BIO INSPIRED NEURAL NETWORKING AMONG
MULTI-ROBOTS

CHAPTER 1
INTRODUCTION

Transportation is one of the most important economic activities of any country. Among the various forms of transport, road transport is one of the most popular means of transportation. Transportation has an element of danger attached to it in the form of vehicle crashes. Road crashes not only cause death and injury, but they also bring along an immeasurable amount of agony to the people involved. Efforts to improve traffic safety to date have concentrated on the occupant protection, which had improved the vehicle crash worthiness. The other important area where research is currently being done is collision avoidance. Technological innovations have given the traffic engineer an option of improving traffic safety by utilizing the available communication tools and sophisticated instruments. Using sensors and digital maps for increasing traffic safety is in its infancy. Systems are being developed to utilize the available state of the art facilities to reduce or possibly prevent the occurrence of crashes. Total prevention of crashes might not be possible for now, but the reduction of crashes could easily be achieved by using the collision avoidance systems.

1.1 NEED FOR COLLISION AVOIDANCE
The development of collision avoidance systems is motivated by their potential for increased vehicle safety. Half of the more than 1.5 million rear-end crashes that occurred in 1994 could have been prevented by collision avoidance systems .Collision avoidance systems can react to situations that humans cannot or do not, due to driver error. Therefore, they are able to reduce the severity of accidents. Figure 1 below indicates that about 45 percent of the crashes that occur are caused by human errors. Human errors that cause crashes include failure to keep in proper lane, failure to yield right of way, inattentive, failure to obey traffic control devices, operating vehicle in negligent manner, drowsy driving, over correcting, driving wrong way and making improper turns. Some of these crashes may have been possibly avoided if the driver was provided with the real time information.

Figure 1.1 Statistical Survey Results

1.2 HISTORICAL BACKGROUND IN DETAIL
The history of neural networks that was described above can be divided into several periods.

1.2.1 First Attempt
There were some initial simulations using formal logic. McCulloch and Pitts (1943) developed models of neural networks based on their understanding of neurology. These models made several assumptions about how neurons worked. Their networks were based on simple neurons which were considered to be binary devices with fixed thresholds. The results of their model were simple logic functions such as "a or b" and "a and b". Another attempt was by using computer simulations. Two groups (Farley and Clark, 1954; Rochester, Holland, Haibit and Duda, 1956). The first group (IBM researchers) maintained closed contact with neuroscientists at McGill University. So whenever their models did not work, they consulted the neuroscientists. This interaction established a multi-disciplinary trend which continues to the present day. 1.2.2 Promising & Emerging Technology
Not only was neuro-science influential in the development of neural networks, but psychologists and engineers also contributed to the progress of neural network simulations. Rosenblatt (1958) stirred considerable interest and activity in the field when he designed and developed the Perceptron. The Perceptron had three layers with the middle layer known as the association layer. This system could learn to connect or associate a given input to a random output unit. Another system was the ADALINE (ADAptive LInear Element) which was developed in 1960 by Widrow and Hoff (of Stanford University). The ADALINE was an analogue electronic device made from simple components. The method used for learning was different to that of the Perceptron; it employed the Least-Mean-Squares (LMS) learning rule. 1.2.3 Period of Frustration & Disrepute
In 1969 Minsky and Papert wrote a book in which they generalised the limitations of single layer Perceptron to multi-layered systems. In the book they said: "…our intuitive judgment that the extension (to multilayer systems) is sterile". The significant result of their book was to eliminate funding for research with neural network simulations. The conclusions supported the disenchantment of researchers in the field. As a result, considerable prejudice against this field was activated. 1.2.4 Innovation
Although public interest and available funding were minimal, several researchers continued working to develop neuro-morphically based computational methods for problems such as pattern recognition.
During this period several paradigms were generated which modern work continues to enhance. Grossberg's (Steve Grossberg and Gail Carpenter in 1988) influence founded a school of thought which explores resonating algorithms. They developed the ART (Adaptive Resonance Theory) networks based on biologically plausible models. Anderson and Kohonen developed associative techniques independent of each other. Klopf (A. Henry Klopf) in 1972 developed a basis for learning in artificial neurons based on a biological principle for neuronal learning called hetero-stasis.
Werbos (Paul Werbos 1974) developed and used the back-propagation learning method, however several years passed before this approach was popularized. Back-propagation nets are probably the most well-known and widely applied of the neural networks today. In essence, the back-propagation net is a Perceptron with multiple layers, a different threshold function in the artificial neuron, and a more robust and capable learning rule.
Amari (A. Shun-Ichi 1967) was involved with theoretical developments: he published a paper which established a mathematical theory for a learning basis (error-correction method) dealing with adaptive pattern classification. While Fukushima (F. Kunihiko) developed a step wise trained multi-layered neural network for interpretation of handwritten characters. The original network was published in 1975 and was called the Cognitron. 1.2.5 Re-Emergence
Progress during the late 1970s and early 1980s was important to the re-emergence on interest in the neural network field. Several factors influenced this movement. For example, comprehensive books and conferences provided a forum for people in diverse fields with specialized technical languages, and the response to conferences and publications was quite positive. The news media picked up on the increased activity and tutorials helped disseminate the technology. Academic programs appeared and courses were introduced at most major Universities (in US and Europe). Attention is now focused on funding levels throughout Europe, Japan and the US and as this funding becomes available, several new commercial with applications in industry and financial institutions are emerging.
1.2.6 Current Trend
Significant progress has been made in the field of neural networks-enough to attract a great deal of attention and fund further research. Advancement beyond current commercial applications appears to be possible, and research is advancing the field on many fronts. Neural based chips are emerging and applications to complex problems developing. Clearly, today is a period of transition for neural network technology.
1.3 EXISTING COLLISION AVOIDANCE SYSTEM
In traditional process of collision avoidance, intelligence is often programmed from above. The programme is the creator, and makes something and inculcates it with its intelligence. The system performs based on the given program in an efficient manner but with a faster pace than the human being. But the major disadvantage of that system is it cannot think or act on its own in non-stationery environment. Thus each time it has to be programmed to match the current situation by the programmer which is practically not possible. Thus a system which is fully monitored and controlled by its programmer which is unable to take its own decisions, this has minimised the rate of collision but have not created collision free environment.
1.4 COLLISION AVOIDANCE USING BIO INSPIRED NEURAL NETWORK
We have designed and tested a real-time cerebellar system for cooperative collision avoidance inspired by organization and functioning of the brain to discover new structural and learning inspirations. This cooperative collision avoidance system is based on the information sharing phenomenon among multi-robots. To conduct the cooperative collision avoidance among multi-robots in unknown and dynamic environments, the robots not only need to take into account basic problems (such as searching, path planning, and collision avoidance), but also need to cooperate in order to pursue a collision free traffic. In this paper, a novel approach based on a bio-inspired neural network is proposed for the real-time collision avoidance among multi-robots, where the locations of the co-working robots and the environment are unknown and changing.
The bio-inspired neural network is used for cooperative pursuing by the multi-robot team. The system was implemented with mixed analog-digital integrated circuits consisting of an analog resistive network and field-programmable gate array (FPGA) circuits so as to take advantage of the real-time analog computation and programmable digital processing. The response properties of the system were examined by using simulated images, and the system was also tested in real-world situations by loading it on a motorized miniature car. In the proposed approach, the pursuing alliances can dynamically change and the robot motion can be adjusted in real-time to pursue without any collision with other robots functioning along with it. It not only prevent itself from colliding with its co-functioning robots but will also send the entire information about the environment in which it is working to other robots, so that they take a different routes even before encountering the situation of collision.
The proposed approach can deal with various situations such as when robots break down, the environment has different boundary shapes, or the obstacles are linked with different shapes. The system was confirmed to respond selectively to colliding objects even in complex real-world situations. Thus the proposed approach is capable of guiding the robots to achieve a collision free traffic in real-time efficiently.

CHAPTER 2
INTRODUCTION TO NEURAL NETWORKS
2.1 NEURAL NETWORKS
An Artificial Neural Network (ANN) is an information processing paradigm that is inspired by the way biological nervous systems, such as the brain, process information. The key element of this paradigm is the novel structure of the information processing system. It is composed of a large number of highly interconnected processing elements (neurones) working in unison to solve specific problems. ANNs, like people, learn by example. An ANN is configured for a specific application, such as pattern recognition or data classification, through a learning process. Learning in biological systems involves adjustments to the synaptic connections that exist between the neurones. This is true of ANNs as well.
2.2 HISTORICAL BACKGROUND
Neural network simulations appear to be a recent development. However, this field was established before the advent of computers, and has survived at least one major setback and several eras.
Many important advances have been boosted by the use of inexpensive computer emulations. Following an initial period of enthusiasm, the field survived a period of frustration and disrepute. During this period when funding and professional support was minimal, important advances were made by relatively few researchers. These pioneers were able to develop convincing technology which surpassed the limitations identified by Minsky and Papert. Minsky and Papert, published a book (in 1969) in which they summed up a general feeling of frustration (against neural networks) among researchers, and was thus accepted by most without further analysis. Currently, the neural network field enjoys a resurgence of interest and a corresponding increase in funding.
The first artificial neuron was produced in 1943 by the neurophysiologist Warren McCulloch and the logician Walter Pits. But the technology available at that time did not allow them to do too much.
2.3 OBJECTIVE OF NEURAL NETWORKS
Neural networks, with their remarkable ability to derive meaning from complicated or imprecise 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.
2.4 NEURAL NETWORKS VERSUS CONVENTIONAL COMPUTERS
Neural networks take a different approach to problem solving than that of conventional computers. Conventional computers use an algorithmic approach i.e. the computer follows a set of instructions in order to solve a problem. Unless the specific steps that the computer needs to follow are known the computer cannot solve the problem. That restricts the problem solving capability of conventional computers to problems that we already understand and know how to solve. But computers would be so much more useful if they could do things that we don't exactly know how to do.
Neural networks process information in a similar way the human brain does. The network is composed of a large number of highly interconnected processing elements (neurones) working in parallel to solve a specific problem. Neural networks learn by example. They cannot be programmed to perform a specific task. The examples must be selected carefully otherwise useful time is wasted or even worse the network might be functioning incorrectly. The disadvantage is that because the network finds out how to solve the problem by itself, its operation can be unpredictable.
On the other hand, conventional computers use a cognitive approach to problem solving; the way the problem is to solved must be known and stated in small unambiguous instructions. These instructions are then converted to a high level language program and then into machine code that the computer can understand. These machines are totally predictable; if anything goes wrong is due to a software or hardware fault.
Neural networks and conventional algorithmic computers are not in competition but complement each other. There are tasks are more suited to an algorithmic approach like arithmetic operations and tasks that are more suited to neural networks. Even more, a large number of tasks, require systems that use a combination of the two approaches (normally a conventional computer is used to supervise the neural network) in order to perform at maximum efficiency.
2.5. HUMAN AND ARTIFICIAL NEURONS - INVESTIGATING THE SIMILARITIES
2.5.1 Learning Process of Human Brain
Much is still unknown about how the brain trains itself to process information, so theories abound. In the human brain, a typical neuron collects signals from others through a host of fine structures called dendrites. The neuron sends out spikes of electrical activity through a long, thin stand known as an axon, which splits into thousands of branches. At the end of each branch, a structure called a synapse converts the activity from the axon into electrical effects that inhibit or excite activity from the axon into electrical effects that inhibit or excite activity in the connected neurones. When a neuron receives excitatory input that is sufficiently large compared with its inhibitory input, it sends a spike of electrical activity down its axon. Learning occurs by changing the effectiveness of the synapses so that the influence of one neuron on another changes. Figure 2.1 Components of a Neuron | Figure 2.2 The Synapse |
2.5.2 Human Neurons to Artificial Neurons
We conduct these neural networks by first trying to deduce the essential features of neurones and their interconnections. We then typically program a computer to simulate these features. However because our knowledge of neurones is incomplete and our computing power is limited, our models are necessarily gross idealisations of real networks of neurones.

Figure 2.3 The Neuron Model
2.6. AN ENGINEERING APPROACH
2.6.1 A Simple Neuron
An artificial neuron is a device with many inputs and one output. The neuron has two modes of operation; the training mode and the using mode. In the training mode, the neuron can be trained to fire (or not), for particular input patterns. In the using mode, when a taught input pattern is detected at the input, its associated output becomes the current output. If the input pattern does not belong in the taught list of input patterns, the firing rule is used to determine whether to fire or not.

Figure 2.4 A Simple Neuron
2.6.2 A More Complicated Neuron
The previous neuron doesn't do anything that conventional computers don't do already. A more sophisticated neuron (figure 2) is the McCulloch and Pitts model (MCP). The difference from the previous model is that the inputs are ‘weighted’; the effect that each input has at decision making is dependent on the weight of the particular input. The weight of an input is a number which when multiplied with the input gives the weighted input. These weighted inputs are then added together and if they exceed a pre-set threshold value, the neuron fires. In any other case the neuron does not fire.

Figure 2.5 A MCP Neuron

In mathematical terms, the neuron fires if and only if;
X1W1 + X2W2 + X3W3 + ... > T
The addition of input weights and of the threshold makes this neuron a very flexible and powerful one. The MCP neuron has the ability to adapt to a particular situation by changing its weights and/or threshold. Various algorithms exist that cause the neuron to 'adapt'; the most used ones are the Delta rule and the back error propagation. The former is used in feed-forward networks and the latter in feedback networks.

CHAPTER 3
PROPOSED SYSTEM
3.1 BIO-INSPIRED NEURAL NETWORKS
Bio-inspired neural network is a bottom-up, decentralized approach.it is simply a set of rules which has to be followed by the robots cooperatively and they adopt themself to the robot, the extra feature compared to other networks is it is capable of learning from past experience and able to train and adopt itself to the current situation.
3.1.1 Training Process
Learning in a neural network is called training. Like training in athletics, training in a neural network requires a coach, someone that describes to the neural network what it should have produced as a response. It is propagated backward through the network. At each neuron in the network the From the difference between the desired response and the actual response, the error is determined and a portion error is used to adjust the weights and threshold values of the neuron, so that the next time, the error in the network response will be less for the same inputs.
The control system of an intelligent agent should be capable of training and memorizing. Neural networks can memorize both through plasticity (changes in the parameters of the neural units, e.g. weights or axonal delays), and through the dependence of their activation state on past states. Given the usual timescales of these changes of the state of the network, plastic memory usually persists on long term, while activation memory usually persists for relatively shorter durations. The output of a feed forward neural network with continuous, spontaneous activation depends exclusively on the current input. The memory of a feed forward integration in time of the inputs has a temporal extent of the order of the integration time constant of the units. The most complex type of activation memory arises in recurrent neural networks, where the activity reverberates in loops and the current state of the network may depend significantly on past input. Recurrent neural networks are thus preferable for building control systems for embodied agents.

Figure 3.1 Weight Calculation
3.1.2 Learning through Past Experience
A genuinely intelligent agent should be adaptive, flexible, and robust: it should adjust its operation to unexpected changes that influence it, and should be creative in finding solutions for completing its tasks, rather than conforming exclusively to preprogramed behaviour. Its categories and concepts should not be imposed by its designer, but should be developed by the agent itself, as a consequence of its interaction with the environment, its goals and its corporal capabilities.
According to this perspective, the designer should not implement exclusively supervised learning schemes where the response of the agent in particular situations is predefined at the level of particular states of the output vector. The designer or the user of the agent may guide its behaviour primarily through reinforcement learning, and occasionally by demonstration–imitation or through physical guidance in the environment. Self-organizing mechanisms should also be implemented. Neural networks possess many known self-organizational capabilities for unsupervised learning, which will be discussed below for some specific types of networks. Reinforcement learning may also be implemented in neural networks, for example by feeding the reinforcement signal throughout the network and regulating learning mechanisms, such as local plasticity, through this neuromodulatory signal.
3.1.3 Learning Process
3.1.3.1. Supervised
• A series of inputs with known expected outputs are provided and the network attempts to minimize error.
– Widrow-Hoff Learning for Linear Function
• Initialize weights to 0
• Until expected output matches network’s output
– Randomly select an input
– Take cross product of weights with input
– Calculate deltai for each wi where deltai = wi *(correct output – actual output)*inputi
– Add deltai to wi

3.1.3.2. Unsupervised
• The neural network is given a cost function that is some function of the input and the neural network’s output
• The network’s goal is to find an optimal, but, in practice, relatively minimized, cost function

3.1.3.3. Reinforcement
• Instead of waiting until the output is evaluated, the neural network is given feedback on its progress throughout its evaluation of the data
• This feedback allows the network to strengthen or weaken links between nodes

3.1.3.4. Back-propagation
• Back-propagation is an example of supervised learning where something like
Widrow-Hoff is used at each layer to minimize the error between the layer’s response and the actual data
• The error at each hidden layer is an average of the previously evaluated error
• Hidden layer networks are trained this way

3.2 OVERVIEW
The below picture explains the exact operation of our project. It has three basic blocks on which it works:
Block 1: system
Block 1 is used for developing a code using Verilog HDL language using Xilinx software and the simulation results are monitored in a pc or laptop.
Block 2: FPGA
Block 2 is used to establish the connection between system and multi robots. It is connected to block 1 and block 3 using RS232 cable.
Block 3:multi-robots
A robot is a mechanical or virtual intelligent agent that can perform tasks automatically or with guidance, typically by remote control. In practice a robot is usually an electro-mechanical machine that is guided by computer and electronic programming. Robots can be autonomous, semi-autonomous or remotely controlled.

Figure 3.2 Pictorial Representation 3.3 ARCHITECTURE
The proposed approach consists of one server and many clients which are embedded as individual sub-routines into a field programmable gate array. We are using single FPGA for effective usage of 2, 50,000 gates. If more clients are into the network the number of gates inside the FPGA will be divided among them and used effectively.

Figure 3.3 Basic Architecture

3.4 BLOCK DIAGRAM
This block diagram is the pictorial representation of our project.it consists of FPGA kit, RS232 cable, serial to parallel convertor embedded inside the kit, LED panel. It basically involves many sections in FPGA kit. Short for Field-Programmable Gate Array, a type of logic chip that can be programmed FPGAs support thousands of gates. They are especially popular for prototyping integrated circuit designs. Once the design is set, hardwired chips are produced for faster performance. The kit is divided into one server and two clients (i.e. the number of gates is shared between server and different clients). This distribution of gates consists of many cells which is the basic training entity. Now the output from FPGA is converted using serial to parallel convertor and fed via RS232 cable and Results are displayed using LED panel. Thus the seven segment display in FPGA gives more information and clear description of the obtained output. Thus the output is obtained and effectively understood.
-
Figure 3.4 Block Diagram
3.5 PRE REQUISITES * Software component * Hardware component * FPGA kit.
3.5.1 Software
The software part consists of 3.5.1.1 Xilinx 9.1
Xilinx Software Development Kit (SDK) is a complete embedded software development environment that supports all Xilinx FPGA architectures including the latest 7 series FPGAs. Xilinx SDK provides a graphical software development environment with comprehensive application edit, build, deploy and debug capabilities for bare-metal and operating system-based application development. * Xilinx SDK Benefits * Full Graphical, Software Development Environment * Based on de facto standard Eclipse environment * Editor, Compilers, Build tools, Flash memory management, and JTAG/GDB debug integration * Custom libraries and device drivers * Off-the-shelf support for bare-metal and Linux development * Also supported by commercial RTOS * Project pre-configuration and set-up * GCC compilers and GNU GDB Debugger pre-configured for bare-metal and Linux development * Includes pre-configured, hardware optimized libraries (e.g. ARM VFP and NEON for Zynq-7000 EPP) * Pre-configured boot image, IP cores, Library, and source file paths * Auto-generates Boot Loader, BSP and Boot Images * SDK uses project-specific configuration data to configure and build key software elements: * Boot loader * Auto-generate Bare-metal and Linux drivers and BSPs * Quickly generate Board Support Packages (BSPs) for other operating systems like Linux, VxWorks, uCos, etc. * Generate System Boot Image comprised of programmable logic bitstream and operating system/application files * Manages Host-Target Connectivity (JTAG, Flash, serial) * Pre-configured, hardware-specific, host-target connections * Serial console * GDB and JTAG debugging * Flash memory read/write

3.5.1.2 Model SIM 6.3
ModelSim PE is the industry-leading, Windows-based simulator for VHDL, Verilog, or mixed-language simulation environments.

* ModelSim PE Features

- Partial VHDL 2008 support
- Transaction wlf logging support in all languages including VHDL
- Windows7 Support
- SecureIP support
- SystemC option
- RTL and Gate-Level Simulation
- Integrated Debug
- Verilog, VHDL and SystemVerilog Design
- Mixed-HDL Simulation option
- Code Coverage option
- Enhanced debug option
- Windows 32-bit * ModelSim PE Benefits
- Cost-effective HDL simulation solution
- Intuitive GUI for efficient interactive debug
- Integrated project management simplifies managing project data
- Easy to use with outstanding technical support
- Sign-off support for popular ASIC libraries
- Award-winning technical support.

3.5.2 Hardware
The hardware part involved in our project consists of a FPGA kit.
3.5.2.1 FPGA kit
A field-programmable gate array (FPGA) is an integrated circuit designed to be configured by the customer or designer after manufacturing—hence "field-programmable". The FPGA configuration is generally specified using a hardware description language (HDL), similar to that used for an application-specific integrated circuit (ASIC) (circuit diagrams were previously used to specify the configuration, as they were for ASICs, but this is increasingly rare). FPGAs can be used to implement any logical function that an ASIC could perform. The ability to update the functionality after shipping, partial re-configuration of a portion of the design and the low non-recurring engineering costs relative to an ASIC design (notwithstanding the generally higher unit cost) offer advantages for many applications.
FPGAs contain programmable logic components called "logic blocks", and a hierarchy of reconfigurable interconnects that allow the blocks to be "wired together"—somewhat like many (changeable) logic gates that can be inter-wired in (many) different configurations. Logic blocks can be configured to perform complex combinational functions, or merely simple logic gates like AND and XOR. In most FPGAs, the logic blocks also include memory elements, which may be simple flip-flops or more complete blocks of memory.
In addition to digital functions, some FPGAs have analog features. The most common analog feature is programmable slew rate and drive strength on each output pin, allowing the engineer to set slow rates on lightly loaded pins that would otherwise ring unacceptably, and to set stronger, faster rates on heavily loaded pins on high-speed channels that would otherwise run too slow. Another relatively common analog feature is differential comparators on input pins designed to be connected to differential signalling channels. A few "mixed signal FPGAs" have integrated peripheral Analog-to-Digital Converters (ADCs) and Digital-to-Analog Converters (DACs) with analog signal conditioning blocks allowing them to operate as a system-on-a-chip. Such devices blur the line between an FPGA, which carries digital ones and zeros on its internal programmable interconnect fabric, and field-programmable analog array (FPAA), which carries analog values on its internal programmable interconnect fabric.

Figure 3.5 Pin-Out Diagram of Spartan 3
3.5.2.2 RS 232 Cable
A serial cable is a cable that can be used to transfer information between two devices using serial communication. The form of connectors depends on the particular serial port used. A cable wired for connecting two data terminal equipment directly is known as a null modem cable. This cable has short transmission distance because of noise limiting the transmission of high numbers of bits per second when the cable is more than 15 meters long. This means that the transmitting and receiving lines are referenced to ground. It is cheap to purchase and is simple to join and connect. It is suitable for unbalanced data standards. Only one device can be connected to the cable. The RS 232 standard states that a compliant port must provide defined signal characteristics for a capacitive load of 2500 pF. This does not correspond to a fixed length of cable since varying cables have different characteristics.

Figure 3.6 Block diagram of RS232

3.5.2.3 Multi-robots The multi-robots can be built using the following components. They are as listed below. * IR obstacle sensors
The basic concept of IR (infrared) obstacle detection is to transmit the IR signal (radiation) in a direction and a signal is received at the IR receiver when the IR radiation bounces back from a surface of the object. Here in the figure the object can be anything which has certain shape and size, the IR LED transmits the IR signal on to the object and the signal is reflected back from the surface of the object. The reflected signals are received by an IR receiver. The IR receiver can be a photodiode / phototransistor or a readymade module which decodes the signal.

Figure 3.7 IR Sensor- Working Principle * AT89S52 (atmel microcontroller) The features of AT89S52 microcontroller are * High performance, low power Atmel * Advanced RISC architecture * High endurance non-volatile memory segments * QTouch * Peripheral features * Two 8-bit timer/counters with separate pre-scaler and compare mode * One 16-bit timer/counter with separate pre-scaler, compare mode, and capture mode * Real time counter with separate oscillator * Six PWM channels * 8-channel 10-bit ADC in TQFP and QFN/MLF package * 6-channel 10-bit ADC in PDIP Package * Programmable serial USART * Master/slave SPI serial interface * Operating voltage: * 1.8V - 5.5V for Atmel ATmega48V/88V/168V * 2.7V - 5.5V for Atmel ATmega48/88/168 * Temperature range: * -40°C to 85°C * Speed grade: * ATmega48V/88V/168V: 0 - 4MHz @ 1.8V - 5.5V, 0 - 10MHz @ 2.7V - 5.5V * ATmega48/88/168: 0 - 10MHz @ 2.7V - 5.5V, 0 - 20MHz @ 4.5V - 5.5V * Low power consumption * Active mode: 250µA at 1MHz, 1.8V 15µA at 32kHz, 1.8V (including oscillator) * Power-down mode: 0.1µA at 1.8V * MAX 232 The MAX232 is an integrated circuit that converts signals from an RS-232 serial port to signals suitable for use in TTL compatible digital logic circuits. The MAX232 is a dual driver/receiver and typically converts the RX, TX, CTS and RTS signals. Figure 3.8 Block Diagram of Multi-Robots * Motors
Electric motors are used to “actuate” something in your robot: its wheels, legs, tracks, arms, fingers, sensor turrets, or weapon systems. There are literally dozens of types of electric motors (and many more if you count gasoline and other fuelled engines), but for amateur robotics, the choice comes down to these three: * In a continuous DC motor, application of power causes the shaft to rotate continually. The shaft stops only when the power is removed, of if the motor is stalled because it can no longer drive the load attached to it. * In a stepping motor, applying power causes the shaft to rotate a few degrees and then stop. Continuous rotation of the shaft requires that the power be pulsed to the motor. As with continuous DC motors, there are sub-types of stepping motors. Permanent magnet steppers are the ones you’ll likely encounter, and they are also the easiest to use. * A special “subset” of continuous motors is the servo motor, which in typical cases combines a continuous DC motor with a “feedback loop” to ensure accurate positioning. There are many, many types of servo motors; a common form is the kind used in model and hobby radio-controlled cars and planes.

CHAPTER-4
IMPLEMENTATION
4.1. VIRTUAL LINKAGE
In the proposed approach, we have a single server and multiple clients which can be embedded into a single FPGA kit. For instance, we take two clients and one server. All the three of them are embedded in the same FPGA kit in order to make use of 2, 50,000 gates available in an efficient manner. Thus, we are creating a virtual link between the server and the clients. The clients operate based on the information provided form the server. The clients work in a mutually co-operative environment which is created by the server. However, the server is only to provide information regarding the path and other obstacles in the path and not to make the client operate. The clients and the server operate individually. Thus the client and the server are mutually independent of each other.
4.2 INFORMATION SHARING
Information sharing is the basic concept in this approach which makes this project be an edge above the other collision avoidance systems existing in the world. The information sharing is carried out based on the information collected in the server. The information is collected by the server from the clients’ operation. The clients send the information about every particular obstacle they face in their path and thus it gets stored in the server. Thus, the server creates a huge database which is sent to each and every client while they operate on the path. Thus it provides the complete detail about the path and the obstacles present in the various parts of the path. This concept of information sharing is a new concept which will be swaying the world in nearest decades. The clients send their previous experiences as the information to the server. The server in turn stores these experiences as its information to update the database maintained by it.
4.3. CO-OPERATIVE MOVEMENT Due to the above two virtual linkage and information sharing concept, there is a co-operative movement between the clients in the path. The clients co-operate with each other and work without collision. The robots assist each other in the working path and move in an effective manner. They help each other by sharing the information it gathers while it moves. They share the information through the server robot and collect the information through the same server robots and thus moves in a collision free manner. This project can be further implemented in an environment where the evaders need to be detected and destroyed as early as they get into the picture. Thus co-operative movement of robots is another important feature of this project which makes it more advanced when compared to other collision avoidance systems.
4.4 SHORTEST PATH
The shortest path problem is the problem of finding a path between two vertices (or nodes) in a graph such that the sum of the weights of its constituent edges is minimized. An example is finding the quickest way to get from one location to another on a road map; in this case, the vertices represent locations and the edges represent segments of road and are weighted by the time needed to travel that segment.

There are several variations according to whether the given graph is undirected, directed, or mixed. For undirected graphs, the shortest path problem can be formally defined as follows. Given a weighted graph (that is, a set V of vertices, a set E of edges, and a real-valued weight function f : E → R), and elements v and v' of V, find a path P (a sequence of edges) from v to a v' of V so that is minimal among all paths connecting v to v' .

The problem is also sometimes called the single-pair shortest path problem, to distinguish it from the following variations: * The single-source shortest path problem, in which we have to find shortest paths from a source vertex v to all other vertices in the graph. * The single-destination shortest path problem, in which we have to find shortest paths from all vertices in the directed graph to a single destination vertex v. This can be reduced to the single-source shortest path problem by reversing the arcs in the directed graph. * The all-pairs shortest path problem, in which we have to find shortest paths between every pair of vertices v, v' in the graph.
These generalizations have significantly more efficient algorithms than the simplistic approach of running a single-pair shortest path algorithm on all relevant pairs of vertices.

4.5 WORKING PROCEDURE
The algorithm for finding the shortest path is detected and from algorithm program is developed and finally the program is simulated and desired output is obtained.
Step 1: First the path is completely visualised and is divided into many small portions.

Step 2: The path is fully scanned for obstacles and then values are assigned to every node a tentative distance value: set it to zero for our initial node and to infinity for all other nodes.
Step 3: Mark all nodes unvisited. Set the initial node as current. Create a set of the unvisited nodes called the unvisited set consisting of all the nodes except the initial node.

Step 4: For the current node, consider all of its unvisited neighbours and calculate their tentative distances. For example, if the current node A is marked with a tentative distance of 6, and the edge connecting it with a neighbour B has length 2, then the distance to B (through A) will be 6+2=8. If this distance is less than the previously recorded tentative distance of B, then overwrite that distance. Even though a neighbour has been examined, it is not marked as visited at this time, and it remains in the unvisited set.

Step 5: When we are done considering all of the neighbours of the current node, mark the current node as visited and remove it from the unvisited set. A visited node will never be checked again; its distance recorded now is final and minimal.

Step 6: If the destination node has been marked visited (when planning a route between two specific nodes) or if the smallest tentative distance among the nodes in the unvisited set is infinity (when planning a complete traversal), then stop.

Step 7: Set the unvisited node marked with the smallest tentative distance as the next "current node" and go back to step 3.

Step8: All the maximum possible paths are registered in the client side by the robot. This information is shared by all the clients since all the clients are in direct contact with the server in FPGA kit.
Step 9: All information of a particular robot is shared via server to all other robots and it is updated constantly.

Step 10: It acts accordingly when information provided.
First the path is recognised by the robot and position is identified by the robot.

Figure 4.1: Path represented with 12 distinct positions 4.6 VISUAL BASIC – SIMULATION RESULTS
The program is executed and simulation results are obtained effectively.
First the shortest path between source and the destination is found out from the program and for effective understanding the simulated result is shown in via VB.

Example 1: when the source and destination are in different points and if the pit is not an obstacle for traversing the path, then the output is obtained effectively as shown below. Now let us consider source at 1 and destination at 9 then the shortest path detected and obtained which are captured effectively.

Figure 4.2 Shortest path-finding process

Example 2: when source is at 1 and destination is at 8 and if the location of pit is at 9 then shortest paths are found effectively which is captured and displayed.

Figure 4.3 Shortest path-finding process with pit.

Example 3: when the source is at 1 and destination is at 4 and pit is in between 2 and 3 and when the gate is in open condition then the obtained simulation result shows this condition is impossible.

Figure 4.4 Failure to find shortest path

4.7 FINAL NETWORK

Figure 4.5 Pin diagram of the final network
The above figure illustrates the final network of the proposed project. The multi robots work based on the principle of neural networking. The multi robots mimic the functional biological neural networks. It is also an adaptive system which changes its structure based on the external or internal information which flows through the network during the learning phase.
These multi-robots travel from source to destination detecting and avoiding collisions on the way as they travel. This is done by the multi-robots themselves. They calculate weight of the objects before travelling from the source to the destination. This is nothing but a closed loop system where the multi-robots send the weights back to the source as feedback controls till they reach the destination.

Figure 4.6 RTL scheme of weight calculation network The neural networking is an advanced version of the adaptive neural networks. They are able to modify their structure based on the information they receive during the learning phase. Due to this they are able to think and act by themselves without any human interference or assistance. This helps to economize time in critical situations. These clients, connected in a network, are able to work in a cooperative manner, in conjunction with the other clients. This cooperative movement of the clients is possible only due to the storage of the data in the server. The server preserves the entire database, storing all the paths which have been used by the client in its previous experiences. This is an example of adaptive neural networking. All the clients are able to access the server and learn the positions of the other clients thereby avoiding collisions. Each client is tasked with the job of finding its shortest path from the source to the destination. This clients use the shortest path algorithm which was coded using VB. The main purpose of the clients is to find the shortest path but they also have to ensure that there is no chance of collision with another client, thus collision detection and avoidance also becomes a priority. If two clients have a chance of colliding with each other then it works in a serial manner, i.e. one client moves first and once the first client has reached its destination then another client starts from its source. If the chance of collision is zero then a different method is used: the parallel method in which both the clients start from their source at the same time.

CHAPTER-5
SIMULATION RESULTS

Figure 5.1 simulation result of change in angle of the stepper motor of the multi-robot when encountered with an obstacle

Figure 5.2 RTL schematic diagram of angle and path detection algorithm

Figure 5.3 RTL schematic diagram for a network with a single source and a single destination with one position vector which gives the position of the multi-robot in the path with a 3 bit vector for the detection of the pit in the path

Figure 5.4 RTL schematic diagram for a single source and a single destination with a single position vector and a reached variable to intimate when the robot reach the exact destination.

Figure 5.5 RTL schematic diagram for two sources and two destinations with two different positions of robot1 and robot2 in two different position vectors with a comp variable to denote if there is any collision in the network

Figure 5.6 RTL schematic diagram for a final network with two sources and two destinations with a single position vector to reveal the positions of two robots

Figure 5.7 simulation results of various positions taken by the two robots when the source and the destination for the two robots are given

Figure 5.8 simulation results of various positions taken by the two robots when the source and the destination for the two robots are given represented using two different position vectors.

CHAPTER-6
CONCLUSION AND FUTURE ENHANCEMENTS

6.1 CONCLUSION
Thus the project is completed successfully and implemented in the real time system and desired result is obtained. The robot is made to work on its own intelligence and many probable collisions are avoided by using this current project. Thus when this project introduced in transportation system it will be a breakthrough.The advantage of this approach is that the system takes its own decision under critical situations and avoids collision. It can train itself and get updated in dynamic environment. On continuous learning process, the system gets its memory elevated as in human brain, where infinite windows gets opened every time an information is stored. The results are obtained as simulated images. This approach can be applied in air, rail and sea traffic to tackle a troublesome environment. It can be applied to situations where the dimensions of the path are unknown and changing where it can adapt itself to different real-time instances. Thus, this approach if implemented will definitely be a breakthrough in creating a collision free traffic.
6.2 APPLICATIONS OF NEURAL NETWORKS
6.2.1 Neural Networks in Practice
Given this description of neural networks and how they work, what real world applications are they suited for Neural networks have broad applicability to real world business problems. In fact, they have already been successfully applied in many industries.
Since neural networks are best at identifying patterns or trends in data, they are well suited for prediction or forecasting needs including: * sales forecasting * industrial process control * customer research * data validation * risk management * target marketing
But to give you some more specific examples; ANN are also used in the following specific paradigms: recognition of speakers in communications; diagnosis of hepatitis; recovery of telecommunications from faulty software; interpretation of multiple meaning Chinese words; undersea mine detection; texture analysis; three-dimensional object recognition; hand-written word recognition; and facial recognition.
The proposed approach can be applied in various situations such as when some robots break down, the environment has different boundary shapes, or the obstacles are linked with different shapes. Due to its dynamic nature, it can adapt in various situations and hence can be used for Collision avoidance in air traffic, Collision avoidance in trains and in other alert and troublesome environment.
6.3 FUTURE ENHANCEMENT
The problem of multi-robot coordination and cooperation has drawn great interest in recent years. Generally speaking, for a given task, utilizing more than one robot may enhance the quality of the solution. Furthermore, many inherently distributed tasks must require a distributed solution. However, if the robots are not properly organized, the interference among them will block the task. Many challenging issues should be considered carefully before they can be successfully implemented. The hunting task concerning mobile robots and its target to be hunted is a particular challenge due to the nature of unknown and irregular motion of the target. In order to coordinate the motion of multiple mobile robots to capture or enclose a target, a novel feedback-control law, linear autonomous system and Multiple Objective Behaviour Coordination have been used. Other related works including pursuit game, whose environments are usually modelled in grid. Because the positions of the robots are not exchanged among them in order to reduce the communication burden, it is hard for each robot to make a global decision. A better idea is to complete the task by local interaction among the robots.

To successfully apply evolutionary algorithms to the solution of increasingly complex problems, we must develop effective techniques for evolving solutions in the form of interacting co-adapted subcomponents. One of the major difficulties is finding computational extensions to our current evolutionary paradigms that will enable such subcomponents to “emerge” rather than being hand designed. Given a simple string matching task, we show that evolutionary pressure to increase the overall fitness of the ecosystem can provide the needed stimulus for the emergence of an appropriate number of interdependent subcomponents that cover multiple niches, evolve to an appropriate level of generality, and adapt as the number and roles of their fellow subcomponents change over time.

Complete coverage path planning requires the robot path to cover every part of the workspace, which is an essential issue in cleaning robots and many other robotic applications such as vacuum robots, painter robots, land mine detectors, lawn mowers, automated harvesters, and window cleaners. In this paper, a novel neural network approach is proposed for complete coverage path planning with obstacle avoidance of cleaning robots in non-stationary environments. The dynamics of each neuron in the topologically organized neural network is characterized by a shunting equation derived from Hodgkin and Huxley's (1952) membrane equation. There are only local lateral connections among neurons. The robot path is autonomously generated from the dynamic activity landscape of the neural network and the previous robot location.

Future implementations may also include complete coverage of the path by the robots which require the robot path to cover every part of the workspace. This is an essential issue in cleaning robots and many other robotic applications such as vacuum robots, painter robots, land mine detectors, lawn mowers, automated harvesters, and window cleaners. With further research a complete coverage path planning with obstacle avoidance of cleaning robots in non- stationary environments may be implemented in this project. The dynamics of each neuron in the topologically organized neural network is characterized by a shunting equation derived from Hodgkin and Huxley's (1952) membrane equation. There are only local lateral connections among neurons. The robot path is autonomously generated from the dynamic activity landscape of the neural network and the previous robot location. The proposed model algorithm is computationally simple.

REFERENCES
[1] R. M. Murray, “Recent research in cooperative control of multivehicle systems,” Trans. ASME J. Dyn. Syst., Meas. Control, vol. 129, no. 5, pp. 571–583, Sep. 2007.
[2] R. K. Sharma and D. Ghose, “Collision avoidance between UAV clusters using swarm intelligence techniques,” Int. J. Syst. Sci., vol. 40, no. 5, pp. 521–538, May 2009.
[3] S. Sariel, T. Balch, and N. Erdogan, “Naval mine countermeasure missions,” IEEE Robot. Autom. Mag., vol. 15, no. 1, pp. 45–52, Mar.
2008.
This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination. NI AND YANG: BIOINSPIRED NEURAL NETWORK FOR REAL-TIME COOPERATIVE HUNTING BY MULTIROBOTS 15
[4] A. Farinelli, L. Iocchi, and D. Nardi, “Multirobot systems: A classification focused on coordination,” IEEE Trans. Syst., Man, Cybern., Part B: Cybern., vol. 34, no. 5, pp. 2015–2028, Oct. 2004.
[5] Z. Cao, M. Tan, L. Li, N. Gu, and S. Wang, “Cooperative hunting by distributed mobile robots based on local interaction,” IEEE Trans. Robot., vol. 22, no. 2, pp. 403–407, Apr. 2006.
[6] K. Tanaka and E. Kondo, “A scalable localization algorithm for high dimensional features and multirobot systems,” in Proc. IEEE Int. Conf. Network., Sens. Control, Sanya, China, Apr. 2008, pp. 920–925.
[7] K. S. Kwok, B. J. Driessen, C. A. Phillips, and C. A. Tovey, “Analyzing the multiple-target-multiple-agent scenario using optimal assignment algorithms,” J. Intell. Robot. Syst.: Theory Appl., vol. 35, no. 1, pp. 111–122, Sep. 2002.
[8] S. X. Yang and M. Meng, “Neural network approaches to dynamic collision-free trajectory generation,” IEEE Trans. Syst., Man, Cybern.,
Part B: Cybern., vol. 31, no. 3, pp. 302–318, Jun. 2001.
[9] S. K. Chalup, C. L. Murch, and M. J. Quinlan, “Machine learning with
AIBO robots in the four-legged league of RoboCup,” IEEE Trans. Syst.,
Man, Cybern., Part C: Appl. Rev., vol. 37, no. 3, pp. 297–310, May
2007.
[10] J. Casper and R. R. Murphy, “Human-robot interactions during the robotassisted urban search and rescue response at the World Trade Center,”
IEEE Trans. Syst., Man, Cybern., Part B: Cybern., vol. 33, no. 3, pp.
367–385, Jun. 2003.