Jump to content

User:Liberuu

From Wikipedia, the free encyclopedia

Mais acrescento que quem tem o polegar da mão direita em cima, quando se cruza os dedos das mãos (ou seja os entrelaçamos alternadamente de maneira natural sem parecer que nos estamos a sentir mal por o fazer), são pessoas dadas a caminhos nas explicações de direcções... dizem algo como virar na segunda à esquerda, na quarta à direita, definindo um caminho contínuo ou analógico... são portanto pessoas típicamente "racionais"...

Quem tem o polegar da mão esquerda em cima, quando se cruza os dedos das mãos, são pessoas dadas a referências ou nomes nas explicações de direcções... dizem algo como virar depois do prédio amarelo, quando se chegar às bombas de gasolina virar à esuerda, etc... definindo um conjunto de referências para chegar ao destino tendo um pensar discreto... são portanto pessoas típicamente "emotivas"... estas pessoas têm o dom de atrvessar paredes em pensamento, converter agua em vinho e construir um predio pelo telhado (perdendo o chão pelo caminho)... pode ser um risco para todo o mundo se não forem chamados à atenção pois fazem coisaS impossiveis de concretizar...


Isto é válido tanto para as Mulheres como para os Homens...

Fazendo com que muitas vezes não tenhamos o mesmo tipo de discurso...


Esquerdo emotivo: O que importa pensar discreto atrai-se pela personalidade discurso com mais adjectivos Saltitam entre assuntos


Direito racional: mensagem a transmitir orientação contínua atrai-se principalmente pela beleza importa a qualidade poucos adjectivos, mais adverbios


PS... eu sou o palhaço que têm visto na internet...


Robot's Brain Architecture

by Paulo Roque Silva ISR-IST-UTL


Abstract

This architecture allows a robot to have full autonomy: of decision; as learning. In all its life and in any environment. Also doesn't depend on robot's body but once at work the body can't be replaced. And intends to be biological viable. The robot must have a motion feedback. And can have any other  sense.
In this architecture we don't tell to do. The breakthrough is: first, the robot acts then waits to “listen”. Also it builds a Past in a temporal line. The robot behavior is based on its Body, the Environment and its Past. The 

robot has emotions and is self motivated but with no freewill.

We can use this architecture in robot pets as in humanoids with more body power (strength, resistance and more acquired senses) to be used in extreme environments.

Lisbon, 22 of June of 2009


Patente de Invenção Nacional nº 104393 Nome: Paulo Jorge Roque Mendes da Silva BI: 6201426 Telemóvel: 91 912 15 70 email: sysliberal@gmail.com INPI patent : 104 393 (Portugal) a publicar 18 Agosto 2010


"Robot Brain Architecture"

Description

Introduction

Nowadays, we can see many robots simulating Human behavior... they show some kind of intelligence with some degree of freedom. Meanwhile this Robot Brain Architecture intends to go in the opposite direction of the state of the art in Artificial Intelligence. That's the Breakthrough...

In engineering we often use the "black box" method to find out how a device works. We do that in the following way: putting signals at the input of the device and analyzing what comes out from it. Then we can figure out how it works.

The architecture here presented is going precisely in that way. The intended system (robot's brain) begins to produce signals for the "world" (our Real world) and it learns with the results returned by the "world". It uses precisely the “black box” method for what surrounds it (body + "world")...

It is that learning that will be important in this system, because its goal is to match the input signals with the output ones...

Basic Module

The system architecture basic module can be analyzed in fig. 1. Here, we consider that each external event (instant sense captured from the "world") is registered in a single node, memory or “neuron”. Those events are being kept in a sequential time way, building a "Past" or a history of the robot.

Fig. 1

When a search of an event kept in the “Past” needs to be done, it will be done in a broadcasting way. The event that the system seeks, in the sequential memory, will be given to all the nodes as if a distribution of a newspaper would be done. Each node sees if it's own memorized event matches with the one delivered. In this case it returns the time or the number of itself.

Decision

Several numbers for the same event can be returned . So, a decision will be made from one of two ways, for which the following Step is going to be given or taken: the event in the most recent Past; or the event that has more number of retrievals...

General Architecture – Sense Identification

To build that Past we need a more enlarged architecture as in the fig. 2. Here the inputted events from the senses of the robot's body (as the vision, hearing, taste, smell and touch), and the proprioception sense (position of the elements of its own body... later its correspondent action), are being classified with an input number (id). It is like a Ticket or an Identity (id) number... That id is given by an identification node (in a Sense Identification module) using a credits learning (fig. 3).

Fig. 2

The identification node is linked to all the possible different signals that can occur from each sense. The Sense Identification module uses the basic module... Each input event is a subset of all signals of a sense that are active at that moment.

An identification node “fires” its own id, if the input event reaches a credit threshold. For instance, when the input event overcomes 80% of all credits (active signals or not). A credit is the number of times that a signal was detected by that node, in a “firing” occasion. Whenever a node "fires", updates the credits (plasticity). Increasing or reducing a credit value if the signal exists or not... The identification node id is sent to the Central Unit...

Fig. 3

Robot's Past

The Past is a straight temporal line of nodes (memories). It has registered a conjunction of all the senses ids at the same node. Those ids are compared with the ones foreseen from the Past (in the previous input events) and the difference is what are registered in the “Past”...

If an id of a sense is different from the foreseen one, the new one is saved, otherwise, if they are equal, it will be an empty register for that sense. We can call at those id's difference, the Attention focus of the robot (or the robot's conscious)...

Out-Wait-In learning

The Cerebellum has the function to "run" the proprioceptive sense. Or better, the correspondent kinesthetic action of the proprioceptive sense (which became equal by the breakthrough learning, here called Out-Wait-In)...

The robot must have some time at the beginning in which it uses the Out-Wait-In learning (fig. 4) to know it's own body. The goal is that it is going to be a numerical equivalence of ids between the identification id and the saved id for action at the cerebellum action node...

Fig. 4

The robot starts to run a random action (Out) (ref. 1 in fig. 4). Then waits for the reaction of the "world" (Wait). And identifies it with an id which comes as input from the body (proprioception sense) (In) (ref. 2). After passing through the Central unit and return to Cerebellum the acting node should be listening to the identified input id... The acting node then memorizes that id (ref. 3). Next time the id is broadcast, the acting node acts. This learning must be repeated a few times to grant no noise (false ids from having to many input proprioception signals at the same time). The interest of the equality between ids is to have a common protocol among which register (brain - Past) and which acts (cerebellum – acting node)...

Cerebellum

So in Cerebellum, the process begins with a random action (ref. 1 of fig. 5).

Fig. 5

Then the corresponding action(s) will be executed (point 2), through the node Z...

The ref. 3 is, precisely, the running action(s) (genetically foreseen, or, in this case, foreseen by the assembler of the robot's body)... It can be something as simple as a small movement by a mechanical arm, or a faster movement in a particular direction... the physical impossibilities of the body itself or of the "world" (go besides a possible angular movement of the arm, for instance) will be registered in the “Past” by the proprioception as a "physically" restriction and an impossibility too (by the difference on the Attention... later it will never run).

The result of that action will be sensed by the body (ref. 4). Then the action will be registered (ref. 5) as an identified id in the Past of the robot (ref. 6).

The brain sends the id back to the cerebellum. That id is going to be memorized (by the referred learning process Out-Wait-In) in the node Z, that was waiting the broadcast from the Bus (ref. 7).

Finally in the cerebellum the "B" module is going to make the transition between the genetic Y, inhibiting it, and activating the node Z (ref. 8). This one has already sensitized to the proprioception id... On the other hand the starting action (in ref. 1) can be originated by the body itself, as a motivation (m) like “hunger”, through the X node. The process is the same as if it was started from the random module. The motivations are genetic or assembled in a body-action nodes association. They also have proprioceptive signals.

Dynamic

The system dynamic is like this (Fig. 6): The Real event identified is compared with the foreseen memory (previously saved in its “Past”). The difference of signals (the robot attention) is search in the Past for a possible knowledge and repetitive resolution of the “world+body” problem. If it is found, it will be foreseen (by the decision method) and running the next step or the following proprioceptive event in the temporal order of the Past. This is compared, again, with the next Real event in a time step... Also, that difference is memorized as the actual moment or present in the robot Past (history).

Fig. 6

The running actions are the correspondent id of the proprioception sense because of the equivalent id protocol...

An “idea” of a full object is a sequential memorized sense in the robot's Past.

The Past simultaneous register the data (Senses) and commands (Actions – proprioceptin sense). It registers in the same "memory" (comparing with Computers which separate the data memory from the instructions memory).

Emotions

The emotions are the internal state of the system, as the result of a genetic evolution. They are used to express to other members of a society a common identity by showing it's internal state. For interaction and possible modification of the other society member’s actions... An emotion can be a set of detectors as simple as: detecting an older event; a new event; “hunger”; or pain... it can also be a broad detecting process (in the Central Unit) of other internal system particularities. Those are genetically (assembled) expressed by actions at the cerebellum. And for the same learning Out-Wait-In process, those actions (as the ones random generated) will be stored, in the Past, as a set of the senses ids...


Paulo Roque Silva Lisbon, 29 of June 2009