INTRODUCTION TO THE SYSTEM:
Basic Concept the System Initial Setting
With the Industrial civilization, a lot of things around the world have changed drastically. Most of these things have changed for good but some of them resulted in pure harm. One of these problems is, the air pollution. In many parts of the world it is a highly critical problem. Many of the world's most polluted cities have been declared as a gas chamber, e.g.: New Delhi (India), Zabol (Iran), Bamenda (Cameroon) and many more. According to WHO, the air in these cities contain dangerously high level of particulate matter, tinny in size to enter the blood stream through the lungs, this problem contributes to an estimated 7 million premature deaths each year. Hence, a problem like this should be addressed at an international level and should be prioritized. To address this problem, many indoor air purifiers are made which focus on the air around the humansspecifically.But,usingindoorpurifierswon'tsolvetheissueofobliteratingairconditions in the ecosystem. This problem can only be dealtby
fixing the main cause rather than by fixing the smaller parts which don't even contribute in curing the problem as a whole. Keeping this in mind, in this air purifier, works on an outdoor air purifier. This purifier will function by suppling fresh or filtered air in an outdoor environment. The purifier has three main integrated functions to perform. These functions are interconnected with each other. They are:
1. To move from one place (where the pollution is less) to another place (where the pollution is
high)
2. To purify the air at thatplace.
3. To purify the air around pedestrians by moving aroundthem

The system mode!
All these functions which are mentioned above in the previous topic will be performed by using the concepts of Reinforcement learning with deep Q network, Arduino, deep learning algorithms like the YOLO algorithms.
Hence, this air purification system can be thought of as an autonomous mobile system, whichuses the reinforcement learning for the movement and is capable of reducing the pollution not only efficiently but alsoeffectively.
Reinforcement Learning for Obstacle Avoidance
The Deep RL network used
In this whole process of reinforcement learning using the deep neural networks is included. The environment is the place where the air purifier is working, this environment can be the apartment lanesoranycolonylanes,theactionoftheagentcanbetomovefromaplaceoflowpollutiontoa place of high pollution also it can be the purification of the air. The state can be any point at a particular time step. The rewards given to the agent can be to accurately travel to a place of higher pollutionandalsoaccuratelyjudgetheairqualityindexofaplaceandthenpurifytheairsuchthat there is a significant difference in the air quality index and that too for good[10][12].
Architecture
Since,thesituationdealswithaprobleminwhichanautonomousrobothastoavoidalltheobjects initspathwayandthattooinverymuchuncertainsituations.So,therearealargenumberofstate- action pairs to deal with. Hence, Pavan uses the Q-learning approach to store all these state action pairs.Nowduringthetrainingthespeedofthewholeprocessmightgetslowassomenoisewillbe added. The main reason for this noise is that, initially and at any point of updation process in the neural network does not know about the Q% this is only known after some episodes. In order to remove the noise and increase the efficiency the experience replay is used or need to record the robot's prior experience at every step, return using the new value of the Q to generate our training update[5].

YOLO Algorithm
The YOLO algorithm in this Pavan is used in the following way[8]:
Stepl. It helps to detect the passengers or any other object in the way veryefficiently. Step2. Once the passenger or the object is detected, its aim should be to avoid thatobstacle. Step3. This is the step where the YOLO connects with the Reinforcement learning concept as
mentionedabove. Step4. Once the object is detected the work of YOLO algorithm isover.
YOLO algorithm is specially used because of its efficiency and it detects the object accurately. Since, the air purifier will be used in real world situation it is important that it doesn't bumps into any other object. This way the YOLO algorithm helps in enhancing the security purpose of the mobile robot.
In Pavan, an updated YOLO algorithm is used. They have used this for- Detecting the Pedestrian using YOLO Network Model. In their paper, in addition to the original YOLO network, 3 pass-through layers were added. These layer consist:
1. Route layer: Which is used to transfer the information regarding the pedestrian of specified layer to currentlayer.
2. Reorg layer: which is used to identify the feature map. This is done so that the route layer which is just introduced is matched with the feature map of nextlayer.
Thismethodlinksthehighandlowresolutionpedestrianfeatures.Itpassestheshallowfeaturesto thedeepnetworkforsomeeffectivelearning.Alongwithallthesechanges,thelayernumberswere alsochangedfroml6tol2,inordertomodifyandsupportthecapabilityofthenetworktoextract information. The basic process of the algorithm is given in the belowdiagram.
Arduino for air purifications
Once the air purifier stops at a point while navigating due to the increase in the air quality index of that place, then, it starts to purify the air there. This purifier has gas sensors which detect the impurities in the air. This sensor uses the Arduino which is coded. The impure air is given as an inputtothesensorandthentheoutput,resultsinthedetectionoftheimpuritiesintheair.Asand when it detects the air, displays the air quality index of that place and if that value is harmfulfor

inhalingthenitringsabuzzerwhichtellsthatthepollutionishighanditistimetostartthesystem. This whole model is a collaboration of the air quality monitoring device + filter consisting of manylayers.
The Monitoring part:
The monitoring part includes, gas sensor, temperature and the humidity sensor. When all these are calculated they are displayed on the LCD screen[2].
• Gas sensor: Helps to detect propane in the carbon content, hydrogen, LPG leakage and other products which arecombustible.
• Temperature and humidity sensor: For this we are using DHT11. It uses the capacitive humiditysensorandthermostatictocheckthesurroundings.Thisprovideswithaveryhigh speed of the collection of the data which is 2sec. The area coverage in order to sense the temperatureandhumidityrangesfromzerototwentymeters.Therangeofthetemperature is zero-fifty°C.
• Arduino: It is a microcontroller based LCD display which consists of ATmegs328. It has 6 analog pins + 8 digital pins. Since it takes input as analog signals, hence it is much beneficial for our purifier as all the inputs are taken as analog signals (gas sensor, temperature and humiditysensor.
• LCD Display: It is liquid crystal display. It displays the temperatures, humidity and other readings.
• Indicators: We are using LED bulks as indicators and a buzzer. The LED's used are the red and the greenones.
The Filtering part:
In the filtering process, a number of filters are used to ensure that the air is properly cleaned. The different filters which we are using are [2]:
• HEPA filter: It is high efficiency particulate air filter. It is used for collecting the particles of dust from theatmosphere.
• UV layer: The second layer is a UV layer. It uses UV lamps through the help of whichit removes the fungus and kill other diseases which are caused by the bacteria in theair.

• Silica Layer: The third layer which used is the Silica layer. It helps to absorb the moisture thereby reducingit.
• Activated carbon layer: The last layer used is the activated carbon layer which is used to reduce that carbon dioxide and other dangerous gases from theair.
Fans are also used in the air purifier. It has two functions:
a) taking air into it andthen
b) Releasing it via many layer and then blowing pure air out through thefilter.
SOFTWARE REQUIREMENT SPECIFICATION. Specific Requirements (Tools and Techniques usectt
A. Hardware Requirements:
RAM: 10GB or more
Processor: Any Intel i7
Arduino
LCD display
Temperature and humidity sensor
Gas sensor
Indicators
HEPA filter
UV layer
Silicalayer
Activated carbon layer
B. SoftwareRequirements:
Operating system: Windows/Linux Technology: Deep learning, Tensorflow, keras Base Language: Python
CONCLUSION AND FUTUREWORK
InPavantheworkwehavedoneiscomprisedofthelatestupgradesinthefieldofdeepIearning and
reinforcement learning which solves one of the biggest problem the world is dealing with, "Air pollution". The work is inspired from the "Deepmind technologies". We gave a different andbetterapproachfromalltheexistingmethods,ourapproachincludedtheuseofthedeepQ-networks.Theobjectiveofthisairpurifieristominimizethepollutionofaparticularoutdoor

place by conducting three functions simultaneously. The mobile automated Robot's work will be to move from a place of less pollution to a place where there is more pollution and then start to purify the air there. While doing these two functions the Robot's aim should be to detect pedestrians and to move around them so that the air around them is clean. The technologies which are combined are the, Reinforcement learning, deep Q network, Yolo algorithm (which has its base from the neural networks) and the Arduino system for sensing the pollutants in the environment. In this way we aim to build a mobile autonomous air purifier which can help to purify the outdoor air.
REFERENCES
[I] Lai, Yang, M. J. (2013, June). Study and development of a negative ion driving circuit.
In Consumer Electronics (ISCE), IEEE 17thISon{pp. 39-40) 13MEEE.
[2] Masuda, Matsuda, N. (1993) The performance of an integrated air purifier for control of
aerosol, microbial, and odor. IEEE transactions on industryapplications. [3] Tsai, P. (2016, May). IOT: IB-TS for smart home. In (ICASI), (pp. 1-4).IEEE. [4] Delgado, Flor, H. (2017, October). Selection of the best air purifier system to urban houses
using AHP. In (CHILECON), 2017 CHILEAN (pp. 1-4).IEEE. [5] Fan, Jingxue, Z. (2016, May). Research of indoor mobile biological air purifier pedestrian
tracking system. In CCDC), 2016 Chinese (pp. 1573-1577).IEEE. [6] Qijun&Zhonghui,L.(2017,June).DesignofcontrolsystemforaninteIligentairpurifierand
sweeper combined robot. In 2017 12th IEEE (ICIEA)(pp. 227-230).IEEE. [7] Huang, & Guo, M. (2005, August). RL neural network to the problem of autonomous mobile
robot obstacle avoidance. In Machine Learning and Cybernetics, 2005. IC on (Vol. 1, pp. 85-
89). IEEE. [8] Lan, W & Wang, S. (2018, August). Pedestrian Detection Based on YOLO Network Model.
In 2018 IEEE (ICMA) (pp. 1547-1551).IEEE. [9] Sharma, &Bachhars A. (2017, October). I2P air purifier. In(ICCES), 2017 2nd Internationa!
Conference on (pp. 478-481).IEEE. [10] Glasius, &Gielen, S. C. (1995). Neural network dynamics for path planning and obstacle
avoidance. Neural Networks, 5(1), 125-133.
[II] Kim & Shin, W. G. (2017). ACPEAP using an activated carbon fiber filter for passenger
cars. IEEE Transactions on Industry Applications, 53(6),5867-5874.
[12] Chen, &Juang, J. G. (2009, August). Intelligent obstacle avoidance control strategy for wheeled mobile robot. In ICCAS-SICE, 2009 (pp. 3199-3204).IEEE.

CLAIMS
1. It is an outdoor air purifier which uses the deep learningtechnology.
Pavan'sdeepneuraltechnologymaximizesthefreshhealthyairthecitizensbreatheinanapartment or a
colony of just around a home anytime anywhere and can just smartly adjust to the uniqueness of
theenvironment.
This air purifier can be suited in the corridors of a building and the streets of a colony or an
apartment.
2. It is a cost effective mobile airpurifier.
It is an ambulatory mobile robot. That is, it moves from one place to another place by simultaneously cleaning the air. The Combination of the reinforcement learning approach with the deep Q networks approach is used, this will help the robot with the mobility part. In this part, the robot's function is to go from a low pollution area to a very high pollution area without getting smashedintoanyotherobject.lnthisway,thesoleaimofthisoperationwillbetoavoidanobstacle. Hence all these facilities are provided by this air purifier in a way that is cost effective.
The details of cost effectiveness are given in the tablel. We claim cost effective because as compared to other smart air purifiers, eg: Dyson's air purifier which lie in the range of 54k to 60k5 our outdoor air purifier provides an efficient solution at a much lesscost.
3. It purifies the air around the pedestrians by moving aroundthem,
Pavan also possesses the technology by which it can detect any pedestrian around it or near to it and then it aims to roam around the area where there are most of the pedestrians so that it can provide them with a really good quality of air to breathe. A classic example of this would be an area or a podium where children play and people walk.