Spherical stealth

立体环场空间状态扫描采主动式射频回波测距在军用上会暴露行踪、在民用上亦易造成干扰；有一种演算法可以从多角度被动式摄得的平面影像合成立体模型可是运算量极为庞大。道路上车辆颇多，有车联网共享，多角度影像是随时大量在取得，若各车备多些运算资源就可以分担中心的运算。车多时倾向多车组成的 near cloud 而疏解与中心的通讯负担，车少时附近运算资源不足而远端通讯带宽反就充足所以倾向 far cloud，如此即需要底层分布式操作系统自动调节运算分布方为完整的 Cloud。

Virtual Reality 和 Reality 是不是同一回事？

在于 Spatial Mapping 的精准度与解析度，只要夠接近，基本上在 VR 里 run 的就是在 reality 实际 run 的。如外在世界不到脑里形成不失真的 VR，大脑又何从思考做出判断以传令予肢体应对 reality ？对 reality 有准确的认识，在脑里 VR 演练过多回，到 reality 实行时方能一次做对。在 VR 里失误不会有损失，在 reality 失误就是实质的损失。

智能行驶

Spatial mapping 是 Google car 首先带到民车的，之前是 DARPA 在 tank 上面搞，tank 才是在人车稀少的野地测试，Google car 是在公路上跑有几年时间 0 事故，那时我还很呕说我光是 VR 模拟就频频当机，他们是怎么办到的？智能车开始出现事故是这一两年的事，看到有一辆直直往分隔岛冲撞去，很惊讶那么明显的障碍物怎么可能没侦测到？得知是把分隔岛另一边道路的右边地标线当成是自己车道的左边地标线，这不是早在世纪初还没有 spatial mapping 以前就开始在公路测试的平面影像辨识与移动追踪吗？现在不是已经多了 spatial mapping 吗？就算看错了地标线怎么会不知道前方有障碍物？火场寻径与道路寻径是通用同套，管它是火场还是人车都是禁止穿越过的障碍物，这还不需用到类神经网络，这是不经过思考几近反射的简单立即判断，除非是同时出现多个多向快速移动中的障碍物才伤脑筋。在 VR game 里都 run 到驾轻就熟很平常的动作，在 reality 意外是怎么发生的？确定不是传感器故障或处理器当机？

AI 寒冬？明明现在我就热到要吹冷气降温

从上世纪末到这世纪初科技界所谓的 AI 寒冬是指 NN 克服不了技术瓶颈表现差强人意沈寂到了近年才在技术上取得突破，而今年非技术人员所谓再一次 AI 寒冬算是属于达康类型的商界情景。当前并非所有 AI 都用 NN 为佳如同人脑亦是有些事用精准逻辑表现较好、有些事用模糊感觉表现较好而采另一套反而表现较差，早就说过要双管齐下同时进行而何者临场结果可接受即采何者，所以不要单方面看到某些时候 NN表 现不如预期就开始在说什么泡沫寒冬。自动驾驶开始出现事故当然是要去改进而非因噎废食，试问在 AI 以前有因为交通事故频传就都禁车而不改进车体的机构安全吗？从来想不劳而获的都是搞金融的，泡沫就只有股市商式炒作在泡沫，技术是不能不持续改进，为的是实用。

Augmented Reality

VR = R ➡ VR + VR objects = AR 

Patterns ↔ Senses

试想，若你的听觉受器是如视觉受器一般是阵列收讯，声讯对你来说是不是如同视讯画面一般？如此你会直觉两种收讯是不同的处里吗？是刚好你天生听觉受器就仅有两个单点，才会直觉视与听不同，在你进化前暂把声讯当成是1x1的2D画面串流，那直觉就会是一样的处理了。

All kinds of signals are patterns, the patterns are resolved to be the senses, and the senses are patterns too.

传统处理器在理解文字是一维循序串列处理，现在将每个字当做一个像素的值看待，不就是一张二维图像？那你觉得可不可以平行并列处理？整个理解过程是整面 operators（ 动词、介系词、连接词... ） 结合旁边的 operands（ 名词、冠词、形容词、副词... ）同时在发生，整个画面感觉就像粒子接触结合的化学反应一般组成了 Sense structure 以于脑内进行思维。

同理逆推，由 Sense structure 合成 pattern 输出是咋样的情形可想而知。

一个让我头抬起来手空出来的头免低手免持新SEN品

现在对着 bar code 付款，我也产生 bar code 让人现场付款给我，这是目前暂时的、过渡性的指向式支付方式，因目前行动装置的声波或磁波 I/O 仍无具阵列孔径合成聚焦之方向性故仅能使用摄像与显像进行可判方向的光影通讯，唯 bar code 必由卖方而非买方提供因提供者可能被任一方向读取而读取者可自主限读特定方向。若以非由己端决定（GPS + g-sensor / 陀螺仪 + 电子罗盘）而由多方共同决定的方位的网通定位定向技术来决定对何支付（因由单方决定可能故意提供错误方位以诈欺），就不用 bar code 了。那是一个什么样的世界？透过眼机，看着何者即对何者付款，每个实体世界中的人或物都被计价与记载拥者于云端：＂请看着我支付，我就是你的。＂无论线上或线下交易都是透过眼机，眼机外面用外显除可显像表达外其内幕亦可不被外人所窥视，这点在线下现场交易输入密码时格外重要。线下和线上的差别仅是在完整对应实体世界的虚拟仿真实境中位置的不同，除了从原地空间跳跃到远处去或者 zoom in 把远处拉近，其余没什么不同。

无论临机或远端操作物联网家电（其实就是15年前的资讯家电、信息家电）都是透过眼机。信息家电无需复杂的固定操作面板而由触控面板取代，又物联网家电需同时满足临机与远端操作，与其临机与远端各一操作界面机不如就让操作界面机跟随人而不跟随家电如此就仅需一机。

Say hello to Project Tango!

The drones vs. the hackers

The drones don't only use the GPS signals to locate themselves but also use the multiple environment data around them to search in the virtual reality to locate, so the hackers can't cheat them by emitting the interference wave. However, the hackers still may palm off the fake instruction to the drones.

The completely bionic artificial neural network

The neural signal scanner is harder to implement than the neural network scanner. The scanner is one, and the other one is the container. No matter how the RD progress of the scanner is, the container is easier to implement. We can use the current virtual reality to record the future scanned 3D image and then simulate the real work of the neural system with slower performance. However, if we want it to run faster, we must make the scanned neural network to be the hardware. In addition to that, the way to have virtual reality in the completely bionic artificial neural network is different from the current processor system, and it will run completely the same as the neural system in our bodies. 

The question is: even one day the scanning technology reaches the resolution of the neural ion stream, how to generalize the formula for simulating out the next state from a series of the scanned states and how to be sure it is working the same as the original brain? 

Record the entire brain map at as much spatiotemporal resolution as possible and for as long as possible. This is the most popular way to commemorate life in contemporary times, just store it in the cloud first. For a large number of continuous brain maps, perform a regression analysis to find out formulas or train the neural network to generate the next time point of the brain map, and then compare it with the original brain maps. This will go through a lengthy iterative process until the simulated brain is fairly consistent with the original brain map record. With the development of science and technology in the times, more people of later generations have more detailed records of spatiotemporal resolution, and simulation technology is also constantly improving. But this is a kind of clone, and the original will be still gone. 

The Natural User Interface

The NUI (Natural User Interface) for the 2D GUI must recognize and trace the captured 3D image to transform to the UI event(s) first, but I don't care about that because I only do the 3D VR/AR UI. The NUI in the 3D VR/AR UI just directly maps the captured 3D image to the virtual reality to become the model added in it then lets themselves interact, so it is not necessary to have the event processing. In the 3D AR/VR UI, the pattern recognition and trace only occurs in the 3D AR/VR space.

The Really Real Virtual Reality: virtual-real synchronization system

VR = R ➡️ VR + VR objects = AR 

The ARDOM ( Augmented Reality Distributed On Mobiles ) maps each minute of the real world by the moving of a large number of users on the mobile network cooperating to scan the real world into the distributed virtual reality with the radar or lidar mobile devices: correspond each pixel from the captured image to the position of ( the located coordinate + the 3D vector * the radar or lidar distance ) in the 3+1D space. The 4D space will also include our bodies and even our neural signals ( our souls ) so welcome to 《The Matrix》.  The objects with the access permissions in the augmented reality are not all stored in one place, they are distributed and are cached only as anyone possibly needs them to present in the augmented reality. Everything occurring in the real world is scanned into the virtual reality; everything happening in the virtual reality could change the real world too but not must be immediately. It could simulate until making sure everything is ok and there is no problem, then let the drones update the real world according to the new right virtual reality. 

The augmented reality does not only embed some virtual objects in the real scanned scenes but also changes the real world according to some of the virtual ones by the drones. The viewers watch the virtual reality from the Cloud directly but do not change it, they change the actors in the real world and then the sensors in the actors update the virtual reality in the Cloud. In the pure simulation mode, the actors with sensors are replaced by virtual ones in the virtual reality in the Cloud. However, the Cloud is not servers, the Cloud is the distribution. The actors in the real world are changed by accessing the Cloud and they are the synchronous parts of the Cloud distribution too. 

The most basic virtual-real synchronization does not update actions from users both to VR and R at the same time, only syncs to R, and then updates VR directly by the change of R. Because there may be exceptions of failure to update to R, and then VR must be restored to before the update.

However, some objects existing in VR but not in R also need to pretend to exist in R, so it is necessary to move objects in VR first to know how to move in R. But, in the process of updating actions from users to VR and then VR to R, it may also encounter objects that have not yet had time to be updated from R to VR. So, in the process of updating actions from users to VR and R at this same time, if one of them meets an exception, it is inevitable to restore both to before actions.

As for switching to the simulation mode decoupled from reality, just turn off the sensing and actuation to reality. 

The NUI (Natural User Interface) for the 2D GUI must recognize and trace the captured 3D image to transform to the UI event(s) first, but I don't care about that because I only do the 3D VR/AR UI. The NUI in the 3D VR/AR UI just directly maps the captured 3D image to the virtual reality to become the model added in it and then lets themselves interact, so it is not necessary to have the event processing. In the 3D VR/AR UI, the pattern recognition and trace only occur in the 3D VR/AR space. 

The virtual-real synchronization system allows you to interact with physical objects anywhere worldwide as long as you have access permission. The ultimate realm is the whole screen presenting the scene of the soul out of the body instantly moving to the other side of the world or the perspective of God, and you can watch and interact with any place in the world through the screen in one place, which is equivalent to the reality and the virtual world are always updated in both directions. 

3D LiDAR Technology

SpaceTop 3D interface lets you reach inside your computer screen

The dynamic network of the mobile spheres

Give you the limitless number of spheres fully with the directional signal cells on their surfaces, please use them to design a network. Further, if the spheres are moving dynamically, how do you make the network? Please design a mechanism for the dynamic network. This may be for the communication among planets far out, and the same architecture also be on a planet or in space nearly. If we cut the sphere in half, it becomes a hat. Establish the network among these ever moving half sphere hats without any base station, so the hat itself is the ever-moving base station in the enemy territory, and must only use the directional signals away from the enemy detection. The same mechanism also can be applied to the communication among the planets in the outer space. There is no way to have a fixed base station too because the planets are always in revolution and rotation. It can not use the nondirectional signal too because they are the big planets. 

The location-based networking: Everyone owns the relative coordinates map created by scanning the neighborhood and merges another far areas from the far ones through the near ones. The more memories one owns; the bigger map one owns. As one wants to transfer something to the other one, decides the path according to the relative coordinates and their states on the map. 

定位方法

由一個天線發出一段訊號，被另一個天線收到立即發出"已收到"之訊號， 然後原發訊號天線收到"已收到"訊號，期間所花總total時間稱之"echo time"，( echo time * 已知平均傳輸速率 ) / 2 即估算出兩個天線之間的距離。

3個不全在同一直線上已知位置的天線，分別以echo所得距離為半徑畫圓圈，3個圓圈交叉得一點，即為第4個未知天線的二維空間位置；4個不全在同一平面上已知位置的天線，分別以以echo所得距離為幅圓成球殼，4個球面相交得一點，即為第5個未知天線的三維空間位置。

定位2Dspace中的未知位置天線需要3個已知位置不全在同一直線上的天線；定位3Dspace中的未知位置天線需要4個已知位置不全在同一平面上的天線。若少一個已知位置天線，就只能獲得兩個正相反方向的點，如果電磁波不是正負交流就可以從天線電流方向判別電磁波磁場方向進而由兩個點中擇出一個來，前提是必須已知欲定位天線電流方向而且方向固定非為正負交流。

但上述方法都已經造成通訊，如何在不讓被定位天線知情的情況下定位之?

將3個不全在同一直線上已知位置的天線所收到之data sequence做位移比對，確認3個天線收波的順序及時間差即可計算定出二維空間中非指向性輻射波的來源方向；同樣地，4個不全在同一平面上已知位置的天線即可定出三維空間中非指向性輻射波的來源方向。再多加一個已知位置的天線，比對其它任兩(三)個不在同一直線(同一平面)上的天線收到的data sequence，即可定出另一個2D(3D)方向，藉由這兩個方向軸夾角與已知位置天線之間的距離透過三角函數計算出與未知位置天線的距離即可定位。

反過來，在一個空域彼此間僅有指向性通訊的天線組也可藉由相同的方法定位空域中一個發射非指向性輻射波的天線。

最後，如何在自己不發波的情況下藉由週遭的發波天線定位自己?