptable driving style for the machine, thereby enhancing the community's acceptance of new hardware.
This work means that pedestrian traffic factors in the environment need to be taken into account, such as approaching roadsides, driveways or entrances to public buildings, and other activities. It also extends to everyone's actions, such as what a person's head looks in one direction and their torso pointing in another direction. Gavrila said: “Identifying the intentions of pedestrians may be a way to save lives. We demonstrated in the real car demo that the unmanned system can react one second faster than humans and that it will not generate false alarms.â€
However, there are limits to what computers can do. Gavril pointed out: "This is not a minority report. No one will tell you what is going to happen. As the level of forecasting increases, the uncertainty of future pedestrians or cyclists also increases rapidly. How many seconds will we expect in the future? Try modeling. The basic behavioral model is no longer useful after one second. A more complex behavioral model may allow us to have a predictive power of two seconds."
Despite this, a one or two second advance warning may be needed for a computerized system because it is still within the time range of human response. But other automation experts believe that we may make the machine fall into an “over-thinking†state every microsecond of driving. Jack Weast, chief systems architect for Intel’s driverless system, said: “When you are trying to predict the future, this is a huge computing task, and of course it only generates a possibility of guesswork. So instead of putting a supercomputer in every car, it's better to think about how to make sure the car never hits anyone. It's a more economical way of doing things."
There is another problem here. The ideal robot can not only understand the surrounding environment, but also understand how it changes the scene. Dragan said that many robot systems have inherent flaws: their manufacturers believe that the presence of driverless cars will not change the behavior of other "actors." But Dragan believes: "Regardless of whether we like it or not, the behavior of driverless cars can affect human behavior. Cars need to start thinking about this effect."
This is why Dragan and his team built a new system, and their model includes the human driver's response to the car. She said: "Our car is no longer super-defensive because it knows that it can also trigger people's reactions. Just as our vehicle merges in front of other vehicles, the latter will slow down. We also use the People are driving cars to actively estimate human intent."
This confidence training may be the key to future traffic. The key to building a robotic car is that it should not only possess human-like awareness, but should also be given human-like capabilities.
SMT Feeder types,
1. different smt machines with different smt feeders, but different model of the same brand machine can use the same feeder,like Panasonic Feeder.
2. Base on the size and types of components, smt feeder could be divided into three types : Tube feeder; Tray Feeder; Bulk Feeder.
Tape feeders with the different size such as 8mm, 16mm, 24mm, 32mm, 44mm, 56mm etc.
3. Base on the feeder condition, smt feeder also could be divided into four types: original new SMT feeder parts, used original SMT feeder parts, imitation/copy new SMT feeder parts, imitation/copy used SMT feeder parts,
Original feeder is made by original smt machine production manufacturers. Due to the large demand for smt feeder, currently there are many copy new feeders parts. Our feeders quality is also very great and looking forward to your cooperation!
Feeder Parts,Smt Feeder Parts,Panasonic Feeder Parts,Parts For Panasonic Feeder
Shenzhen Keith Electronic Equipment Co., Ltd. , https://www.aismtks.com