Updated: Nov 26, 2020
INTRODUCTION / PROBLEM STATEMENT
Earth is our planet and on this planet, we have a lot to explore. But still, there is something around which is affecting a huge part of our planet which includes biodiversity of species especially flora and fauna. It is not only about the individual countries, but it is also happening worldwide, and being a citizen of our respective countries it is our responsibility to save our planet and use emerging technologies to solve real-life problems.
Deforestation and forest fires are one of the major concerns which affect the most precious wealth of our country, endangers the life of plants, animals and even our farmers who work hard over their land. Many wildlife sanctuaries and national parks are also affected by these incidents and endanger both flora and fauna species. People do think about their fire extinguishers in offices, houses, gardens, and other workplaces but we hardly think about those innocent lives which could be affected by major incidents in California and Amazon jungles.
We have many devices available in our market which help to extinguish a fire, but the major problem is that they are not easily available for those National Sanctuaries and Parks. It is very important to do something that can alert us about those fires and a few preventive measures which can be taken before time. The products available in the market with which companies are dealing are based on some individual Modules like simply Bluetooth based, IOT based and even their cost is too high.
Our motive is to make a worldwide feasible device which can be cost-effective and decrease the cost of the fire fighting robot with some more impactful features in a single device. This robot will be Eye Controlled, Voice-controlled, and Manually controlled which will help forest fire stations to keep live track of those forests' weather, climate, and humidity status using algorithms. This device can also be used by people who are physically handicapped or disabled and this would become the best feature of this device.
HARDWARE AND SOFTWARE REQUIREMENTS
Wooden board (High quality)
Software and other components-
LM 35 Temperature sensors- used for detection of the temperature of climate and environment at the site of fire locations.
HC05 Bluetooth module- For communication between the android app, firemen, and robot. The HC-05 Bluetooth Module can be a great solution for wireless communication as it can be used in a Master or Slave configuration. To establish or maintain a connection between MCU and GPS, PC to your embedded project, etc you can use it mainly for a serial port replacement. The HC-05 Bluetooth Module has 6 pins namely-VCC, GND, TX, RX, Key and LED. It is usually pre-programmed in the Slave mode, so there is no need to connect the Key pin, unless or until you need it to change it to Master mode. The main difference between the Master and Slave modes is that the Bluetooth module can’t initiate a connection, it can accept incoming connections in the Slave mode. The Bluetooth module can transfer and receive data irrespective of the mode it is running in after the connection is made. For the connection of a mobile phone to the Bluetooth module, We can simply use it in Slave mode. This module has a default data transmission rate of 9600kbps and the ranges are also predefined for Bluetooth communication which is usually 30m or less than 30m.
L293D Motor driver- It is used to drive the motors. L293D is a typical 16-pin Motor driver IC which allows the DC motor to drive in different directions. It also controls a set of DC motors in any direction simultaneously. It can control two DC motors with a single IC of L293D by the use of a Dual H-bridge Motor Driver integrated circuit(IC). This IC can drive small as well as big motors. Its working principle is based on the concept of H-bridge. It is the kind of circuit that allows the voltage to be flowing in either direction. H-bridge ICs are ideal for driving a DC motor, As we know voltage will need to change its direction for being able to rotate the motor in both clockwise and anticlockwise directions or say 360 degrees. Two dc motors can rotate independently by a single L293D chip where it contains two H-Bridge circuits inside the IC. It is widely used in robotic and AI-based applications for controlling the DC motors due to its smaller size.
Relay circuit- It is used for controlling water pumps. Relay Circuit is an electrically operated device that has a control system and a controlled system. A control system is also called an input circuit or input contactor and the controlled system is also called an output circuit or output contactor that is most frequently used in automatic control of a circuit. It also works as the automatic switch that controls a high-current circuit with a low-current signal. Some advantages of using relay are the lower moving inertia, stability factor, high reliability, and smaller volume. It has a wide number of applications in power protection devices, automation and artificial intelligence technology, sports, reconnaissance and communication, and in electromechanics and power-based electronics devices. A relay contains an induction part that is capable of reflecting the input variable like current, voltage, temperature, pressure, speed, light, power, resistance and frequency, etc to energize or de-energize the connection of a controlled circuit. An actuator module (output) is present inside it. For coupling and decoupling and to isolate input current as well as to actuate the output there is an intermediary part between the input and output part which can be used for the operation execution. The controlled output circuit of the relay will be energized or de-energized when the rated value of the input (voltage, current, and temperature, etc.) must be greater than the critical value.
Node MCU (WIFI- ESP8266)-It provides mobility to the fire fighting robot. It is both software and hardware with an integrated chip which is ESP8266. NodeMCU is an open-source-based firmware and software development board specially made for IoT based applications. It contains firmware that runs on the ESP8266 Wi-Fi SoC from Espressif Systems, and hardware works on the ESP-12 module.
Arduino nano- It is used for making our robot operational and giving commands to our robot. The fire-fighting robot works under the control of the Arduino nano. It is open-source hardware that uses the Arduino IDE that is based on easy-to-use software and hardware. We have a circuit board referred to as a microcontroller that can be pre-programmed according to the user's needs. It consists of ready-made software called Arduino IDE (Integrated Development Environment), in which the computer code can be written and uploaded directly to the physical board using a USB port. Arduino boards have the capability of reading analog or digital input signals from a diversity of sensors and gives output by turning LED on/off, activating the motor, connecting to the cloud, etc. And interfacing can be done. A simplified version of C++ is used in the Arduino IDE that makes it easier to learn the program and its working. It provides a standard way that breaks the functions of the micro-controller into accessible packages and modules.
Filter capacitor- It is used for smoothing voltages. It divides the data into different voltages by using batteries of 12V and divides the voltage respectively into 5V for better efficiency and working in the circuit.
We will make a device that can help us to deal with this real-time situation. The Working methodology for this device will be divided into two different modules. In our first module, we will use machine learning the emerging technology in the field of analytics, and then make a device that will extinguish the fire by again using different modules one is Bluetooth and another one is IOT mode. We can also use this device manually and make it voice-controlled also. We will use MATLAB free source software for the EYE controlled feature of fire fighting robot we will send commands using this software and the robot will move accordingly likewise we can say if we move our eye in the right direction it will move to the right directions and if we direct our eye in the left direction it will move in the left direction and if we blink our eyes three times it will stop automatically. We can also use the concept of GSM which will automatically send the message to a nearby fire station.
The first part will complete work on the dataset of the particular areas of the data, available in the UCI Repositories. And we use different algorithms to find how much area will be burned by Analysing the previous fire data present or area burned previously
Forest Fire Prediction:-
It is the forest fire prediction using Meteorological Data with the help of machine learning algorithms. (SVR, Random Forest, Decision tree, and Deep Neural Networks).
The forest Fire Weather Index (FWI) is the Canadian system for rating fire danger and it includes six components Fine Fuel Moisture Code (FFMC), Duff Moisture Code (DMC), Drought Code (DC), Initial Spread Index (ISI), Buildup Index (BUI) and FWI.
This the dataset inputs which we use and then train our model using different algorithms and predict the area burned in the forest fire prediction.
Forest Fire prediction
It is a Data Mining Approach to Predict Forest Fires using Meteorological Data and find the area burned with better accuracy. This data is easily available at Kaggle and UCI repository.
X - x-axis spatial coordinate within the Montesinho park map: 1 to 9
Y - y-axis spatial coordinate within the Montesinho park map: 2 to 9
month - month of the year: "jan" to "dec"
day - day of the week: "mon" to "sun"
FFMC - FFMC index from the FWI system: 18.7 to 96.20
DMC - DMC index from the FWI system: 1.1 to 291.3
DC-DC index from the FWI system: 7.9 to 860.6
ISI - ISI index from the FWI system: 0.0 to 56.10
temp - temperature in Celsius degrees: 2.2 to 33.30
RH - relative humidity in %: 15.0 to 100
wind - wind speed in km/h: 0.40 to 9.40
rain - outside rain in mm/m2 : 0.0 to 6.4
area - the burned area of the forest (in ha): 0.00 to 109
Firstly we made some scatter plots to better understand the data by comparing and taking logarithms of data and different dependent and independent variables. We made some predictions based on these variables using various numbers of libraries pandas, NumPy, matplotlib. pyplot and os. We use NumPy for statistical computations and pandas for file manipulation, matplotlib is used for effective data visualization.
We Plot the boxplots of how the categorical features (month and day) affect the outcome we find out the extreme values and outliers too.
Regression Error Characteristic (REC) estimation
Receiver Operating Characteristic (ROC) curves are used as one of the powerful tools for visualizing and comparing classification results. Regression Error Characteristic (REC) curves generalize ROC curves to regression. ROC curves plot the error tolerance on the 𝑥−𝑎𝑥𝑖𝑠 versus the percentage of points predicted within the tolerance on the 𝑦−𝑎𝑥𝑖𝑠. The resulting curve estimates the cumulative distribution function of the error. The REC curve visually presents commonly-used statistics. The area-over-the-curve (AOC) is a biased estimate of the expected error. The 𝑅2R2 value can be estimated using the ratio of the AOC for a given model to the AOC for the null-model. Users can quickly assess the relative merits of many regression functions by examining the relative position of their REC curves. The shape of the curve reveals additional information that can be used to guide modeling.
We use the grid search for Finding the right parameters for machine learning models. Scikit-learn has the functionality of trying a bunch of combinations and see what works the best, built-in with GridSearchCV. The CV stands for cross-validation.
GridSearchCV takes a dictionary that describes the parameters that should be tried and a model to train. The grid of parameters is defined as a dictionary, where the keys are the parameters and the values are the settings to be tested.
Four algorithms used-
2. Random forest
3. Decision Tree
4. Deep Neural network
1. SVR (Support vector regression)
Support Vector Machine is one of the discriminative algorithms that always tries to find the optimal hyperplane that distinctly classifies the data points in N-dimensional space(N - the number of features). In a two dimensional space, a hyperplane is a line that optimally divides the data points into two different classes. In a higher-dimensional space, the hyperplane would have a different shape rather than a line.
2. Random Forest
Random forest is like one of the bootstrapping algorithms with the Decision tree (CART) model. Suppose we have 1000 observations in the complete population with 10 variables. Random forest tries to build multiple CART models with different samples and different initial variables. For instance, This will take a random sample of 100 observations and 5 randomly chosen initial variables to build a CART model. It will also repeat the process (suppose)10 times and then make a final prediction on each observation. The final prediction is a function of every prediction. This final prediction can simply be the mean of each prediction.
3. Decision Tree
A decision tree is the mapping of the possible outcomes of a series of related choices. It allows an individual or organization to weigh possible actions against one another based on their costs, probabilities, and benefits. They can also be used either to drive informal discussion or to map out an algorithm that predicts the best choice mathematically.
A decision tree typically starts with a single node, which branches into possible outcomes. Each of those outcomes leads to additional nodes, which branch off into other possibilities. This gives it a treelike shape.
4. Deep Neural Networks
Building Neural Network
Keras is a very simple tool for constructing a neural network. It is a high-level framework that is based on TensorFlow, theano, or cntk backends.
Sequential, which is specified to Keras, is for creating the model sequentially and the output of each layer we add will be input to the next layer we specify.
model. add is used to add an input layer to our neural network. We need to specify our argument on what type of layer we want to add to our neural network. The Dense is used to specify the fully connected layer. The activation function to be used is relu in this case.
RESULTS OBTAINED USING SVR REGRESSOR
RESULTS OBTAINED USING DECISION TREE REGRESSOR
RESULTS OBTAINED FROM RANDOM FOREST AND DEEP NN
FINAL CONCLUSION BASED ON COMPARISON OF DIFFERENT MODELS
We used all the models using the macintosh park dataset.
Finally, we compared all the graphs and got the maximum efficiency using the deep neural networks model using Keras as the backend with maximum accuracy, stability, and consistency in the deep neural networks can be shown in the above-mentioned ROC curves.
This is the best way to reduce the cost and make firemen and people more independent. This will help the firemen to protect the life of people who are in danger during fire incidents. And they can also take preventive measures to protect the people around the national parks and wildlife sanctuaries.
Future scope and Applications
Complete and automatic control of the fire fighting robot.
Camera and video transmission can be added.
Improve the weight capacity of the robot.
Make it an eye, voice, and manually controlled device.
Further use of some more algorithms.