Data visualization is an inevitable way for various systems in the future

In recent years, concepts such as digital twins, three-dimensional modeling, and data visualization have become popular, but in the past few years, many people think that these are seemingly lively displays and have no practical value.
However, on the other hand, as the cutting-edge technology of 3D modeling, SLAM (real-time positioning and map construction), it gathers the most advanced deep neural networks and the most advanced mathematical theories in the world, such as group theory, manifolds, etc. It has become a battlefield for fierce competition among major technology companies. So what is the use of SLAM?
You may be very surprised that the largest and most necessary application scenario for 3D modeling is the field of autonomous driving and robotics. I told this conclusion to many friends who studied science and engineering. They were all surprised. What is the relationship between unmanned robots and 3D modeling? A simple understanding is of course to prevent collisions, so if you install enough and good enough sensors, isn't it enough? In fact, I thought so a few years ago, but as smart devices such as unmanned vehicles and robots with large space and fast environment changes, simple sensors are far from enough.
Everyone knows that "people have no long-term worries, there must be near-worries." Unmanned machines cannot wait for the sensing signals in front of them to respond. Instead, they must make predictions in advance. At the same time, they must make predictions and real feedback. Make constant adjustments to achieve the desired goal.
Think about it carefully, all of our human intelligence activities are a process of prediction. The so-called prediction is the advancement of time and space and the variety and selection of plans. In order to achieve this goal, we humans have to spend decades learning the knowledge accumulated by our ancestors. The principle is to establish a mapping relationship between the spatial structure of individual life practice in the brain, which is similar to robot SLAM, and is the result of millions of years of human evolution.
The 2014 Nobel Prize in Physiology or Medicine was awarded to scientists such as John O'Keeffe who discovered cells in the brain's positioning system. They found that there are three types of cells in the hippocampus and adjacent brain regions, namely, place cells, head direction cells and grid cells. These three types of cells constitute the navigation in the brain. system.
After these cells receive spatial information from various sources, they process the information to form a cognitive map in the hippocampus to form a permanent memory of the spatial location.
Maybe many people still don't understand, why do we have to build a spatial model in the brain first, and directly use the currently seen (perceived) scene to make decisions? Of course it is possible, but you will have very little room for decision-making. For example, if a sweeping robot uses collision sensors to sweep the floor, although it can run, it has been in random collisions and can hardly do effective things. It's like asking you to go to a strange city to find a friend who has no specific address. You can only keep inquiring. If you don't analyze and record the result of the inquiry and decide the next step, you will never find a friend.
However, if the sweeping robot can model the space after several random collision explorations, then it can actively plan the sweeping route and analyze which places have not been scanned. In the case of insufficient power, it can follow the shortest path Locate the charging dock. Of course, in order to achieve these goals, it also benefits from the development of lidar and machine vision in recent years, as well as some awesome algorithms, such as Kalman filtering.
It can be seen that the three-dimensional modeling of environmental space is not just for good-looking, it contains very profound principles of human cognition. The seemingly random environmental information is built into a specific structured model within a three-dimensional a priori frame, and then based on the advance of the current information (visual information always arrives at the fastest speed), it can be in the space path or There is plenty of leeway in the choice of countermeasures to truly "work with ease."