Neural Network Application in Robotics

Neural Network Application in Robotics

Development of Autonomous Aero-Robot and its Applications to Safety and Disaster Prevention with the help of neural network

Sharique Hayat1, R. N. Mall2

1. M.Tech. Final Year CIM, MMMEC Gorakhpur

2. Asstt. Professor, MMMEC Gorakhpur

Abstract- To develop an autonomous robot with the application of neural network and to apply it for monitoring and rescue activities in case of natural or manmade disaster and also implementing the neural network in Maruti Udyog Gurgaon for increasing the productivity and more quality improvement of the system.

1. NTRODUCTION

   The term neural network is used to refer to a network or circuit of biological neurons. The modern usage of the term often refers to artificial neural networks, which are composed of artificial neurons or nodes

   OVERVIEW

   A biological neural network is composed of a group or groups of chemically connected or functionally associated neurons. A single neuron may be connected to many other neurons and the total number of neurons and connections in a network may be extensive. Connections, called synapses, are usually formed from axons to dendrites, though dendrodendritic microcircuits and other connections are possible

   Use of neural network

   Neural networks, with their remarkable ability to derive meaning from complicated data, can be used to extract patterns and detect trends that are too complex to be noticed by either humans or other computer techniques. A trained neural network can be thought of as an "expert" in the

   category of information it has been given to analyse. This expert can then be used to provide projections given new situations of interest and answer "what if" questions. Other advantages include:

   1. Adaptive learning: An ability to learn how to do tasks based on the data given for training or initial experience.

   2. Self-Organisation: An ANN can create its own organisation or representation of the information it receives during learning time.

   3. Real Time Operation: ANN computations may be carried out in parallel, and special hardware devices are being designed and manufactured which take advantage of this capability.

   4. Fault Tolerance via Redundant Information Coding: Partial destruction of a network leads to the corresponding degradation of performance. However, some network capabilities may be retained even with major network damage.

FUNDAMENTAL DESIGNING OF NEURAL NETWORK

• The design of neural networks and how to use them as a robot brain, simple neural network consisting of only two inputs and two outputs.

• In this diagram there are the inputs 'sensors' and the outputs 'motors'. The relationship between the sensors and the motors can be described in the following table (

• means on and -1 means off). If a sensor is on it means the switch is activated

1. HOW TO DESIGN A NEURAL BRAIN 1.Determine what you want the robot to do. 2.Determine the number and types of input you require. 3.Determine the number of outputs you require.

   1. Map inputs and outputs to a vector (both must be in the same vector). If it has less inputs and outputs than points in the vector spread them evenly

   2. Make a complete set as described in point 4 for all possible combinations of inputs and outputs. 6.Repeat step 5 but this time only put the inputs in the vectors.

   7.Now play around with the different variables in neuroqb.zip and try to find the most effective combination

2. The following areas in the flight system where application of neural network are enable.

   Hierarchy structure of Autonomous Flight Control of UAVs top Situation Awareness

   Command Interface middle Switching Flight Mode

   Velocity Control Positioning Control etc.

   Reconfiguring Flight Control Fault Detection

   Flight Controller bottom

   Designing OF Flight Controller

   Two methods are important for designing of flight controller;

   1. Knowledge of Many Experts

   2. Results of Many Experiments

   Designing Control Systems for Complex Systems

   When there are complex problem comes different methods are implemented;

   1. Conventional methods

   2. Linearizing of nonlinear dynamics

   3. Switching linear controllers

   4. Dividing the whole system into some sub-systems

   5. Singular Perturbation are required to design control systems

   Proposed method

   Using neural network training

   Treating complex systems directly and in holistic approach

   Controller using Neural Network

   Ability of neural network Learning

   Training

   Off-line Training

   Training method based on Gradient

   Training method based on Powells conjugated direction algorithm

   On-line Training

   Designing and Developing Control Systems Reconstruction or Reconfiguring Control Systems

3. Training a neural network Optimization of a performance index

   In developing autonomous flight controller of UAVs, the algorithm enables to use mat lab software.

   On-line Training of Neural Network

   Indoor Experiment using a small helicopter(electrically powered) Case1.

   Under disturbance

   A: without network(No disturbance) B: without network(witdisturbance) C: with network(with disturbance)

   Case2.

   Efficiency of the control is reduced

   A:with network

   B:without network

   For the reliability of the autonomous flight

   Numerical Simulations Inputs of a neural network Altitudez velocity vz

   Pseudo-Input U= -Kp(z-d)-Kd vz Output of a neural network Collective control collective

   Nonlinear dynamics is easily transformed to a linear dynamics

   Results of Flight Experiments

   Hovering by PD Controller

   200

   150

   100

   50

   0

   -50

   -100

   	E[err] (cm)	Var[err] 2 (cm )
   without online training	37.8	3832.4
   with online training	22.3	554.4

   	E[err] (cm)	Var[err] 2 (cm )
   without online training	37.8	3832.4
   with online training	22.3	554.4

   0 5 10 15 20 25 30 35 40

   	E[err] (cm)	Var[err] 2 (cm )
   without online training	68.5	77.9
   with online training	41.6	174.5

   	E[err] (cm)	Var[err] 2 (cm )
   without online training	68.5	77.9
   with online training	41.6	174.5

   Hovering by Neural Networks

   200

   150

   100

   50

   0

   -50

   -100

   0 5 10 15 20 25 30 35 40

   Training Controller for Linearization

   U f ( y, y,u) Kp ( y d) Kd y

   With the help of Hovering and PD Controller method the actual flight position are shws in the graph by red and blue curves and resulted parameters are calculated.

   By the analysis it comes out that when the neural network implement in the flight control its become more accurate then the before.

4. Applications of neural networks

   sales forecasting industrial process control customer research

   data validation risk management target marketing

   APPLICATION OF ROBOTS IN ASSEMBLY

   The final assembly is still the most labour intensive of all the automobile production shops. The share of the final assembly in total man-hours required for the manufacture of a car is almost 50%. Under the condition, the assembly line becomes most sensitive to labour attitude and productivity. Automation through robot is the universal approach in final assembly operations. However, the robot used in these operations require precise control, that is attained in one of the following two ways:

   The control function is embedded in the tools or implements, and the robot works in a play back mode based on comparatively simple message exchange.The robot is provided with pattern recognition through visual and tactile sensors and is made to operate in an intelligent manner.

   The robot functions are improved through combination with peripheral tools, and the superiority of robots over human workers is enhanced.

   • Neural Networks have been used in a variety of linear and non-linear controllers.

   • Neural networks can handle one or more inputs and outputs.

   • Neural networks do not work well when dealing with the mathematical problem of converting space coordinates to joint coordinates.

   • Neural networks have been used in most popular control schemes including controlling un modelled processes.

   • Various sensors have been used successfully with neural networks.

   • Back propagation is the most popular neural network paradigm for robotics research.

• .The total sheduling time for the manufacturing of car will be reduced and it reduces the number of robots , so that more workstation will be there.

• The application of neural network in robotics to the flight control for monitoring and rescue activities

FUTURE WORK;

• Neural network application are enabled in maruti udyog gurgaon.

• The total sheduling time for the manufacturing of a car will be reduced and it reduces the number of robots , so that more workstation will be there.

  .

  LIST OF FIGURES

  fig. Manufactuing of Cars in Maruti Udyog Gurgaon

  Past and Present SCOPE

  The applications of neural network in robotics are following.

  • Financial Analysis — stock predictions .

• Signature Analysis — the banks in America have taken to NNs to compare signatures with what is stored.

• Process Control Oversight — NNs are used to advise aircraft pilots of engine problems.

• Direct Marketing — NNs can monitor results from a test mailing and determine the most successful areas.

Neural network have been used in variety of linear and non linear controllers.it can handle one or more inputs and outputs , neural network have been used in most populer control schemes including controlling un modelled processes .

Implementation of neural network in maruti udyog

• The number of robots are implementing in the maruti udyod plant. Trere are different operations are done by the number of robots in the plant .so by applying the neural network all the operations are done by a single robot.

• By applying the neural network application in the plant the number of robots reduced and the multifunctional task will be done in a single plateform.

References

1. Industrial Applications of Neural Networks (research reports Esprit, I.F.Croall, J.P.Mason)

2. An introduction to neural computing. Aleksander, I. and Morton, H. 2nd edition

3. Demarse,Thomas B., Wagenaar,Douglas A., Blau,Axel W., & Potter,Steve M., The Neutrally Controlled Animate: Biological Brains Acting with Simulated Bodies, Autonomous Robots, Kluwer Academic Publishers, 2001

4. Pomerleau, Dean A. Neural Network Vision for Robot Driving, The Handbook of Brain Theory and Neural Networks, M. Arbib, ed., 1995

BIBLIOGRAPHY

1. www.sciencedirect.com

2. www.nitrkl.ac.in