GB2574045A

GB2574045A - A system for detecting air pollution

Info

Publication number: GB2574045A
Application number: GB1808550.6A
Authority: GB
Inventors: Momahed Heravi Behzad
Original assignee: Vortex Iot Ltd
Current assignee: Vortex Iot Ltd
Priority date: 2018-05-24
Filing date: 2018-05-24
Publication date: 2019-11-27
Anticipated expiration: 2038-05-24
Also published as: GB201808550D0; WO2019224545A2; GB2574045B; WO2019224545A3

Abstract

A server 200 for determining air pollution causality is configured to receive data from a sensor unit, analyse the data by applying sensor fusion 202, determine the source of data by applying machine learning algorithm 203 to said data, and analyse the data so as to establish a cause of air pollution 212. The sensor unit comprises a plurality of sensors such as gas, particulate, meteorological, and acoustic sensors. A network or grid of sensor units are configured to send measurement data to a gateway for forwarding to the server. The sensor units may include means to apply sensor fusion and/or machine learning. The arrangement can enable hyperlocal street level air monitoring over a wide area.

Description

In today's society, people are more aware of their surrounding environmental conditions and what contributory factors that may influence the balance of healthy living in their natural environments. Variation of environmental pollution especially caused by air pollution is identified as one of the key challenges that need continuous tackling to improve the health and wellbeing of people, and maintain balance in the environment. The air pollution may be caused by man-made environmental factors and meteorological factors. The major threat to clean air is now caused by vehicle emissions due to increased number of vehicles on the road and frequent traffic congestion in cities. These man-made causes are much more prominent in urban environments and difficult to detect, and can be carried for long distances across large cities. Vehicles that use petrol and diesel emit wide range of pollutants i.e., carbon monoxide (CO), nitrogen oxide (NO), nitrogen dioxide (NO2), volatile organic compounds (VOCs), Ozone (03) and particulate matter (PM1, PM2.5 and PM10) and these pollutants influence condition of the urban air quality.

Meteorological factors variations such as temperature change, wind speed, wind direction, air pressure can directly or indirectly influence the urban air quality and local climate. Smoke from chimneys (fireplaces and wood-burning stoves), factories, vehicles are man-made environmental factors that influence air pollution.

When pollutants contaminate the air quality and environment beyond acceptable levels prescribed by national and international regulatory guidelines, then this adversely impacts human health. Breathing of pollutant contaminated air may result in respiratory diseases for human population. These respiratory diseases can be asthma, lung cancer etc. Release of hazardous gases into the air causes global warming and acid rain. This increases temperature, erratic rains, acid rain, droughts adversely impacting human and animal life on a global scale.

With ever increasing population density in urban environments mean that the air quality is required to be monitored adequately to guarantee that the local population is protected from short term and long term adverse health effects. In urban environments street level air quality is considered as “Hyperlocal street level air quality”, where the people can know the air quality they breath at street level. Daily people are affected by air quality changes and having the air quality information surrounding their local living environment is highly beneficial. It can help people that are vulnerable to respiratory conditions, heart conditions, diabetes, children, pregnant women and elderly.

Brief Description of the Drawings

Figure 1 is a block diagram of the air pollution causality links according to an embodiment of the invention;

Figure 2 is a schematic block diagram of a causality detection system according to an embodiment of the invention;

Figure 3 is a schematic block diagram of a grid according to an embodiment of the invention; and

Figure 4 is a schematic block diagram of an on-street device according to another embodiment of the invention.

Detailed Description

The detailed description set forth below is intended as a description of grind, on street device, system and method; are parts of the deployment system. The appended drawings are incorporated herein and constitute a part of the detailed description. The detailed description includes specific details for the purpose of providing a thorough understanding of the subject and technology.

At present air pollution is measured at a city level, however air quality can change from block to block, hour to hour and day to day, at a hyperlocal level, hyperlocal meaning in this context that the level of air pollution relates to a small geographical area. In order to achieve a more detailed or hyperlocal level of monitoring, an air pollution monitoring system that can be installed in various locations of an urban environment such as streets and that can receive data real time is hereby provided.

The challenges that embodiments of the present invention address, include how to determine what actually causes local variance in air pollution and to quantify contributing factors in air pollutant concentration in a hyperlocal, local and/or urban area. The embodiments also address what measures (effective plans) can be potentially taken to efficiently reduce air pollutant levels in a hyperlocal area. Furthermore, embodiments of the present invention apply an aggregation of causality detection, correlation analysis, spatial statistical inference, sensor fusion and artificial intelligence. Deployment of said air pollution causality detection system can benefit health and wellbeing of local residents, support better city planning in turn having economic benefits locally and nationally. The results can be used by city planners and local authorities to create resilient cities and better-managed urban areas.

An objective of this invention is to provide an air pollution detection system that determines causality (causes and effects), in other words, what causes air pollution in a hyperlocal area, what are the influential factors of high (or low) level of air pollutants and what are their effects on the air quality. This is achieved by establishing hyperlocal monitoring system for air quality by measuring spatial characteristics and variations of airborne pollutants in a city, in order to identify influences on hyperlocal concentrations of airborne pollutants.

Figure 1 shows a method that is utilized for establishing causality. Street level local effects 116 are caused by local causes 100 where the relationships are mapped by causality links 119. The local causes are categorised into environmental factors 101 and meteorological factors 110. The environmental factors 101 comprises dynamic 102 and static 103 contributory factors. Dynamic environmental contributory factors may be traffic flow 104 and vehicles types 105 that are dynamically changing with the varying needs of the environment. Static contributory environmental factors may be street canyons 106, industry density 107, urban density 108 and vegetation density 109. Meteorological factors 110 may be contributory factors humidity 111, air pressure 112, temperature 113, wind speed 114 and wind direction 115. Local effects 116 are caused by local causes 100, particulate matter pollution 117 and gas pollution 118. Although examples of environmental and meteorological factors are described above, it should be understood that the environmental and meteorological factors are not limited to these. It should be also understood that Figure 1 is a representation of what local causes may be and the local effects on air quality.

Figure 2 shows a system for detecting air pollution. The system comprises a cloud-based computing unit 200 which processes data in order to find a link between cause and effect in terms of air pollution in a hyperlocal area or small area such as a street, pathway, park, square, courtyard etc. The cloud-based computing unit 200 receives input or grid data 201 from a grid or network of sensors or sensor units, and input or information of the relevant hyperlocal area stored in a cloud storage 211. The grid data 201 from a grid of sensor units will be described in more detail below, however briefly, it is an aggregation or collection of sensor units distributed in a hyperlocal or a wider area that is configured to measure a variety of factors. For example, the sensor units may include gas sensors, particulate sensors, meteorological sensors and acoustic sensors. Returning now to the information stored in the cloud storage 211, this information is processed by an urban model generator 207, street canyon model generator 209, meteorological data miner 208 and historical data miner 210 before it is being sent to the cloud-based computing unit 200. The urban model generator 207 and the street canyon model generator 209 quantify static environmental factors of the hyperlocal area that is being analysed. For example, the urban model generator 207 processes data to provide the cloud-based computing unit 200 with information of the urban layout of the hyperlocal area. The street canyon model generator 209 processes data to provide the cloud-based computing unit 200 with information on street canyons (a street canyon is a street that is flanked by buildings on both sides creating a canyon-like environment). Meteorological data miner 208 processes data relating to meteorological factors so as to provide the cloud-based computing unit 200 with information of the meteorological information of the hyperlocal area that is being analysed. The cloud-based computing unit 200 further comprises a sensor fusion 202 (sensor fusion module) which receives preprocessed grid data from the grid of sensor units. The sensor fusion 202 may combine data from multiple sensors of a sensor unit and correct for any deficiencies of the individual sensor of the sensor unit in order to calculate more accurate information. In one embodiment, the sensor fusion receives data relating to wind speed and/or wind direction measured by ultrasonic sensors as will be discussed in more detail below. The cloud-based computing unit further comprises an artificial intelligence (Al) algorithm 203 (Al module) which may be a machine-learning module. The grid data 201 and processed data of the sensor fusion 202 may become inputs to the Al algorithm 203, in an alternative embodiment the Al algorithm 203 receives only grid data 201 as input. In one embodiment, the Al algorithm 203 processes grid data relating to acoustic sensors. The Al algorithm 203 can for example be configured so as to determine what type of vehicle (i.e. car, van, lorry, bus, motorbike, diesel, petrol etc.) that has been sensed in a hyperlocal area based on acoustic measurements of acoustic sensors. (Acoustic sensors form part of the grid of sensors which will be explained in more detail below.) In one embodiment, the cloud-based computing unit comprises a sensor fusion 202 and an Al algorithm 203 for each sensor unit. The data processed by the sensor fusion 202 and the Al algorithm 203 is then being processed in order to establish a link between cause and effect which will now be described in more detail. In one embodiment, the data processed by the sensor fusion 202 and the Al algorithm 203 is sent to a correlation analyser and Mutual Information Estimator 204. This component is optional and it identifies candidates of local causes (i.e. environmental factors and meteorological factors as discussed with reference to figure 1). The correlation analyser and mutual information estimator 204 then forwards the information to a causality detection engine 205. If the cloudbased computing unit 200 does not comprise a correlation analyser and mutual information estimator 204 then the sensor fusion 202 and the Al algorithm 203 send their information straight to the causality detection engine 205. The causality detection engine 205 includes Granger Causality and Convergence Cross Mapping (CCM) 205. The Granger Causality is a statistical hypothesis test for determining whether one series of data is useful in forecasting another. CCM 205 is a statistical test for a cause-and-effect relationship between two series variables of data, or two time series variables, that, like the Granger causality test, seeks to resolve the problem that correlation does not imply causation. Once the data has been processed by the causality detection engine the data is sent to a module of spatial statistical inference 206. This module gives the probability of a link between cause and effect. Although it is described above that the flow of processed data is from the correlation analyser and Mutual Information Estimator 204, to the causality detection engine 205 and then to the module of spatial inference 206, it should be understood that data may be sent in the opposite direction or they may share pre or post processed data in any order. Similarly, although data is described as being sent from the sensor fusion 202 to the Al algorithm 203 it should be understood that the correlation analyser and Mutual Information Estimator 204, the causality detection engine 205 and the module of spatial inference 206, may send messages to the sensor fusion 202 and Al algorithm 203 querying received data, or ask for more or alternative data.

In an alternative embodiment, the cloud-based computing unit 200 does not comprise all of the correlation analyser and Mutual Information Estimator 204, the causality detection engine 205 and the module of spatial inference 206. For example, the cloud-based computing unit 200 may only comprise the causality detection engine and the module of spatial inference 206.

Once, the data has been processed, the data output of the cloud- based computing unit 200 may be sent to a predictive model (not shown) which analyses the data in order to determine how the air quality (effect) changes as a result of changing any of the causes (environmental factors, meteorological factors). For example, if you were to reduce the number of cars in the analysed area, what effect would that have on the air quality.

The data output from the cloud-based computing unit provides local causes of air pollution 212 as seen in figure 2. The table below shows an example of the result of each step of the data analysis and the final result (Causal Decision) of the cloud-based computing unit 200.

Cause (501)	Effect (502)	Null Hypothesis (503)	Granger Causality (504)	Convergent Cross Mapping (505)	Causal Probability (506)	Causal Decision (507)
F-Test	Chi-squared Test
		CtoE1 Unidirectional Coupling

C1	E1	C1-Z^E1 Non-coupled with external forcing
C1«E1 Multidirectional Coupling
C1<-(X«Y«Z)->E1 Complex Model
E2




En
C2	E1
E2

En

Cn

The above table shows the cause 501 and effect 502 mapping approach that is taken to determine the dynamics of the contributory causal factors and distinguish their influences on the hyper local pollution at street level. The environmental factors 101 and meteorological factors 110, are the contributory causal factors for this system. The causality detection is governed by 4 types of coupling relationships between cause 501 and effect 502. The null hypothesis 503 column shows the relationship mapping of the four possible couplings: unidirectional coupling, non-coupled with external forcing, multidirectional coupling and complex model. There are multitude of candidate causes that are required to be processed to determine the effect.

To reach an effect E1 from cause C1, it is required to consider 4 coupling types as given in the null hypothesis. Each coupling type is made up of 5 processes: Granger causality with Ftest and Chi-squared test 504, convergent cross mapping 505, causal probability 506 and causal decision 507 to identify the ground truth cause at a street level.

Unidirectional coupling (C1^E1), where the cause C1 unidirectionally is coupled to the effect

E1. It is considered as a direct coupling between C1 and E1 without an intermediary external forcing state (Z). Here the C1 influences the dynamics of E1, but E1 has no effect on C1.

Non-coupled with external forcing (C1^Z^E1), where the cause C1 is not directly coupled with effect E1 and has no direct cross mapping. C1 and E1 are coupled through external forcing (Z) factors and Z is considered as a common factor to coupling type.

Multidirectional coupling (C1*>E1), cause (C1) and effect (E1) are mutually coupled and cross mapped in both directions, which allows each to predict from the other.

In complex model(C1^(X*>Y*>Z)^E1), the cause C1 and effect E1 are coupled through a complex model of mutually interacting external factor (X,Y and Z).

sub processes:

Granger causality 504 as a statistical hypothesis when applied with F-test and Chi-squared test, determines the predictive causality for the system statistically. F-test and Chi-squared test use F-distribution and Chi-squared distribution respectively, applying to captured pollutant data under the null hypothesis and then determine the variances. However Granger causality statistical testing is not limited by F-test and Chi-squared test and may also consider use of other tests. By applying convergent cross mapping to the data, the system determines the deviation between the statistical levels and expected causality levels.

Causal probability establishes the probability of the causality based on the outputs of the convergent cross mapping. A causal decision can use the determined causal probability to identify the ground truth pollutant cause and also provide a pollutant forecast at a street level.

Although not shown in figure 2, it should be understood that the cloud-based computing unit utilises a server or a network of servers comprising processor(s) and memory(ies).

The grid of sensor units which provide grid data to the cloud-based computing unit will now be described with reference to figure 3. The grid of sensor units comprises sensor units 302, 305, 308 positioned in the streets, parks, pathway, courtyard, squares or an urban area etc. The street, park, etc. form a hyperlocal area which may be divided into a grid so as to form a plurality of (hyperlocal) cells, 303, 306, 309. Each cell is monitored by a sensor unit 302, 305, 308. The sensor units may be positioned such that the grid they form is of a mesh or star topology, or it may be of any other suitable random or structured topology. It should be understood that the invention is not limited to a specific number of sensor units but can comprise any desired number of sensor units. Furthermore, it should also be understood that the grid of sensor units can also be interchangeably referred to as a network of sensor units, wherein both terms are not limited to a particular topology nor communication structure. For example, if one sensor unit is removed or faulty then the grid/network of sensor units is still fully functional.

The sensor units detect or measure the air quality. The sensor units are wirelessly communicating with an IP Gateway 312, 313, 314. In one embodiment, sensor units may report their measurements to a main sensor unit which will forward an aggregation of measurements to the IP gateway. In another embodiment, all sensor units report their measurements to the IP gateway. The IP gateway 312, 313, 314 then sends the accumulated grid data to the cloud-based computing unit 200. In one embodiment, the IP gateway 312, 313, 314 sends the grid data to a cloud storage which will store the grid data until the cloudbased computing unit is ready to receive and process the data.

The sensor units 302, 305, 308 may be configured to communicate to the IP gateway via Zigbee, LORAH or SIGFOX, or any other suitable wireless network, the wireless network is denoted 311 in figure 3. The IP gateway may be configured to communicate with the cloud storage or cloud-based computing unit via any cellular network such as GPRS, WIFI, NB-loT, LTE.

In an urban environment there may be multiple cells, or hyperlocal cells, 309, 303 and 306 representing a portion of a street, a whole street or multiple streets. In a city there maybe multiple local grids 315, 316 with multiple wireless networks.

The on street device deployment positions of the sensor units are identified and selected according to the grid as explained above; and has predetermined grid dimensions. The grid is an input in determining the wireless sensor network configuration topology according to deployment requirements (x,y,z) and wireless link budget key performance indicators (KPI). Different information sources are identified, information extracted and processed to generate the hyperlocal grid. Local model generator, historical data miner and meteorological data miner are inputs to the grid generation process.

On street devices, or sensor units, consists of multiple sensors, processor chip with memory, wireless transceiver, power source, firmware and ruggedized enclosure for installation in outdoor and harsh environments. Power source to the on street device can be electric, battery, solar, kinetic, tidal and wind.

Figure 4 shows an on-street device, sensor unit, for sensing air pollutants in accordance with embodiment of various parts. On-board Sensor part 400 comprises of gas sensor 401, particulate sensor 407, meteorological sensor411 and acoustic array sensor 416. Gas sensor may include one or more sensors such as Nitrogen Dioxide (NO2) 402, carbon monoxide (CO) 403, Sulphur dioxide (SO2) 404, Ozone (03) 405 and Volatile Organic Compounds (VOC) 406. Particulate sensor may include one or more sensors such as PM1 sensor 408, PM2.5 sensor 409 and PM 10 sensor 410. Meteorological sensor may include one or more sensors of wind speed sensor 412, wind direction sensor 413, humidity sensor 414 and temperature sensor 415.

Sensors 401,407, 411 and 416 are all combined on a single onboard device (sensor unit) and data generated from these sensors are provided to the processor 421 over the sensor interface 418. Processor 421 has access to memory 420 that provides extra data processing and storage capability for the on-street device. Power source 417 provides power to the processor 421 and on-board sensors 400. Power management 422 provides the power control and management options for on-street device. Onboard processor is loaded with firmware 422 that has device drivers 423, state machines 424, protocols 425 complying with standards. When the sensor data are processed and ready to transmit offboard then the processed data is transferred to Radio Frequency (RF) transceiver 419 , where the RF transceiver establishes external wireless link to a gateway device, for example gateway 312, 313, 314 as described above.

In an alternative embodiment, the sensors 401, 407, 411 and 416 are on separate onboard devices such that there are controlled by separate controllers or processors and have different memories.

In another alternative embodiment, the sensor fusion 202 and Al algorithm 203 as described above in connection with figure 2, are located in each sensor unit rather than the cloud-based computing unit.

Alternative embodiments will now be described.

An air pollution causality detection system that detects and quantifies the causation between hyperlocal air pollution (effect) and environmental (man-made) and meteorological (natural) factors (causes); comprising the subsystems:

• Multi-sensor wireless device that measure air pollutants and meteorological changes for indoor and outdoor installation according to a hyperlocal area deployment grid; and requires multiple devices to establish a wireless sensor network.

• Wireless gateway device that establishes a wireless IOT network for the deployed sensor devices; and requires minimum of one gateway device to establish the network.

• Remote cloud server is a storage and data processing computer wirelessly connected to the gateway device.

• Software consisting of Sensor Fusion and Artificial intelligence (resides on remote cloud server) that process the multi-sensor data of each device individually and gives the measures and information required for Causality Detection.

• Grid-based multi-sensor spatial statistical methods that uses local environmental data (both static and dynamic) to measure spatial statistical moments and quantitative parameters required for causality detection methods.

• Causality detection method that uses the spatial statistical measures and determines a probabilistic model of local air pollution causes. The probabilistic causality model describes the influential factors contributing in local airborne pollutant levels.

The effect of embodiment 1 is that it provides hyperlocal air pollution intelligence at street level.

Embodiment 2

An air pollution causality detection system as set forth in embodiment 1 is an aggregation of sensors, hardware, firmware, power supply options, wireless sensor network, cloud-based computing software, Artificial Intelligence, sensor fusion, spatial statistical inference methods and causality detection methods.

Embodiment 3:

An air pollution causality detection system as set forth in embodiment 1, wherein providing of hyperlocal air pollution heat map by storing and analysing the multi-sensor data.

Embodiment 4:

An air pollution causality detection system as set forth in embodiment 1, wherein providing environmental (man-made) factors include:

Static Factors:

street canyons model generated from local 3D maps, local industry density model generated from local 3D maps and local authority database local population/building density model generated from local 3D maps and local authority database

Local vegetation density model generated from elevation maps.

Dynamic Factors:

Traffic flow estimated from acoustic sensor array

Vehicle speed and direction from acoustic sensor array and doppler effect analysis Vehicle types estimated from Artificial Intelligence methods

Embodiment 5:

An air pollution causality detection system as set forth in embodiment 4, wherein generating street canyons model by processing local 3D maps. Said processing includes determining information of interest, obtaining data associated with both determined information of interest and geo-location of the interest.

Embodiment 6:

An air pollution causality detection system as set forth in embodiment 5, wherein generating local industry density model by processing local 3D maps. Said processing includes determining information of local industry, obtaining data associated with both determined information of interest and geo-location of the interest.

Embodiment 7:

An air pollution causality detection system as set forth in embodiment 5, wherein generating local population and building density model by processing local 3D maps and local authority data. Said processing includes determining information of local population and building density, obtaining data associated with both determined information of interest and geolocation of the interest.

Embodiment 8:

An air pollution causality detection system as set forth in embodiment 5, wherein generating local vegetation density model by processing local elevation maps and local authority data. Said processing includes determining information of local vegetation type, location and density.

Embodiment 9:

An air pollution causality detection system as set forth in embodiment 2, wherein providing quantified traffic flow information by storing and analysing the acoustic sensor array data received from sensor device.

Embodiment 10:

An air pollution causality detection system as set forth in embodiment 8, wherein providing vehicle speed and direction information by storing and analysing the acoustic sensor array data received from sensor device using doppler effect analysis.

Embodiment 11:

An air pollution causality detection system as set forth in embodiment 9, wherein providing vehicle type identification by storing and analysing the acoustic sensor array data using Artificial Intelligence and deep learning methods.

Embodiment 12:

An air pollution causality detection system as set forth in embodiment 4, wherein capturing meteorological factors influencing hyperlocal air pollution by storing and analysing the multisensor data: Humidity, Temperature and Air Pressure.

Embodiment 13:

An air pollution causality detection system as set forth in embodiment 1, wherein each onstreet device comprising of on-board sensors, each sensor communicates with a processor where output data from sensors are fused together. Said on-board sensors comprising: air pollutant sensor, humidity sensor, temperature sensor, acoustic array sensors, wind speed and direction sensor and air pressure sensor.

Embodiment 14:

An air pollution causality detection system as set forth in embodiment 1, wherein wireless sensors are installed according to a grid for the area. Said grid is created from data obtained locally, third party; using available data by collection, aggregation or filtering.

Embodiment 15:

An air pollution causality detection system as set forth in embodiment 1, wherein uses causality detection methods for hyperlocal air pollution using methods: Convergent cross mapping; probabilistic causation using Markov condition; Granger Causality; Mutual Information Estimation.

Embodiment 16:

An air pollution causality detection system as set forth in embodiment 9, wherein a method that find out where to deploy the sensors and to increase the sensor coverage efficiency. It uses to train the machine learning algorithm using data from hyperlocal area. It can use data and models: Data from similar areas, Street canyons Model.

Embodiment 17:

An air pollution causality detection system as set forth in embodiment 1, wherein gateway device provides wireless connectivity to cloud-based computing unit by internet protocol. It acts as an access point to all sensor devices and sensor devices are all connected wirelessly to gateway device.

Embodiment 18:

An air pollution causality detection system as set forth in embodiment 1, wherein cloud-based computing unit stores and processes multi-sensor data utilizing artificial intelligence and machine learning algorithms that are residing in a server computer.

Embodiment 19:

An air pollution causality detection system as set forth in embodiment 1, wherein correlation analysis, causality detection and spatial statistic inference of hyperlocal air pollutant concentration gives a contribution mapping matrix of each air pollution causes as set forth in embodiment 4.

Embodiment 20:

An air pollution causality detection system as set forth in embodiment 2, wherein power is supplied to the on street device by means of tapping to electricity available at street furniture, batteries or solar panels.

Embodiment 21:

An air pollution causality detection system as set forth in embodiment 5, said grid provides on street device deployment positions; input in determining the wireless sensor network configuration topology according to deployment requirements (x,y,z coordinates); Determine wireless link budget key performance indicators for the said grid; Determine minimum strength of a signal received according to the topology; Determine the loss based upon the determined strength.

Embodiment 22:

An air pollution causality detection system as set forth in embodiment 1, take multiple input factors to model causality (cause and effect) and determine quantitative relationship among the factors. This method provides outputs for a sensitivity analysis and forecast the hyperlocal concentration of the airborne pollutants.

Embodiment 23:

An air pollution causality detection system as set forth in embodiment 1 and embodiment 22, hyperlocal real time processed data output from the said methods and sensitivity analysis can provide quantifiable correlation of human exposure levels to pollutants and length of time; can lead to insurance claims and health benefits; local authority resourcing adequately to maintain hyperlocal street level air quality to higher standards.

Further embodiments of the present invention will now be described.

In one embodiment, a server, for example a cloud server, is provided for determining air pollution causality. The server may be configured to receive data from a sensor unit or a network of sensor units. The server may then analyse the data by applying sensor fusion and machine algorithm to said data and so as to establish a link between cause of air pollution and the effect on air quality. This may alternatively be described as the server may process the data by running the data through a sensor fusion module and an artificial intelligence module so as to establish a link between cause of air pollution and the effect on air quality.

The cause of air pollution may be environmental factors for example traffic flow, vehicle types, street canyons, industry density, urban density or vegetation density. The cause of air pollution can also be meteorological factors, for example humidity, air pressure, temperature, wind speed or wind direction.

The link between cause and effect may be explained as the relationship between cause and effect.

The effect may be the gas pollution and/or particulate matter pollution.

In one embodiment, the data comprises ultrasonic sensor measurement and the ultrasonic sensor measurement is analysed by applying the sensor fusion. The ultrasonic sensor measurement may be a measurement of wind speed or wind direction.

In one embodiment, the data comprises acoustic measurement and the acoustic measurement is analysed by applying a machine learning algorithm so as to determine the origin of the data. The origin of data may be considered to be the cause of the acoustic reading. For example, the origin of the data may be determined as a particular type of vehicle or type of engine. The acoustic measurement may be measured using a microphone. An alternative wording for origin could be “source”.

In one embodiment, a statistical test is applied to the data so as to establish the probability of a link between a cause and an effect. In such an embodiment, the server may further comprise a causality engine for applying Granger causality test and/or convergent cross mapping test so as to establish a cause of air pollution, or a probability of a cause of air pollution.

In one embodiment, the server further comprises a module for spatial statistical inference for analysing the data.

In one embodiment, the server further comprises a correlation analyser and mutual information estimator for analysing the data. In another embodiment, the server is further configured to receive data relating to street layout and/or meteorological data. This data may be received from the urban model generator 207, street canyon model generator 209, meteorological data miner 208, historical data miner 210 as discussed above in connection with figure 2.

In another embodiment of the invention, there is provided a network of sensor units for measuring environmental factors and/or meteorological factors, wherein the sensor units are configured to send their measurements to a gateway. The network of sensor units are configured to communicate with a gateway for forwarding the measurement to a server.

In one embodiment, one sensor unit is configured to aggregate measurements from other sensor units of the network and is further configured to sends the measurements to a gateway for forwarding to a server. The sensor units may be positioned in an area wherein the area is divided into cells, and each cell is monitored by a sensor unit. In another embodiment, sensor units are distributed in a hyperlocal area.

In one embodiment, a sensor unit is provided, the sensor unit comprises a plurality of sensors and is configured to analyse the measurements of the plurality of sensors by applying sensor fusion and/or machine learning algorithm similar to the sensor fusion and/or machine algorithm described above in connection with figure 2. This provides the advantage that part of the analysis stage of the data happens in the sensor units rather than in a server or cloud-based computing unit.

A system may also be provided for detecting air pollution. The system comprises a server as described above, or a cloud-based computing unit as described above. The system may also comprise a sensor unit as described above. In one embodiment, the system may comprise a network of sensor unit. The network of sensor unit may be configured according to any of the descriptions mentioned above.

The present invention also relates to methods performed by a server, cloud-based computing unit, network of sensor units and sensors.

The embodiments of the present invention, also relates to a computer program configured, when run on a computer, to carry out a method performed by a server, cloud-based computing unit, network of sensor units and sensors.

It should be noted that the above-mentioned embodiments illustrate rather than limit the invention, and that those skilled in the art will be able to design many alternative embodiments without departing from the scope of the appended claims. The word “comprising” does not exclude the presence of elements or steps other than those listed in a claim, “a” or “an” does not exclude a plurality, and a single feature or other unit may fulfil the functions of several units recited in the claims. Any reference signs in the claims shall not be construed so as to limit their scope.

Claims

1. A server for determining air pollution causality, the server being configured to receive data from a sensor unit, analyse the data by applying sensor fusion, determine the source of data by applying machine learning algorithm to said data, and analyse the data so as to establish a cause of air pollution.

2. A server according to claim 1, wherein the data comprises ultrasonic sensor measurement and the ultrasonic sensor measurement is analysed by applying the sensor fusion.

3. A server according to claim 2, wherein the ultrasonic sensor measurement comprises a measurement of wind speed and/or wind direction.

4. A server according to any preceding claim, wherein the data comprises acoustic measurement and the acoustic measurement is analysed by applying a machine learning algorithm so as to determine the source of the data.

5. A server according to claim 4, wherein the source of the data may be determined as the type of vehicle or the type of engine.

6. A server according to any preceding claim, wherein the server further comprises a causality engine for applying Granger causality test and/or convergent cross mapping test so as to establish a cause of air pollution.

7. A server according to any preceding claim, wherein the server further comprises a module for spatial statistical inference for analysing the data.

8. A server according to any preceding claim, wherein the server further comprises a correlation analyser and mutual information estimator for analysing the data.

9. A server according to any preceding claim, wherein the server is further configured to receive data relating to street layout and/or meteorological data.

10. A server according to any preceding claim, wherein the server is configured to determine a link between the cause of air pollution and the effect on air quality.

10. A server according to any preceding claim, wherein the server comprises a single server or a plurality of servers.

11. A server according to any preceding claim, wherein the server is a virtual server utilising resources from a cloud computing network.

12. A network of sensor units for measuring environmental factors and/or meteorological factors, wherein the sensor units are configured to send their measurements to a gateway.

13. A network of sensor units according to claim 12, wherein the network of sensor units are configured to communicate with a gateway for forwarding the measurement to a server.

14. A network of sensor units according to claim 12, wherein one sensor unit is configured to aggregate measurements from other sensor units of the network and is further configured to send the measurements to a gateway for forwarding to a server.

15. A network of sensor units according to any of claims 12 to 14, wherein the sensor units are positioned in an area such that the area is divided into cells, wherein each cell is monitored by a sensor unit.

16. A network of sensor units according to any of claims 12 to 15, wherein environmental factors may be any of traffic flow, vehicle types, street canyons, industry density, urban density or vegetation density and meteorological factors can be any of humidity, air pressure, temperature, wind speed or wind direction.

17. A sensor unit comprising a plurality of sensors and is configured to analyse the measurements of the plurality of sensors by applying sensor fusion and/or a machine learning algorithm.

18. A sensor unit according to claim 17, wherein sensor fusion is applied to wind speed measurement and/or wind direction measurement, and machine learning algorithm is applied to acoustic measurements.

19. A system for detecting air pollution, the system comprising a server as claimed in any of claims 1 to 11 above and a sensor unit comprising a plurality of sensors.