Navigation: =>

Home / Inventions / Entertaining inventions /

NEW INVENTIONS AND MODELS. INVENTION INVENTIONS || NEW INVENTIONS And MODELS

SCIENTIFIC GROUP "PSIGMA"
MOTIVATED BRAIN ANIMATE

Interesting inventions and models

Tsygankov VD, Sharifov SK, Sharifov N.K.

Leave a comment

Introduction

A team consisting of an animat , i.e., an autonomous robot - a CRAB-2 trolley (author Sharifov SK) and its electronic brain in the form of a virtual neurocomputer (designed and tested) was developed and tested by the team of the scientific group PSIGMA under the leadership of Tsygankov VD. VNK) of the "EMBRYON-10.1" type (developer Sharifov NK), which is implemented in the same package of FPGA firm "ALTERA".

The purpose of the development was to solve two problems: The first task is the realization of a long-developed virtual neurocomputer of the EMBRYON type (see the book Tsygankov VD Neurocomputer and the brain , M., SYNTEG, 2001) on a modern microelectronic technology base in the form of FPGAs (Programmable Logical Integral Scheme) of a high degree of integration. The second task is the use of this neurocomputer as the electronic brain of the second new author's version of the autonomous robot, the KRAB animat, which itself forms and conducts its behavior without any initial programming, and which has only general targets in the form of motivational signals, for example , The kind of "pain" or "food".

This publication gives brief information about the device and technical characteristics of the elements of this complex and describes the experiments on the effect on the behavior of the robot "KRAB-2" modification of the properties of its brain, in particular its "subcortical motivational sphere."

Functional diagram

In Fig. 1. the general block diagram of the robotic complex is given. It consists of an electronic brain in the form of a virtual neurocomputer "EMBRYON-10.1", connected via the RS-232 channel to the control object - an animat or an autonomous robot, executed as a trolley with independent drives.

Block diagram of the brain-robot system

Fig. 1. Block diagram of the system "brain-robot"

The virtual neurocomputer "EMBRYON-10.1" consists of the following functional units (Figure 2.):

Functional diagram of the virtual neurocomputer "EMBRYON-10.1"

Fig. 2. Functional diagram of the virtual neurocomputer "EMBRYON-10.1"

  1. A virtual neuroprocessor is a generator of virtual quasineuron networks in the form of a probabilistic discrete coherent field Ψ (n, P / Sj, Uj, NS). The processor is located in the case of FPGA MAX70724 firm "ALTERA".

  2. Alphanumeric display.

  3. Indicator X-codes of the status of the internal memory register.

  4. The neuroprocessor is connected by a bus with a chip of the BVG controller (Hypothesis Extension Unit) executed on a PIC16F874 chip. The controller serves the keyboard, and stores in memory the image that appears on the sensor matrix (CM).

  5. Decoder of groups of motoneurons with a light indicator of the number of the Y-group.

  6. The controller of the RS-232 channel, through which two-way communication with the robot is supported (with the control object).

  7. Keyboard for manual input of parameters or sensory image.

The robot "CRAB-2" (Figure 3) is a three-wheeled trolley with independent drives on two side-wheel engines. The front wheel is passive. The cart is equipped with a group of optical sensors located along the perimeter of the cart body. They have some distant painful zones of detection of an external stimulus.

Robot - animat "CRAB -2"

Fig. 3. Robot - animat "CRAB -2"

On the body of the trolley there are also the controller of the RS-232 channel for two-way communication with a neurocomputer, sensors or transducers of signals from sensors and a block of drives with reverse motor control.

The sensors of the left and right halves of the body, as well as the sensors in the "head" and the "tail" of the robot body, are connected to the four inputs of the left and right engine by some rigid nerve net or genetics by the hereditary scheme of the body. In the ruptures of the nodes of the intersection of the sensory and motor fibers through the RS-232 channel, the EMBRYON neurocomputer is connected, which functions as a spinal coordinating mechanism, dynamically flexibly, in real time moving the cart, incorporating its virtual neural networks and autonomously actively modifying the behavior of the robot, correlating it with Current state of the external environment. At the same time, the main task of the robot is constantly solved - satisfaction of the motivational need, namely, withdrawal from pain and, if there is no pain stimulus, movement in search of "food".

The sensor matrix (CM) of the neurocomputer has two rows. The first S1 line receives signals from pain sensors of the external environment. It displays the state of the external environment or the external world of "Crab". The second S2 line CM is the inner world, the motive or one of the goals of the robot. These, for example, are the commands for moving forward, random walking in search of "food" or harassment. The complex "neurocomputer-robot" performs closed, cyclic operations of the type "perception-action" or complete sensorimotor behavioral acts.

Experiments with the complex

Virtual neurocomputer "EMBRYON-10.1" - robot "CRAB-2"

Experience 1. The first line of the sensor array S1 is given signals from the pain sensors, i.e. S1 = S B. The second line is disabled and the robot is not perceived ( U2 = 0). The activation level of the first line is U1 = 3. This is some analog of the activity of the motivational center or the action of its mediator, for example, a certain concentration of adrenaline. When trying to cover the "CRAB" from above with a cloth or when the hand approaches the body, the robot reacts gently, timidly, gently, smoothly, in pain. Movement is mostly accurate and slow.

Experience 2. Increase the level of excitation by setting U1 = 8. Movements have become more energetic, more productive and rational. The response time slightly increased.

Experience 3. Set the level of U1 = 13. Robot moves become more aggressive and it tries to influence the source of pain.

Experience 4. We establish the motivational level of excitation U1 = 50. The energy of the robot's movements has increased. The time of reaction and inhibition increased markedly. The aggressiveness of "Crab" is obvious.

Experience 5. Level U1 = 255. Robot movements are fast, energetic and long lasting. The robot initially gathers for a long time with "thoughts", and then persistently and persistently long follows in the chosen direction. If you simultaneously touch all the pain sensors, the robot comes to a "stupor" or freezes on the spot.

Now we complicate the robot's behavior by connecting the second row of the sensor array S2 together with its hypothesis of perception U2 .

In the neurocomputer or in the brain of the robot, there is a "Pain Analyzer" block that redistributes "attention" in the form of a vector U1, U2 or selects a certain type of activity, depending on the content of the information on the two lines of the SM. In this case, the total excitation level U1 + U2 = const is maintained constant.

Experience 6. On the second line, "hereditary" is given the "goal of life" S2 = Sn - move quickly and resolutely forward in the absence of pain. We define the hypothesis of perception U = U1 + U2 = 8. The robot starts to move slightly zigzag, basically, in a rectilinear chosen direction. When encountering an obstacle, the pain reflex is triggered in the form of "withdrawal from pain". Attention U2 from the second line is redistributed to the first line U1 CM. As soon as the pain ceases, the robot immediately restores its original movement along a straight line, i.e., the attention of U1 from the first line passes or "flows" to the second row of U2, while fulfilling the condition U1 + U2 = const . The robot bypasses the obstacle long and neatly, without destroying it.

Experience 7. The total excitation level for two lines U = 15. When encountering an obstacle in its motion, the robot "thinks" and makes the right decision to bypass the obstacle. The chosen direction of movement maintains longer than in experiment 6.

Experience 8. We establish a general level of motivation U = 50. The robot rushes very energetically, rarely changes direction, shows inertness of its behavior. The reaction time when encountering an obstacle is slow and the robot is his, the obstacle sweeps off, if the obstacle is weak and unstable. After an aggressive "struggle" with the obstacle, the robot finds a way to get away from the pain.

In conclusion, it should be noted that in the brain of "CRAB" there is no any preliminary programming of his behavior, ways of movement. There are no computational operations in the neurocomputer and there is no normal program control. The robot's brain itself creates motion algorithms and immediately automatically implements them.

The described principles of operation and application implementation of algorithms are designed for use in automatic control systems of unmanned aerial or ground objects in conditions that are inaccessible or dangerous for human stay.

print version
Authors: Tsygankov VD, Sharifov SK, Sharifov NK Moscow
PS The material is protected.
Date of publication 28.08.2004гг