Air quality metrics

Abstract

The technology is generally directed to using a collection of consumer devices to create a virtual sensor for a geographic region. The sensors may be air quality sensors. The collection of consumer devices may include small sensors, such as those typically owned and installed by users. Rather than measure the particulate counts of one or more types of pollutants, the data received from the sensors may be calibrated to provide the mass of the pollutant(s) per volume. By calibrating the sensor data and combining the calibrated sensor data from sensors within the geographic region, accurate air quality metrics may be determined. The air quality metrics may be a current or predicted air quality metric. The current and / or predicted air quality metrics may be used to identify a clean navigational route, e.g., a route with a lower expected exposure metric, identify locations for urban planning, or the like.

Description

GOOGLE-4087 AIR QUALITY METRICS CROSS REFERENCE TO RELATED APPLICATIONS

[0001] The present application claims the benefit of the filing date of U.S. Provisional Application No.63 / 619,764, filed January 11, 2024, entitled Air Quality Metrics, the disclosure of which is hereby incorporated herein by reference. BACKGROUND

[0002] Air pollutants are substances in the air that can have negative effects on humans. The substances can be solid particles, liquid droplets, gasses, and so on. An air pollutant can be of natural origin or man-made.

[0003] Although air pollution is a well-known phenomenon worldwide, the problem is that people are usually unaware that the air they breathe may harm them. Ironically, there are many cases where people can immediately improve the air quality in their close environment by simple actions.

[0004] While some air pollution monitoring systems exist, these systems are typically designed to measure a specific type of pollutant in a specific geographical region, and thus are limited in their applicability to other types of pollution that may be found in other regions. Moreover, because of their specific designs, measurements from one type of air pollution monitoring system are often not compatible or consistent with measurements from other types of air pollution monitoring systems. BRIEF SUMMARY

[0005] The technology is generally directed to using a collection of consumer devices to create a virtual sensor for a geographic region. The collection of consumer devices may include small sensors, such as those typically owned and installed by users. The sensors may be air quality sensors. Rather than measure the particulate counts of one or more types of pollutants, the data received from the sensors may be calibrated to provide the mass of the pollutant(s) per volume. By calibrating the sensor data and combining the calibrated sensor data from sensors within the geographic region, accurate air quality metrics may be determined. The air quality metrics may be a current or predicted air quality metric. The current and / or predicted air quality metrics may be used to identify a clean navigational route, e.g., a route with a lower expected exposure metric, identify locations for urban planning, or the like.

[0006] One aspect of the technology is directed to method, comprising receiving, by one or more processors from a plurality of sensors, sensor data f, calibrating, by the one or more processors, at least a portion of the sensor data, receiving, by the one or more processors, the sensor data including a least the portion of the calibrated sensor data as input into an artificial intelligence (AI) model trained to predict air quality metrics, wherein the AI model corresponds to a virtual sensor for the geographic region, and determining, by the one or more processors based on the virtual sensor, an air quality metric for the geographic region. The plurality of sensors may include at least one air quality sensor. One or more of the plurality of sensors may be within a geographic region. At least a portion of the sensor data is for one or more pollutants.

[0007] Calibrating the sensor data may include translating the sensor data for the one or more pollutants from particle counts of the one or more pollutants to mass per volume of the one or more pollutants. Generating theGOOGLE-4087 virtual sensor may be further based on an algorithm. The virtual sensor may comprise one or more layers. Each layer of the one or more layers may correspond to a type of sensor data. Each layer of the one or more layers may be a model trained to predict a preliminary air quality metric based on the type of sensor data.

[0008] The method may further comprise identifying, by the one or more processors based on the calibrated sensor data, one or more data outliers. When identifying the one or more data outliers, the method may further comprise comparing, by the one or more processors, the calibrated sensor data for a given air quality sensor in the geographic region to average calibrated sensor data for the geographic region. When the calibrated sensor data for the given air quality sensor is more than a threshold amount above the average calibrated sensor data, the calibrated sensor data may be identified as an outlier. When the calibrated sensor data for the given air quality sensor is more than a threshold amount below the average calibrated sensor data, the calibrated sensor data may be identified as an outlier.

[0009] The method may further comprise predicting, by the one or more processors executing a second artificial intelligence (AI) model, a future air quality metric for the geographic region. The method may further comprise receiving, as input into the second AI model, the determined air quality metric for the geographic region, at least one of a time or a day in the future, and one or more of a future time of day, current weather for the geographic region, predicted weather forecast for the geographic region, and environmental factors, and providing as output, by the one or more processors executing the AI model, the future air quality metric. The environmental factors may include one or more of wind speed, wind direction, fires within a radius of the geographic region, earthquakes within the radius of the geographic region, and tornadoes within the radius of the geographic region.

[0010] The method may further comprise receiving, by the one or more processors, a navigation request including a starting location and a destination location, identifying, by the one or more processors, a first navigational route from the starting location to the destination location, the first navigational route having a first accumulated exposure metric, identifying, by the one or more processors, a second navigational routes from the starting location to the destination location, the second navigational route having a second accumulated exposure metric, comparing, by the one or more processors, the first accumulated exposure metric with the second accumulated exposure metric to determine which of the first route or second route represents a cleaner route, and providing, by the one or more processors, navigation instructions relating to the cleaner route. Determining the first accumulated exposure metric or the second accumulated exposure metric may comprise aggregating, by the one or more processors, one or more air quality metrics for regions along the respective first or second navigational routes. Aggregating he one or more air quality metrics may comprise determining, by the one or more processors, a respective air quality index for each type of pollutant along the respective first or second navigational routes, determining, by the one or more processors, a worse one of the respective air quality index among the types of pollutants along the respective first or second navigational routes, and identifying, by the one or more processors, the worse one of the respective air quality indexes for the respective first or second navigational routes as the respective first or second accumulated exposure metric.GOOGLE-4087

[0011] The method may further comprise receiving, by the one or more processors, a geographic planning request for a geographic region, identifying, by the one or more processors based on the geographic planning request, available locations within the geographic region, determining, by the one or more processors, air quality metrics of one or more discrete regions within the geographic region, wherein the available locations are within at least one of the discrete regions, comparing, by the one or more processors, the air quality metrics of the one or more discrete region, and identifying, by the one or more processors based on the comparison, a suggested available location, wherein the discrete region associated with the suggested available location has a lower numerical value for the air quality metric as compared to other discrete regions. The geographic planning request may include one more or more planning criteria. The one or more planning criteria may include at least one of intended use, rent, sale, price, property size, zoning requirements, and parking requirements.

[0012] Another aspect of the technology is directed to a system comprising one or more processors. The one or more processors may be configured to receive, from a plurality of air quality sensors within a geographic region, sensor data for one or more pollutants, calibrate the sensor data, generate, based on the calibrated sensor data, a virtual sensor for the geographic region, and determine, based on the virtual sensor, an air quality metric for the geographic region.

[0013] Yet another aspect of the technology is directed to one or more non-transitory computer readable media encoding instructions that, when executed by one or more processors, cause the one or more processors to perform operations. The operations may comprise receiving, from a plurality of air quality sensors within a geographic region, sensor data for one or more pollutants, calibrating the sensor data, generating, based on the calibrated sensor data, a virtual sensor for the geographic region, and determining, based on the virtual sensor, an air quality metric for the geographic region. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 is an example screenshot illustrating an example interface providing air quality metrics according to aspects of the disclosure.

[0015] Figure 2 is an example map indicating the location of sensors and a geographic region according to aspects of the disclosure.

[0016] Figure 3 is a block diagram of an example air quality prediction model according to aspects of the disclosure.

[0017] Figure 4A is an example screenshot illustrating cleanest routes identified based on air quality metrics according to aspects of the disclosure.

[0018] Figure 4B is a chart of example air quality indexes for different types of pollutants according to aspects of the disclosure.

[0019] Figure 5 is an example heat map overlay providing an indication of air quality metrics for a geographic region according to aspects of the disclosure.

[0020] Figure 6 is a block diagram of an example clean route model according to aspects of the disclosure.GOOGLE-4087

[0021] Figure 7A is an example map identifying available locations within a geographic region according to aspects of the disclosure.

[0022] Figure 7B is an example map identifying suggested transportation routes within a geographic region according to aspects of the disclosure.

[0023] Figure 7C is an example map identifying recommendations for improving air quality metrics within the geographic region according to aspects of the disclosure.

[0024] Figure 8A is a functional diagram of an example system according to aspects of the disclosure.

[0025] Figure 8B is a pictorial diagram of the example system of Figure 8B according to aspects of the disclosure.

[0026] Figure 9 is a flow diagram of an example method for determining an air quality metric for a geographic region according to aspects of the disclosure. DETAILED DESCRIPTION

[0027] The technology is generally directed to using information from discrete sensors located within a geographic region to generate air quality metrics. The sensors may be small sensors that are part of a consumer device. In some examples, the sensors may be owned and installed by users at their home, their car, etc. In some examples the sensors may be part of a personal computing device, such as a smartphone, watch, helmet, etc. The sensors may measure the number of particulates in the air for a specific pollutant. The pollutant may be any solid or liquid suspended in the air, such as smoke, dust, combustion byproduct, pollen, water vapor, etc. In some examples, the pollutants may be ammonia, odor, nitrogen dioxide, ground-level ozone, particulates, ozone, sulfur dioxide, carbon monoxide, and benzene. The sensor data may be calibrated for measuring the concentration of a given pollutant to the mass of the pollutants in the air as using mass, instead of particle concentration, provides a more accurate representation of the air quality. The calibrated sensor readings from the sensors located within the geographic region may be aggregated and extrapolated to determine an accurate representation of the air quality in the geographic region. The representation of the air quality may be, for example, an air quality metric.

[0028] According to some examples, the air quality metric determined based on sensor readings within a geographic area may be used for navigation. For example, using the air quality metric for a given geographic location, a starting location, and a destination, an accumulated exposure metric may be determined. The accumulated exposure metric may be determined based on the air quality metrics in the geographic regions along a navigational route between the starting location and a destination. In some examples, an alternative navigational route having a lower accumulated exposure may be identified. The alternative route may be determined by an artificial intelligence (“AI”) model, such as a machine learning (“ML”) model. For example, the AI may be trained to identify, based on the air quality for geographic regions and known roads and travel passages, an alternative navigational route having less accumulated exposure than the originally suggested route. For example, the originally suggested route may be the route having the shortest distance and / or time of travel between the starting location and the destination, regardless of the accumulated exposure along theGOOGLE-4087 route. The alternative route identified by the AI model may have a lower accumulated exposure as compared to the originally suggested route. The AI model may provide, as output, the alternative route and the predicted accumulated exposure along the route. A navigational application may provide, as output to the user, the originally suggested route and the alternative route.

[0029] In some examples, the air quality determined based on sensor readings may be used for urban planning. Urban planning may include, for example, identifying possible locations for buildings, factories, or the like, identifying possible locations for public transportation, public transportation routes, etc. According to some examples, the AI model may be trained to identify possible locations for urban planning purposes based on a planned addition, e.g., building, transportation route, etc., and the air quality within a geographic region. The identified possible locations and / or routes may distribute air pollutants more equally within the geographic region. For example, rather than suggesting that multiple factories be built side by side, the AI model may identify a candidate location for each factory such that the factories are spread apart within a geographic region, thereby distributing the air pollutants within the geographic region.

[0030] Collecting data from sensors throughout a geographic region allows for a more robust and accurate model of the air quality metrics to be determined. For example, rather than relying on the data taken from a single sensor or from a single location, the air quality metrics are determined based on sensor data, publicly available information, satellite data, weather data, and / or traffic data. The use of multiple sources of data allows for outliers to be determined and disregarded when determining the air quality metrics for a region. Being able to identify outliers provides for a more accurate determination of the air quality for the region.

[0031] Further, by generating a virtual sensor based on sensor data collected from a plurality of sources, air quality metrics can be determined in real time, based on, at least in part, real time data obtained from the sensors. For example, rather than solely relying on governmental sources, weather data, or the like, which are either predictions of what is going to happen or is outdated as the data represents what has already happened, the virtual sensor further uses real time data from sensors integrated into user devices, home devices, vehicles, or the like. The data from these sensors is shared in real time, allowing for the air quality metrics to be determined in real time as well. Determining Air Quality

[0032] Figure 1 is an example screenshot illustrating example air quality metrics. Air quality metrics may include air quality readings based on one or more indices, such as the air quality index (“AQI”). The air quality metrics may be determined based on sensor data from a given geographic region.

[0033] The sensor data may be data from a plurality of sensors. The sensors may be, for example, sensors purchased and set up by users, governmental sensors, mobile sensors, such as those in a smart device or vehicle, or the like. In some examples, the sensors may be small in size such that the sensors are integrated with other devices, such as smartphones, watches, doorbells, helmets, cars, motorcycles, bikes, buses, trains, electric bikes, electric cars, etc. For example, device 102 may be a smartphone with one or more sensors 108. While sensor 108 is shown as located at the bottom of device 102, sensor 108 may be integrated anywhere on deviceGOOGLE-4087 102 that allows sensor 108 to capture the requisite data for determining one or more air quality metrics. While the examples discussed herein are based on the sensors 108 being integrated in a smartphone, the sensors 108 may be integrated in any device or vehicle and, therefore, the example of a smartphone is not intended to be limiting.

[0034] According to some examples, the sensors may be air quality sensors. The air quality sensors may be configured to determine a concentration of a given pollutant in the air near the sensor. The pollutant may be any solid or liquid that is suspended in the air. For example, the pollutant may be smoke, dust, dander, soot, pollen, sea salt, spores, water vapor (e.g., humidity), fumes, combustion byproducts (e.g., exhaust from a vehicle), or the like. The concentration may be, for example, a measure of the number of counts, or particles, for the given pollutant. The number of particles of a pollutant, however, does not provide an accurate representation of the air quality. For example, a user would not be able to determine whether the air quality is worse if there were 1,000 particles of smoke as compared to 1,000 particles of dust or 1,000 particles of water vapor. A more accurate way to determine the air quality is to determine the mass of the pollutants per volume, e.g., micrograms per cubic meter. In some examples, the mass of the pollutants per volume may be compared to an air quality index that will provide an indication of how the air quality can affect a user’s health.

[0035] The geographic region may be, for example, determined based on a location 104, such as a current location of a user of device 102. The geographic region may, in some examples, be a neighborhood, town, city, county, state, country, etc. based on the location 104.

[0036] According to some examples, the device 102 may include a plurality of sensors. For example, the device 102 can include an air quality sensor for particles, an air quality sensor for gasses, a light sensor, humidity sensor, temperature sensor, precipitation sensor, or the like. The data obtained by each sensor of the plurality of sensors can be used to generate a virtual air quality sensor. For example, each type of sensor, e.g, an air quality sensor for particles, an air quality sensor for gasses, a light sensor, humidity sensor, temperature sensor, precipitation sensor, etc., can be a different layer, model, or engine within the air quality prediction model 300, discussed with respect to Figure 3, below. For example, each layer may be an AI model, such as a machine learning model, trained to predict air quality metrics based on the respective type of sensor data (e.g., weather, traffic, air quality, publicly accessible sensor data, weather, wind, temperature, humidity, precipitation, light, satellite, etc.). The layers may be executed in parallel or substantially in parallel such that the air quality prediction model 300 can use the outputs of each layer to generate a prediction of air quality metrics for the location 104. The output of the air quality prediction model 300 may be one or more air quality metrics for the location 104, corresponding to a virtual sensor for location 104.

[0037] Figure 2 illustrates an example map indicating the location of sensors and a geographic region. Sensors 220 may be located throughout a geographic area. For example, as illustrated on map 200, a plurality of sensors “S” are located throughout the mapped geographic area. For clarity purposes, only one sensor “S” is identified with reference number 220. However, each “S” represents a sensor within the geographic area defined by map 200.GOOGLE-4087

[0038] The sensors “S” may be sensors on user devices, vehicles, or the like. In some examples, some of the sensors “S” may be on vehicles, such as cars, bikes, electric vehicles, buses, trains, etc., that are moving through the geographic area defined by map 200. The data from the sensors “S” moving through the area may be collected, or obtained, by the system and used to determine air quality metrics for the area. In some examples, as explained above and herein, the data from sensors moving through the area may be used to determine, or predict, pollutants and, therefore, air quality, based on the traffic within the area.

[0039] A geographic region 222 may be within the geographic area defined by map 200. The geographic region 222 may be determined based on the location 204 of the user. In some examples, the geographic region 222 may be determined based on a location provided by the user. As illustrated in Figure 2, the geographic region 222 may be defined based on a threshold distance “X” from location 204. In such an example, the geographic region 222 may be defined by the threshold distance “X”, e.g., radius, centered on location 204.

[0040] According to some examples, the geographic region 222 may be defined based on other properties. For example, in response to receiving an input corresponding to the selection of “X”, location 104 in Figure 1, or another input, a pop-up 224 may be provided for output on the display of device 102. While illustrated as a pop-up 224, pop-up 224 may be provided as an overlay, a new screen output, or the like. The pop-up 224 may provide one or more options for defining the geographic region 222. For example, pop-up 224 may include one or more inputs corresponding to selections for defining the geographic region 222. The inputs may include check boxes, as shown, radio buttons, a drop down list, or the like to select a method for defining the geographic region 222. The methods may include, for example, selecting a threshold distance “X” or defining the geographic region based on geographical boundaries, such as by neighborhood, city, state, or the like. In some examples, the threshold distance “X” may be changed such that the geographic region 222 may be increased or decreased.

[0041] The sensor data from sensors within the geographic region 222 may be used to determine one or more air quality metrics. The sensor data may be received after confirming that users have provided authorization for the sensor data to be shared with the system. For example, the user may provide authorization to a website, application, or system when setting up the sensor. The authorization may be for the application or system to access, store, track, or the like the sensor data captured by the sensor. In addition, certain data may be treated in one or more ways before it is stored or used, so that personally identifiable information is removed. The user may have control over what information is collected about the user, how that information is used, and what information is provided to the user.

[0042] In some examples, the sensor data from sensors within the geographic region 222 may be obtained by scraping the internet for publicly accessible sensor data, connecting to publicly available interfaces, or the like. For example, the system may obtain sensor information from governmental sensors in which the data collected and / or predictions made based on the data are publicly available. However, the information obtained from the publicly accessible sensor data, publicly available interfaces, or the like is typically outdated by the time the data is obtained by the system. The system may include an artificial intelligence (AI) model, such as a machineGOOGLE-4087 learning (ML) model, that can predict what the actual sensor reading would be at the present time or at a future time. The prediction may be, for example, the concentration level of one or more pollutants at a given time at the location of the sensor. The predictions made by the AI model for a given sensor may be provided as input into another AI model, as discussed further in here.

[0043] According to some examples, other publicly available information may be obtained and used to determine air quality metrics. For example, data from weather sensors that detect the wind, temperature, humidity, precipitation, etc. may be collected and provided to one or more models for determining, or predicting, air quality metrics. The weather data may be scraped from the internet, collected by connecting to publicly available interfaces, collected by sensors integrated within user devices, vehicles, or the like, satellites, etc. In some examples, data from pollution dispersion models, traffic information for the geographic region 222, etc. may be obtained and provided as input into models.

[0044] In some examples, an AI model may be trained to translate, or convert, traffic data to air quality data. The AI model for translating traffic data to air quality data may be another layer in the air quality prediction model 300. For example, the model may be trained to determine air quality metrics based on a severity of a traffic jam, a speed at which vehicles are moving on a road, the number of lanes on a given road, the typical type of vehicles driving on the road, the pollutants output by the vehicles on the road, etc. The model may be trained based on data obtained by sensors at or near the traffic. The data obtained by the sensors may be used to determine an ambient, or typical, air quality metric for the location, road, etc. As a traffic jam occurs, the data obtained by the sensors may be used to determine an air quality metric for that location, road, etc. during the traffic jam. The difference between the air quality metric during the traffic jam and the ambient air quality metric corresponds to the air pollution that can be contributed to the traffic jam. As an example, if the ambient air quality metric is a ten (10) and then the air quality metric during the traffic jam is fifteen (15), then the air quality metric attributed to the traffic jam is five (5).

[0045] The air quality metric attributed to the traffic jam may be associated with characteristics of the traffic jam. The characteristics can include, for example, a distance of the traffic jam, an amount of time the traffic jam lasts, the characteristics of the vehicles in the traffic jam (e.g., combustion engine, electric vehicles, buses, etc.), and the like. The characteristics associated with the traffic jam and the associated air quality metric attributed to the traffic jam may become training pairs, or a labeled dataset, used to train the model for translating traffic data to air quality data. The model can use the training pairs or labeled data sets to predict air quality metrics that can be attributed to traffic jams in a given location. In some examples, the output of the model may be used by the air quality prediction model 300. In another example, the output of the model may be a heatmap that can be overlaid on a geographic map of the location, road, etc. The heatmap may provide an indication of the air quality metrics on the roads based on the traffic data and / or other sensor data.

[0046] Referring back to Figure 1, the sensor data from sensors 220 within the geographic region 222 may be used to determine a current air quality metric 106. The sensors 220 may, for example, be air quality sensors in a consumer device, such as home devices and smart devices. Air quality sensors in consumer devices areGOOGLE-4087 typically laser based sensors that measure the particulate counts of the pollutants. For example, the sensors may measure the pollutant based on the particular matter (PM), rather than the mass of the pollutant. The PM is measured based on the size of the pollutant. The sensors may measure the particulate counts of one or more types of pollutants based on the size, or PM, of the pollutant. For example, the sensors may determine the particulate count of water vapor, pollen, dust, smoke, etc. However, knowing the count of the pollutant does not provide an accurate representation of the dangers associated with the pollutant as larger pollutants, e.g., pollutants with PM 10, can typically be filtered by the human nose whereas smaller pollutants, e.g., pollutants with PM 2.5, are typically not filtered by the human nose and can enter into the bloodstream. To provide a more accurate determination of the air quality in the given geographic region, the sensor data received from the sensors may be calibrated to provide the mass of the pollutant(s) per volume, rather than the particulate count. For example, for each type of pollutant detected by the sensor, the particulate count of that pollutant may be calibrated to provide the mass of that pollutant per volume. As each particle of the pollutant has its own mass, the count of that pollutant can be converted to mass based on the specific, or typical, mass associated with that pollutant. The volume may be, for example, a cubic meter. In some examples, the calibrated sensor data, e.g., the translation from counts to mass per volume, may be measured in micrograms per cubic meter (µg / m3).

[0047] The calibrated sensor data may be combined. For example, for each pollutant, a mean, average, median, or mode of the calibrated sensor data may be determined.

[0048] In some examples, outliers of the calibrated sensor data for a given pollutant may be identified and not included when determining air quality metrics. For example, calibrated sensor data may be identified as an outlier if the mass per volume is too high and / or too low as compared to the other sensor data. For example, a sensor may be included in a smart doorbell. A bonfire may be burning in close proximity to the smart doorbell, thereby providing a higher mass per volume of smoke near that particular sensor as compared to the geographic region as a whole. The sensor data from that smart doorbell may be identified as an outlier if the mass per volume of smoke is above a threshold beyond the average mass per volume of smoke for the other sensors in the geographic region.

[0049] The combination, or compilation, of calibrated sensor data may generate a virtual sensor for the location 104 and / or geographic region 222 encompassing location 104. According to some examples, the virtual sensor for the location 104 and / or geographic region 222 may include data from publicly available sources, in addition to the sensors within the region. For example, the publicly available sources may be sensor data from governmental sensors, weather data, satellite data, traffic data, etc. The sources may be publicly available due to the data being accessible via the internet, an interface in which the user and / or system can consent to data sharing during setup or initialization, or the like. Each set of sensor data, e.g., user sensor data, governmental sensor data, weather data, traffic data, etc., may correspond to a layer within the AI model used to determine air quality metrics. According to some examples, the layers within the model may be a respective model trained to predict different concentrations of different pollutants based on a given type of sensor data.GOOGLE-4087 The predictions from each layer may be combined, or aggregated, into an air quality metric for the location 104 and / or geographic region 222.

[0050] The calibrated sensor data for sensors within geographic region 222 may be combined and used as a virtual sensor to determine the current air quality metric 106. Referring back to Figure 1, the virtual sensor may provide an indication of the current air quality 106 for the location 104 and / or geographic region 222 encompassing location 104 based on aggregate calibrated sensor data. The aggregated calibrated sensor data may be, for example, the air quality metric prediction generated by one or more models.

[0051] The current air quality 106 may be provided as output as a numerical score, e.g., “11 AQI”, a relative score, e.g., “Good”, and / or or as an indication along an AQI representation 112, e.g., a number line. Providing the current air quality in one or more formats, e.g., numerical score, relative score, on a representation, or the like, allows for a user to readily determine whether the air quality is acceptable or dangerous.

[0052] The virtual sensor may leverage sensor data from a plurality of sensors 220 to determine the air quality. Leveraging sensor data may improve the accuracy of the air quality determination as the virtual sensor uses data from all over the geographic region rather than a single sensor in one location within the geographic region. By using a plurality of air sensors throughout a geographic region to determine an air quality metric for that region, the system beneficially determines the air quality for the region with greater accuracy. The use of a plurality of sensors to generate a virtual sensor ensures that the air quality metric for a given geographic region is based on samples taken across the region, rather than in a particular location within the region. Moreover, the accuracy of the virtual sensor is increased, as compared to using a single sensor, by removing outlying sensor data that may be caused based on the placement of the sensor. Predicting Future Air Quality Metrics

[0053] The air quality metrics 102 may, in some examples, include predicted metrics 110. For example, current air quality metrics 106 may be used to predict future air quality metrics 110. Future air quality metrics 110 may, for example, be a forecast of predicted air quality metrics for a given time in the future as compared to the current air quality metric 106. An AI model may be trained to predict what the air quality for a given geographic region is going to be in the future, e.g., in one-hour, three-hours, tomorrow, etc.

[0054] Figure 3 depicts a block diagram of an example air quality prediction model 300, which can be implemented on one or more computing devices. The air quality prediction model 300 can be configured to receive inference data and / or training data for use in predicting air quality metrics for a given location or geographic region. The predicted air quality metrics may be for a time in the future as compared to when the prediction was made. For example, the predicted air quality metrics may be for a time one hour, two hours, one day, etc. from when the prediction was made. For example, the air quality prediction model 300 can receive the inference data 330 and / or training data 332 as part of a call to an application programming interface (API) exposing the air quality prediction model 300 to one or more computing devices. Inference data 330 and / or training data 332 can also be provided to the air quality prediction model 300 through a storage medium, such as remote storage connected to the one or more computing devices over a network. Inference data and / orGOOGLE-4087 training data can further be provided as input through a user interface on a client computing device coupled to the air quality prediction model 300.

[0055] The inference data 330 can include data associated with predicting air quality metrics for a geographic region for a time in the future. The inference data 330 may include, for example, current air quality metrics 106, sensor data from sensors 220 within the geographic region 222 and / or around the geographic region 222, current weather data, predicted weather data, or the like. For example, if the weather data indicates that winds are moving air west to east with gust of up to 6 knots, the inference data 330 may include sensor data from sensors to the west of the location 104 and / or geographic region 222 as the air from the west is likely going to move towards the location 104 and / or geographic region.

[0056] The training data 332 can correspond to an artificial intelligence (AI) learning task, such as a machine learning (“ML”) task for predicting air quality metrics for a time in the future for a given geographic location. The training data 332 can be split into a training set, a validation set, and / or a testing set. An example training / validation / testing split can be an 80 / 10 / 10 split, although any other split may be possible. The training data 332 can include examples for predicted air quality metrics. The training data 332 may include, for example, historical air quality determinations, the relationship between weather and air quality, environmental factors, human produced factors, the current air quality determination, etc. Historical air quality determinations may be, for example, the air quality for a given geographic region as determined by the virtual sensor, historical determinations made by governmental or publicly available sensors, or the like. The relationship between weather and air quality may be, for example, pairs of data providing an association between the weather, e.g., temperature, humidity, precipitation, or the like, and the determined air quality. Natural environmental factors may include, for examples, wildfires, earthquakes, hurricanes, tornados, etc., Human produced factors may include, for example, combustion due to traffic, combustion due to business operations, e.g., factory operations, pollutants from fireworks, smoke from a bonfire, exhaust from a bar-b-que, etc.

[0057] The training data 332 can be in any form suitable for training a model, according to one of a variety of different learning techniques. Learning techniques for training a model can include supervised learning, unsupervised learning, and semi-supervised learning techniques. For example, the training data can include multiple training examples that can be received as input by a model. The training examples can be labeled with a desired output for the model when processing the labeled training examples. The label and the model output can be evaluated through a loss function to determine an error, which can be back propagated through the model to update weights for the model. For example, if the machine learning task is a classification task, the training examples can be images labeled with one or more classes categorizing subjects depicted in the images. As another example, a supervised learning technique can be applied to calculate an error between outputs, with a ground-truth label of a training example processed by the model. Any of a variety of loss or error functions appropriate for the type of the task the model is being trained for can be utilized, such as cross-entropy loss for classification tasks, or mean square error for regression tasks. The gradient of the error with respect to the different weights of the candidate model on candidate hardware can be calculated, for example using aGOOGLE-4087 backpropagation algorithm, and the weights for the model can be updated. The model can be trained until stopping criteria are met, such as a number of iterations for training, a maximum period of time, a convergence, or when a minimum accuracy threshold is met.

[0058] From the inference data 330 and / or training data 332, the air quality prediction model 300 can be configured to output one or more results related to predicted air quality metrics for a given time in the future generated as output data 334. In some examples, the air quality prediction model 300 may output one or more results based on the type of data provided as input. For example, the air quality prediction model 300 may determine a predicted air quality metric based on user sensors, vehicle sensors, traffic data, weather data, satellite data, publicly available sensor data, etc. Each output, or prediction, may be, for example, a layer within the air quality prediction model 300. The air quality prediction model 300 may use the output of each layer to then predict the air quality metrics for a given time, e.g., the current time or a time in the future, for a geographic region. The predicted air quality metric, based on the output of each layer, may be an aggregated air quality metric for the geographic region.

[0059] As examples, the output data 334 can be any kind of score, classification, or regression output based on the input data. Correspondingly, the AI or machine learning task can be a scoring, classification, and / or regression task for predicting some output given some input. These AI or machine learning tasks can correspond to a variety of different applications in processing images, video, text, speech, or other types of data to predict air quality metrics for the geographic region for a future period of time.

[0060] As an example, the air quality prediction model 300 can be configured to send the output data 334 for display on a client or user display. For example, referring to Figure 1, the output data 334 may be the predicted metrics 110. As another example, the air quality prediction model 300 can be configured to provide the output data as a set of computer-readable instructions, such as one or more computer programs. The computer programs can be written in any type of programming language, and according to any programming paradigm, e.g., declarative, procedural, assembly, object-oriented, data-oriented, functional, or imperative. The computer programs can be written to perform one or more different functions and to operate within a computing environment, e.g., on a physical device, virtual machine, or across multiple devices. The computer programs can also implement functionality described herein, for example, as performed by a system, engine, module, or model. The air quality prediction model 300 can further be configured to forward the output data 334 to one or more other devices configured for translating the output data into an executable program written in a computer programming language. The air quality prediction model 300 can also be configured to send the output data 334 to a storage device for storage and later retrieval.

[0061] According to some examples, when executing the air quality prediction model 300, the location 104 and / or geographic region 222 along with the current air quality for said location 104 and / or geographic region 222 may be provided as input to the air quality prediction model 300. In some examples, the time of day, weather, and / or environmental factors may be provided as input. Based on the inputs received by the air quality prediction model 300, the air quality prediction model 300 may provide, as output, a prediction of the futureGOOGLE-4087 air quality. As illustrated in Figure 1, the prediction of the future air quality may be provided as a forecast of predicted air quality metrics 110. Cleanest Routes

[0062] Figure 4 illustrates an example screenshot of cleanest routes identified based on air quality metrics. For example, a navigation application 440 may identify a cleanest route and / or one or more alternative routes for traveling from a starting location 404 to a destination location 442. The starting location 404 may be the current location of the user “U” or any location provided as input as the starting location 404. The destination location 442, represented by destination “D” on map 400, may be any location provided as input into destination location 442.

[0063] By identifying the cleanest route from the starting location 404 to the destination location 442, the user can make a well informed decision to reduce their accumulated exposure metric for a given trip. As some trips may be repetitive in nature, such as a commute to work, weekly trips to the food store, or the like, reducing a user’s accumulated exposure metric can increase the overall long term health of the user. Moreover, in examples where the cleanest route has a reduced carbon footprint, identifying the cleanest route may mitigate increasing the accumulated exposure metric along the route.

[0064] Any use of location information, travel history, vehicle information, and / or travel metrics of a user is authorized by the respective user. For example, the user may provide authorization to an application for identifying cleanest routes by setting certain permissions for the application. The authorization may be for the application to access one or more databases or sub-databases in the memory of the device, vehicle, remote server, etc. According to one example, the user may select specific sub-databases to which the application is granted access. For instance, the user may grant access to the location history database but not the calendar archive database.

[0065] The navigation application 440 may identify one or more navigational routes between the starting location 404 and the destination location 442. The navigational routes may be identified and / or filtered based on one or more travel metrics. The travel metrics may include, for example, time of travel, distance of travel, travel expenses, carbon footprint, accumulated exposure metric, or the like.

[0066] According to some examples, the navigational application 440 may identify a route as being the “cleaner route.” A cleaner route 440 may be, for example, a navigational route between a starting location 404 and a destination location 442 in which the accumulated exposure metric along the route is lower than an initial navigational route 446. For example, the initial navigational route 446 may be the route between the starting and destination location having the shortest travel distance, the shortest travel time, etc. The cleaner route 444 may be the navigational route between the starting and destination location with a lower accumulated exposure metric than the initial route while still fulfilling other navigational filters, such as shortest travel distance, shortest time of travel, most fuel efficient, etc.

[0067] The accumulated exposure metric may be determined based on aggregating one or more air quality metrics for geographic regions along the navigational routes. In some examples, the accumulated exposureGOOGLE-4087 metric may be determined based on the exposure to different concentrations of pollutants along the navigational routes. For example, an AI model, such as a machine learning model, or another type of model may be trained to receive sensor data from the sensors within a graphical region encompassing the starting location 404 and the destination location 442. The model may analyze the sensor data to identify the different types of pollutants, such as Ammonia, odor, Nitrogen Dioxide (NO2), ground-level Ozone (O3), particulates (e.g., PM2.5 and / or PM10), Ozone, Sulfur Dioxide (SO2), Carbon Monoxide (CO), Benzene, combinations therebetween, or the like. The model may then determine, based on the possible routes between the starting location 404 and the destination location 442 an estimated exposure time to each of the different types of pollutants. The exposure time for each pollutant may correspond to a value or metric, as shown in Figure 4B.

[0068] The accumulated exposure metric may be determined based on the worst, e.g., most hazardous, air quality index value for a given pollutant that the user would be exposed to during a given navigational route. For example, the model may determine the value or metric of each pollutant based on the amount of time the user would be exposed to said pollutant. The model would then determine the air quality index value based on a given pollutant. The air quality index value corresponding to the accumulated exposure metric would be the pollutant that results in the worst, e.g., most hazardous, air quality index value w.

[0069] According to some examples, the accumulated exposure metric may be determined based on the mode of transportation being used to travel from the starting location “U” to the destination location “D”. For example, the accumulated exposure metric may be different based on whether the user is in a car, on a bus, on a bicycle or motorcycle, on a train, etc. In some examples, the interface shown in Figure 4A may include an input configured to allow a user to select the mode of transportation. In yet another example, the interface may include one or more routes for alternative transportation. For example, if the accumulated exposure metric for a bus route to reach the destination location “D” is less than the accumulated exposure metric for driving a vehicle, the interface may provide, as output, the bus route as the cleanest route.

[0070] In some examples, the cleanest route may be determined based on the users within the mode of transportation. For example, the cleanest route may be determined based on specific types of pollutants that are known to be problematic for a specific user within the mode of transportation. As part of the navigation application and / or as part of the system as a whole, the user may have a profile indicating problematic pollutants. For example, a first user may indicate that they have bad seasonal allergies. The cleanest route model 600 may be provided with this information as input such that the cleanest route model 600 may identify the cleanest route that will route the first user away from places with a high pollen count, even if it means routing the first user through another region with poor air quality metrics due to another pollutant, e.g., a region with factory combustion fumes. In contrast, if a second user profile indicates that the second user has asthma, the cleanest route model 600 may identify a cleanest route that routes the second user away from the factory combustion fumes even though the route may go through an area with high pollen count, as the high pollen count may not affect the second user as much. In some examples, the system may automatically determine the users in the vehicle based on their account, profile, or the like. Additionally or alternatively, the interface mayGOOGLE-4087 include one or more inputs to add users and / or profiles, input health conditions and / or problematic pollutants, or the like. In such an example, the cleanest route model 600 may identify the cleanest route based on the collective conditions for the users within the vehicle.

[0071] According to some examples, based on the accumulated exposure metric, the interface may provide, as output, a preventative action. The preventative action may, in some examples, reduce the risk of encountering an air quality metric above a predetermined threshold. In some examples, the preventative action may decrease the accumulated exposure metric. The preventative action may include, for example, a proposed travel time to reach the destination location, a suggestion to delay exiting the mode of transportation, turning on an air purifier inside of the mode of transportation, closing one or more windows of the mode of transportation, turning on an air conditioning unit of the mode of transportation, modifying a circulation of the air conditioning of the vehicle to circulate internal air within the mode of transportation, or the like.

[0072] The preventative action may be determined based on the accumulated exposure metric for the route, the current air metrics, and / or predicted air metrics. For example, the mode of transportation may be predicted to encounter air pollution at a future location, based on the air pollution data along the driving path thereof. The driving path of the vehicle may comprise a future location that the vehicle is expected to arrive thereto at a future time, which may be predicted to encounter air pollution. In some exemplary embodiments, an arrival time of the vehicle to each location on the driving path, may be determined. An estimated air quality metric at each location on the driving path, at the associated arrival time thereto may be predicted. The prediction may be performed based on historical pollution data along the driving path, based on monitored real-time pollution levels along the driving path, or the like. Locations having an air quality metric above a predetermined threshold may be determined to encounter air pollution and be risky to passengers of the vehicle. The predetermined threshold may be set in accordance with the demographic characteristics of the passengers of the vehicle, based on health characteristics of the passengers, may be manually set by a user, a driver or a passenger of the vehicle, or the like. Additionally or alternatively, the predetermined threshold may be set based on the lowest air quality metric within the vehicle along the driving path. Additionally or alternatively, the predetermined threshold may be set in an absolute manner, such as based on recommendations of authorities, or the like.

[0073] In some examples, the air preventative action may be configured to prevent the air pollution predicted at the future location to reach a passenger cabin of the vehicle. The preventative actions may be performed at least a predetermined time or distance prior to the vehicle reaching the at least one future location. The predetermined time may be a time period large enough to prevent the air pollution from reaching the passenger cabin, such as about 10 minutes, about 5 minutes, or the like. Additionally or alternatively, the predetermined time may be determined based on the preventative action and the time it is predicted to take, the severity of the air pollution, or the like.

[0074] In some examples, the preventative action may include turning on an air purifier inside the vehicle, closing one or more windows of the vehicle, turning on an air conditioning of the vehicle, modifying aGOOGLE-4087 circulation of the air conditioning of the vehicle to circulate internal air within the vehicle, or the like. Additionally or alternatively, the preventative action may comprise delaying an arrival of the vehicle to the location with the air pollution, until the air quality metric is reduced, such as by slowing down the vehicle, making a stop at a station or location with a lower, e.g., healthier, air quality metric, or the like.

[0075] In some examples, the vehicle may be predicted to arrive to a location with enhanced air quality, based on the air pollution data along the driving path thereof. The driving path may comprise a future sub-path that the vehicle is expected to arrive at a future time with an estimated air quality metric below a threshold. The threshold may be based on an air quality metric at a passenger cabin of the vehicle. As an example, the threshold may be equal to the air quality metric at a passenger cabin, below the air quality metric at a passenger cabin, or the like, in order to enable reducing the air quality metric in the passenger cabin. In some examples, the air quality metric at the passenger cabin may be determined based on sensor readings within the vehicle. Additionally or alternatively, the air quality metric at the passenger cabin of the vehicle may be determined using one or more sensors, such as air quality sensors. The sensors may be integrated into a user device, the vehicle, or the like. In some examples, the sensors may be connected to a network such that the sensor data can be used to determine one or more air quality metrics.

[0076] According to some examples, the system may automatically enact one or more of the preventative actions. For example, the system may detect that there is a traffic jam along the navigation route and automatically provide a suggested alternate route. The suggested alternative route may be a route in which the air quality metrics are lower, e.g., healthier, than staying in the current route with the traffic jam. In some examples, the system may automatically turn on an air filter, such as a high efficiency particulate air filter, change the air circulation from external to internal, close the windows, or the like. The system may automatically enact one or more of the preventative actions based on predicted air quality metrics along the route. For example, if the system determines that the air quality metric for the current location of the vehicle or user is within a threshold level corresponding to a category of “good” or better but predicts that the air quality for a future location along the route is worse, the system may enact the preventive actions. For example, upon the user being within a threshold distance or time of the future location, the system may automatically enable the air filter, change the type of air circulation, suggest an alternative route, close the windows, or the like.

[0077] The sensor data may be received after confirming that users have provided authorization for the sensor data to be shared with the system. For example, the user may provide authorization to a website, application, or system when setting up the sensor. The authorization may be for the application or system to access, store, track, or the like the sensor data captured by the sensor. In addition, certain data may be treated in one or more ways before it is stored or used, so that personally identifiable information is removed. The user may have control over what information is collected about the user, how that information is used, and what information is provided to the user.GOOGLE-4087

[0078] According to some examples, in the absence of sensors integrated with and / or within the vehicle, air quality metrics at a passenger cabin may be determined based on one or more air quality metrics of similar vehicles on the same driving path, such as vehicles from the same fleet of the vehicle that share information with the vehicle, crowdsourced air quality metrics from passengers from similar vehicles with the same driving path, or the like. The vehicle and the other vehicle providing the air quality metrics may be required to have the same ventilation status. As an example, the other vehicle may be driving over the same driving path, sharing the same properties such as size and window shape, performing the same ventilation habits, such as driving with an opened / closed window, with the same A / C activation and circulation, or the like.

[0079] In some examples, a time window at when the vehicle is expected to be located on the future sub-path may be determined. Estimated air quality metrics at the future sub-path within the time window may be predicted. In some exemplary embodiments, a plurality of estimated air quality metrics along several locations within the future sub-path at respective expected arrival times thereto may be predicted. The estimated air quality metrics may be determined based on the plurality of estimated air quality metrics along the several locations.

[0080] According to some examples, the ventilation action may be performed during the time that the vehicle is located on this future sub-path, in order to decrease the air quality metric inside a passenger cabin of the vehicle. The ventilation action may comprise performing at least one of: opening one or more windows of the vehicle, turning on an air conditioning of the vehicle, modifying a circulation of the air conditioning of the vehicle to circulate external air into the vehicle, or the like. The ventilation action may be terminated before the vehicle exits the future sub-path.

[0081] The preventative actions and the ventilation actions may be similar, opposites, or the like. As an example, when a preventative action is performed, the associated ventilation action may be terminated. Additionally or alternatively, the preventative actions and the ventilation actions may be simultaneously and automatically performed or terminated based on the air quality metrics inside and outside the vehicle. As an example, windows may be opened or closed automatically according to the current air quality metrics at the current location. When a window is open, internal air quality may reach the external air quality over time. If the external air quality metric is above the threshold, or above the internal air quality metric, the windows may be closed automatically. If the external air quality metric is below the internal air quality metric, the windows may be opened automatically. Additionally or alternatively, the windows may be automatically opened if the external air quality metric is both lower than the internal air quality metric and below a predetermined threshold. As another example, air conditioners (A / C) can also affect air quality. A / C can use external air in its operation, and accordingly may affect the air quality in a similar manner to a window. In some cases, the A / C may be activated or turned off automatically according to the current air quality metrics inside and outside the vehicle. Additionally or alternatively, the A / C may be instructed to change its circulation property, to circulate the air in or out of the vehicle, automatically according to the current air quality metrics. Similarly, the vehicle windows may be opened, A / C turned on or circulation turned off, when the external air quality isGOOGLE-4087 higher than the internal air quality, in the current location of the car and in the succeeding locations along the driving path of the car, so as to allow the ventilation of the car, to increase the internal air quality. As yet another example, an air purifier inside the vehicle may be automatically turned on or off in accordance with the air quality metric. When the vehicle is predicted to drive through a polluted area, prior to reaching the polluted area, the air purifier may be automatically turned on. Simultaneously, the windows may be automatically closed, the A / C may be automatically turned off or modified to circulate the internal air, before the vehicle reaches the polluted area.

[0082] In some exemplary embodiments, a user, such as the driver of the vehicle, the owner, the passengers, or the like, may be notified of relevant information regarding the air pollution. As an example, the user may be notified when the vehicle parks in a polluted area, and accordingly may avoid exiting the vehicle. As another example, the user may be notified when the vehicle drives through a polluted area (or about to drive through), when the vehicle exits the polluted area, or the like. As yet another example, special notifications may be issued to special passengers, such as vulnerable passengers, children, elderly, pregnant women, people with heart and respiratory diseases, or the like. Additionally or alternatively, a recommendation for air filter replacement may be issued based on the vehicle's actual exposure to pollution.

[0083] It may be noted that the air quality metrics may be determined based on the amount of one or more pollutants in the air, different types of pollutants, pollutants of different sizes, or the like. Different air pollution measurements may indicate different types of air pollution, such as indicating different types of pollutants (such as Ammonia, odor, Nitrogen Dioxide (NO2), ground-level Ozone (O3), particulates (e.g., PM2.5 and / or PM10), Ozone, Sulfur Dioxide (SO2), Carbon Monoxide (CO), Benzene, combinations therebetween, or the like. Additionally or alternatively, Different air pollution measurements may indicate different subtypes of atmospheric particles having different diameters, such as between 2.5 and 10 micrometers (μm) (PM10), fine particles with a diameter of 2.5 μm or less (PM2.5), ultrafine particles, or the like. In some exemplary embodiments, different pollutants may affect different people differently. Accordingly, in some exemplary embodiments, the air quality metric in the passenger cabin, the air quality metric in the future sub-path, the relevant threshold, or the like, may be determined to be in the same type of measurement (e.g. the same pollutants, the same PM, or the like). The type of the measurement may be determined based on demographic characteristics of the passengers of the vehicle, based on health characteristics of the passengers, or the like.

[0084] The use of preventative actions and / or ventilation actions may transform the composition of the air inside a vehicle to reduce pollen. For example, the preventative and / or ventilation actions automatically change the composition of the air inside the vehicle, based on outdoor air quality metrics in real time without direct human intervention.

[0085] Figure 5 illustrates an example heat map overlay providing an indication of the air quality metric for the geographic regions 550, 552, 554, 556, 558. The air quality metric for a given region 550, 552, 554, 556, 558 may correspond to a pattern or overlay on map 500. For example, each pattern may correspond to a range of air quality metrics. As one example, a first pattern may correspond to a range of 0-50 AQI, a second patternGOOGLE-4087 may correspond to a range of 51-100, a third pattern may correspond to a range of 101-150, and so on. The ranges may be determined based on ranges provided by environmental agencies or the like. In some examples, rather than providing a pattern, the regions may be color coded, where each color corresponds to a range of air quality metrics, a level of concern associated with the air quality metrics, or the like. In another example, the air quality metrics may be provided as a numerical score overlaid on map 500. Accordingly, the pattern overlay on map 500 indicating the air quality metrics for a given region is just one example and is not intended to be limiting.

[0086] The air quality metrics for the geographic regions may be used to determine an overall accumulation exposure for a given navigational route. For example, if an initial navigational route 446 guides a user through geographic regions 550, 552, 554 the accumulated exposure metric may be determined based on the air quality metrics for geographic regions 550, 552, 554. The accumulated exposure metric may be, in some examples, an average of the air quality metrics for the geographic regions along the route. In some examples, the accumulated exposure metric may be determined based on the air quality metric for the geographic region, the predicted amount of travel time through the geographic region, or the like. The predicted amount of travel time through the geographic region may be determined based on the distance of the route that travels through the geographic region, known speed limits in the geographic region along the route, predicted traffic along the route, or the like.

[0087] According to some examples, as explained above with respect to Figure 4B, the accumulated exposure metric may be determined based on the exposure to different types of pollutants for a given amount of time along the navigational route.

[0088] In some examples, the accumulated exposure metric may be an indication of health related effects for traveling along the navigation route. The health related effects may be, for example, short term health effects, such as difficulty breathing or reduced visibility, or long term effects, such as potential for diseases. According to some examples, the health related effects may be determined based on the frequency of traveling the navigational route. For example, the system may determine the frequency the user travels between the starting location and the destination location, e.g., between a user’s home and an address designated as “work.” Based on the frequency of traveling the navigational route and the air quality metrics along the route, the system may determine potential health related effects. For example, if a user travels along the navigational route twice a day for five days a week and there is a factory producing combustion byproducts, continued and / or excess exposure to the byproducts may have a long term health effect.

[0089] According to some examples, the alternative route, e.g., cleaner route 444, may be determined based on one or more route criteria, transportation options, and the air quality metrics for geographic regions between the starting and destination location. For example, an AI model may be trained to identify one or more alternative routes based on known transportation options, available roads, route criteria, etc. The transportation options may include, for example, public transportations such as buses, trains, and subways, as well as ride shares, bicycle paths, or the like. The AI model may receive, as input, the starting location 404, destinationGOOGLE-4087 location 442, and one or more route criteria. The route criteria may include, for example, minimum travel distance, minimum travel time, maximum fuel efficiency, avoid toll roads, use of car, use of public transportation, or the like. The AI model may identify, based on the received input, alternative routes, e.g., cleaner route 444, having a lower accumulated exposure as compared to the initial route. In some examples, the lower accumulated exposure may be determined based on the air quality metrics for the geographic regions the cleaner route 444 travels through. For example, as shown in Figure 5, as compared to initial route 446, cleaner route 444 travels through geographic regions 550, 554. The accumulated exposure for traveling through geographic regions 550, 554 may be less than the accumulated exposure for traveling through geographic regions 550, 552, 554.

[0090] Figure 6 depicts a block diagram of an example clean route model 600, which can be implemented on one or more computing devices. The clean route model 600 can be configured to receive training data 660 for use in identifying a cleaner route as compared to an initial route. In some examples, the clean route model 600 may be trained to identify a cleaner route that fulfills other travel metrics or criteria defined by the navigational search query.

[0091] The clean route model 600 may receive and / or process the training data 662 in a method similar to those described above with respect to the air quality prediction model 300. The clean route model 600 may provide, as output, one or more results related to the prediction. The results of the clean route model 600 may be generated as output data 664.

[0092] The training data 662 can correspond to an AI task for predicting a cleaner navigational route from a starting location to a destination location that satisfies the navigational search criteria.

[0093] From the training data 662, the clean route model 600 can be modified, or trained, to identify a cleaner route, e.g., a route with a lower accumulated exposure, as compared to an initial route. According to some examples, the clean route model 600 may provide, as output, one or more results related to the identification. The results of the clean route model 600 may be generated as output data 664. In some examples, the output data 664 may be a navigational route between the starting location and the destination location. The navigational route, e.g., cleaner route 444, may be provided as an overlay on a map 400. For example, the clean route model 600 may be configured to send the output data 664 to a navigational application 440 such that the navigational application 440 may provide the output data 664 for display on a display of a client or user device 402.

[0094] According to some examples, inference data 660 may be provided as input into the clean route model 600. The clean route model 600 may receive and / or process the inference data 660 in methods similar to those described above with respect to the air quality prediction model 300.

[0095] The inference data 660 for the clean route model 600 can include data associated with identifying one or more navigational routes between the starting location and the destination location that satisfies the navigational search and / or identifies the cleaner route as compared to an initial route. The inference data 660GOOGLE-4087 may include, for example, current air quality metrics, predicted air quality metrics, available roads for traveling between the starting location and destination location, available public transportation, weather data, or the like.

[0096] When executed, the clean route model 600 may receive, as input, the starting location 404 and destination location 442. The clean route model 600 may provide, as output, a cleaner route 444. The cleaner route 444 may be cleaner as compared to the initial, or alternative, route 446. A cleaner route may be a route having a lower accumulated exposure metric as compared to the initial route 446.

[0097] By using air quality metrics to identify navigational routes, the system beneficially determines navigational routes with objectively better air quality metrics that provides improved values of accumulated exposure metrics. Urban Planning

[0098] The determined air quality metric for geographic regions may be used when a user, neighborhood, city, state, geographic region, or the like is planning additions to a geographic region. For example, the air quality metrics may be used to identify a candidate location for said addition within the geographic region. The addition may be, for example, a building, such as a house, apartment, business, warehouse, factory, or the like. In some examples, the addition may be transportation routes, such as bus routes, train lines, carpool lanes, new highways, or the like.

[0099] Figure 7A illustrates an example map 700 identifying available locations within the geographic region. The air quality metrics for the regions on the map 700, in conjunction with the available locations, may be used to recommend a location for a building, business, factory, or the like. The locations may be, for example, locations, property, buildings, land, etc. The available locations may be, for example, locations that are available for rent “R” or for sale “FS”. The available locations may be identified based on buildings and / or land that are available for sale or rent within the geographic region. These locations may be identified based on publicly available listings, private ads listing the locations, or the like. According to some examples, the available locations may be further identified or filtered based on intended use. The intended use may be, for example, commercial, residential, or the like. In some examples, the available locations may be filtered based on approved zoning, such as factory, warehousing, offices, medical, single family homes, multi-family homes, apartments, or the like. In some examples, the available locations may be further filtered based on the intended use, needs, etc. of the location. For example, if a company is seeking to build a new office building having parking for more than 50 vehicles, locations within the geographic region having enough land space to accommodate the building and parking may be identified.

[0100] According to some examples, when identifying the candidate location for the addition within the geographic area, one or more weights may be applied to each factor. The factors may include, for example, the size of the land and / or building, the amount of parking, zoning requirements and / or restrictions, air quality metrics, population density, commutability, or the like. In some examples, the factors may correspond to and / or include the filters. The weights for each factor may be different such that more weight is given to the size of the land and / or building as compared to the commutability to the location. In some examples, theGOOGLE-4087 weights for one or more of the factors may be the same, such that the weight for the amount of parking and the population density of the geographic region are the same but the weight for air quality metrics is different.

[0101] Air quality metrics at the available locations, around the available locations, and / or throughout the geographic region may be determined. For example, the geographic region, shown on map 700, may be divided into smaller discrete regions 750, 752, 754, 756, 758. The air quality metrics for the discrete regions and / or the geographic region may be determined based on sensors “S” within the said regions. In some examples, the air quality metrics may be determined based on a virtual sensor generated based on the calibrated sensor data from the sensors within the region(s).

[0102] According to some examples, the air quality metrics from the geographic region and / or discrete regions may be determined based on current air quality metrics, predicted air quality metrics, and / or historic air quality metrics. For example, historical air quality metrics may be used to determine the air quality metric for a region for a given time of day, day of week, or the like. The historical air quality metrics may be combined with the current air quality metrics and / or predicted air quality metrics for the given time, day, day of week, etc. to determine an expected air quality metric for the region for that time, day, day of week, or the like.

[0103] As illustrated in Figure 7A, each region 750, 752, 754, 756, 758 may have a different expected air quality metric, as indicated based on the pattern overlay. While the expected different air quality metrics are shown as different patterns, the different expected air quality metrics may be represented by numerical values, colors, or the like. As one example, the diagonal pattern in regions 750, 756 may represent an expected air quality metric corresponding to “good”, the dot pattern in regions 752, 758 may represent an expected air quality metric corresponding to “moderate”, and the vertical line pattern of region 754 may represent an expected air quality metric corresponding to “poor”. However, the patterns may, in some examples, represent different metrics, ranges of metrics, or the like such that the examples provided herein are not intended to be limiting.

[0104] According to some examples, a proposed location for an addition to the geographic region 700 may be identified based on the available locations and the expected air quality metrics. For example, the expected air quality metrics of the regions 750, 752, 754, 756, 758 may be compared to identify areas within the geographic region having poorer air quality metrics as compared to other areas. In the example shown in Figure 7, region 754 has a poorer expected air quality metric as compared to regions 750, 752, 756, 758. If the system identifies available locations within a first area and a second area that fulfills the requirements of the user, business, town, etc., and the available locations within the first region has a poor expected air quality metric and another available locations within the second region has a better expected air quality metric than the first region, the system may suggest choosing the available location in the second area to distribute pollutants more equally.

[0105] As one example, if a user, business, etc. is seeking locations that are for rent, the system may identify regions 750, 754, and 758 as having available locations for rent. The system may compare the expected air quality metrics for regions 750, 754, and 758. In some examples, if the intended use of the locations is for aGOOGLE-4087 business that is open Monday-Friday, the system may compare the expected air quality metrics for typical business hours Monday-Friday. In other examples, if the intended use of the locations is for residential use, the system may compare the average daily expected air quality metrics. The system may determine that the expected air quality metrics of region 750 is better than the expected air quality metrics of regions 754, 758. For example, if the expected air quality metric of region 750 is “good” and the expected air quality metrics of regions 754, 758 is “poor” and “moderate”, respectively, the expected air quality metric of region 750 is better than the expected air quality metric of regions 754, 758. In such an example, the system may suggest choosing the locations for rent in region 750 rather than the locations for rent in regions 754, 758. The suggestion may, in some examples, cause the pollutants to be more equally distributed amongst the regions in map 700.

[0106] Figure 7B illustrates an example map 700 including a candidate transportation route identified based on the air quality metrics for the geographic region. For example, if a city, state, region, or the like is planning to add additional bus routes, train routes, carpool lanes, or the like to their infrastructure, the expected air quality metrics may be used to identify regions with lower, e.g., better, expected air quality metrics such that the pollutants within a geographic region illustrated on map 700 may be more equally distributed.

[0107] The candidate route 760 may be determined based on one or more factors, such as the current location of bus stops (“BS”), train stations, car pool lanes, etc., current bus routes, requested bus routes, air quality metrics for discrete regions 750, 752, 754, 756, 758 within the geographic region, population density within the region, or the like. Different weights may be applied to each factor. In some examples, an AI model can be used to identify the candidate routes. In such an example, the AI model may apply different weights to the inputs provided.

[0108] As shown on the map 700 in Figure 7B, the candidate route 760 may be identified to minimize the amount of time, distance, etc. within a discrete region having poor air quality metrics. The candidate route 760 may be the one in which the mode of transportation, e.g., a bus, spends the least amount of time and / or distance within region 754 as the actual and / or expected air quality metrics may be “poor” as compared to the actual and / or expected air quality metrics for regions 750, 752, 756, 758.

[0109] By determining candidate locations for businesses or candidate transportation routes within a geographic region, the system beneficially determines locations and / or routes that objectively distribute pollutants more equally throughout the region. Equal distribution of pollutants within a geographic region can objectively reduce accumulated exposure metrics for people within the geographic region.

[0110] Figure 7C illustrates an example map including recommendations for improving air quality metrics within the geographic region. The recommendations may include, for example, planting trees “T” within the regions, designating areas of the region as green space “GS”, or the like. The recommendations may be determined based on current and / or expected air quality metrics for the geographic region.

[0111] In the examples shown in Figure 7C, each region 750, 752, 754, 756, 758 may have a different actual and / or expected air quality metric, as indicated based on the pattern overlay. While the actual and / or expected different air quality metrics are shown as different patterns, the different actual and / or expected air qualityGOOGLE-4087 metrics may be represented by numerical values, colors, or the like. As one example, the diagonal pattern in regions 750, 756 may represent an actual and / or expected air quality metric corresponding to “good”, the dot pattern in regions 752, 758 may represent an actual and / or expected air quality metric corresponding to “moderate”, and the vertical line pattern of region 754 may represent an actual and / or expected air quality metric corresponding to “poor”. However, the patterns may, in some examples, represent different metrics, ranges of metrics, or the like such that the examples provided herein are not intended to be limiting.

[0112] In areas in which the actual and / or expected air quality metrics are poor, e.g., region 754, the recommendation may be to designate a portion of the region as green space “GS” such that more plants, trees, and wildlife can grow to help clean the air. In some examples, the green space “GS” designation may prevent additional buildings, factories, or the like from being built in the area. This may prevent additional pollutants from being added to the region. In some examples, such as regions 752, 756 in which the actual and / or expected air quality metrics are moderate, the recommendation may be to plant trees in the region. The trees may help clean the air thereby improving the air quality metrics of the region. Example Systems

[0113] Figure 8A illustrates an example system 800 in which the features described above may be implemented. It should not be considered limiting the scope of the disclosure or usefulness of the features described herein. In this example, system 800 may include device(s) 808, vehicle(s) 818, server computing device 830, storage system 840, and network 880.

[0114] Each of devices 808 may include one or more processors 838, memory 848, data 868 and instructions 858. Each of devices 808 may also include an output 878, user input 888, and location sensor 898898. The devices 808 may be any device that includes one or more sensors 898, such as a smartphone, tablet, laptop, smart watch, AR / VR headset, smart helmet, etc., as shown in Figure 8B, or a home device, e.g., a smart doorbell, smart lights, home assistant, smart thermostat, etc.

[0115] Memory 848 of devices 808 may store information that is accessible by processor 838. Memory 848 may also include data that can be retrieved, manipulated or stored by the processor 838. The memory 848 may be of any non-transitory type capable of storing information accessible by the processor 838, including a non- transitory computer-readable medium, or other medium that stores data that may be read with the aid of an electronic device, such as a hard-drive, memory card, read-only memory ("ROM"), random access memory ("RAM"), optical disks, as well as other write-capable and read-only memories. Memory 848 may store information that is accessible by the processors 838, including instructions 858 that may be executed by processors 838, and data 868.

[0116] Data 868 may be retrieved, stored or modified by processors 838 in accordance with instructions 858. For instance, although the present disclosure is not limited by a particular data structure, the data 868 may be stored in computer registers, in a relational database as a table having a plurality of different fields and records, XML documents, or flat files. The data 868 may also be formatted in a computer-readable format such as, but not limited to, binary values, ASCII or Unicode. By further way of example only, the data 868 may compriseGOOGLE-4087 information sufficient to identify the relevant information, such as numbers, descriptive text, proprietary codes, pointers, references to data stored in other memories (including other network locations) or information that is used by a function to calculate the relevant data.

[0117] The instructions 858 can be any set of instructions to be executed directly, such as machine code, or indirectly, such as scripts, by the processor 838. In that regard, the terms “instructions,” “application,” “steps,” and “programs” can be used interchangeably herein. The instructions can be stored in object code format for direct processing by the processor, or in any other computing device language including scripts or collections of independent source code modules that are interpreted on demand or compiled in advance. Functions, methods and routines of the instructions are explained in more detail below.

[0118] The one or more processors 838 may include any conventional processors, such as a commercially available CPU or microprocessor. Alternatively, the processor can be a dedicated component such as an ASIC or other hardware-based processor. Although not necessary, computing devices 808 may include specialized hardware components to perform specific computing functions faster or more efficiently.

[0119] Although Figure 8A functionally illustrates the processor, memory, and other elements of devices 808 as being within the same respective blocks, it will be understood by those of ordinary skill in the art that the processor or memory may actually include multiple processors or memories that may or may not be stored within the same physical housing. Similarly, the memory may be a hard drive or other storage media located in a housing different from that of the devices 808. Accordingly, references to a processor or device will be understood to include references to a collection of processors or devices or memories that may or may not operate in parallel.

[0120] Output 878 may be a display, such as a monitor having a screen, a touch-screen, a projector, or a television. The display 878 of the one or more computing devices 808 may electronically display information to a user via a graphical user interface ("GUI") or other types of user interfaces. For example, as will be discussed below, display 878 may electronically display air quality metrics, navigational routes including a cleaner route, maps for urban planning, or the like.

[0121] The user input 888 may be a mouse, keyboard, touch-screen, microphone, or any other type of input. The user input may receive the user’s authorization to use the sensors 898. In examples where the sensors 898 are air quality sensors, the user input may receive the user’s authorization to obtain air quality data for determining air quality metrics. In some examples, where the sensors 898 are location sensors, the user input may receive the user’s authorization to obtain location information for determining a geographic region. For example, the user can select particular applications for which to allow location services, particular times during which location services can be enabled, or other permissions or limitations for the location services.

[0122] The location sensor may be, for example, a global positioning system (“GPS”) sensor, wireless communications interface, etc. The location sensor, when enabled by the user, may provide a rough indication as to the location of the device. According to some examples, when authorized by the user, the location sensors may provide location information indicating a geographic region for determining air quality metrics.GOOGLE-4087

[0123] The air quality data may be stored locally on the device 808 or shared to a remote location, such as a remote server 830 or storage system 840. The air quality data may be used to determine air quality metrics, such as current air quality metrics, predicted air quality metrics, expected air quality metrics, or the like.

[0124] The location information may be stored locally on the device 808 or navigational system, such as part of an application or integrated into vehicle 818. In some examples, the location information may be shared to a remote location, such as a remote server 830 or storage system 840. According to some examples, the location information may be used to identify types of destinations visited and the frequency such that the system does not require or obtain the specific destination location.

[0125] The devices 808 can be at various nodes of a network 880 and capable of directly and indirectly communicating with other nodes of network 880. Although one device is depicted in Figure 8A, it should be appreciated that a typical system can include one or more devices, with each device being at a different node of network 880. The network 880 and intervening nodes described herein can be interconnected using various protocols and systems, such that the network can be part of the Internet, World Wide Web, specific intranets, wide area networks, or local networks. The network 880 can utilize standard communications protocols, such as WiFi, Bluetooth, 4G, 5G, etc., that are proprietary to one or more companies. Although certain advantages are obtained when information is transmitted or received as noted above, other aspects of the subject matter described herein are not limited to any particular manner of transmission.

[0126] In one example, system 800 may include one or more server computing devices 830 having a plurality of computing devices, e.g., a load balanced server farm, that exchange information with different nodes of a network for the purpose of receiving, processing and transmitting the data to and from other computing devices. For instance, one or more server computing devices 830 may be a web server that is capable of communicating with the one or more client computing devices 808 via the network 880. In addition, server computing device 830 may use network 880 to transmit and present information to a user of one of the other computing devices 808 or a passenger of a vehicle. In this regard, vehicle 818 may be considered a client computing device. Server computing device 830 may include one or more processors, memory, instructions, data, air quality sensors, location sensors, etc. These components operate in the same or similar fashion as those described above with respect to computing devices 808.

[0127] According to some examples, the server computing device 830 may be connected over the network to a data center 810 housing any number of hardware accelerators. The data center 810 can be one of multiple data centers or other facilities in which various types of computing devices, such as hardware accelerators, are located. Computing resources housed in the data center can be specified for deploying models related to determining air quality metrics, predicting air quality metrics, identifying cleaner routes, urban planning, or the like.

[0128] The server computing device 830 can be configured to receive requests to process data from the client computing device 808 on computing resources in the data center 810. For example, the environment can be part of a computing platform configured to provide a variety of services to users, through various userGOOGLE-4087 interfaces and / or application programming interfaces (APIs) exposing the platform services. The variety of services can include determining air quality metrics, predicting air quality metrics, identifying cleaner routes, urban planning, or the like. The client computing device 808 can transmit input data associated with sensor data, navigational queries, urban planning, or the like. The air quality prediction model 300, discussed above in conjunction with Figure 3, and the cleanest route model 600, discussed above in conjunction with Figure 6 can receive the input data, and in response, generate output data including air quality metrics, predicted air quality metrics, cleaner routes, suggested locations for additions to a geographic region, or the like.

[0129] As other examples of potential services provided by a platform implementing the environment, the server computing device can maintain a variety of models in accordance with different constraints available at the data center. For example, the server computing device can maintain different families for deploying models on various types of TPUs and / or GPUs housed in the data center or otherwise available for processing.

[0130] As shown in Figure 8B, device 808 may be a personal computing device intended for use by a respective user 888, and have all of the components normally used in connection with a personal computing device including one or more processors (e.g., a central processing unit (CPU)), memory (e.g., RAM and internal hard drives) storing data and instructions, an output, such as a display (e.g., a monitor having a screen, a touch-screen, a projector, a television, or other device such as a smart watch display that is operable to display information), and user input devices (e.g., a mouse, keyboard, touchscreen or microphone). The devices may also include a camera for recording video streams, speakers, a network interface device, and all of the components used for connecting these elements to one another. Devices 808 may be capable of wirelessly exchanging or obtaining data over the network 880.

[0131] Although the client computing devices may each comprise a full-sized personal computing device, they may alternatively comprise mobile computing devices capable of wirelessly exchanging data with a server over a network such as the Internet. By way of example only, devices 808 may be mobile phones or devices such as a wireless-enabled PDA, smartphones, a tablet PC, a wearable computing device (e.g., a smartwatch, AR / VR headset, smart helmet, etc.), a netbook that is capable of obtaining information via the Internet or other networks, or a smart home device, such as a home assistant, smart thermostat, smart doorbell, smart light, etc..

[0132] User 888 may operate a respective vehicle 818. The vehicle 818 may include a location sensor and / or air quality sensor. In some examples, vehicle 818 may include an integrated navigation system. According to some examples, the navigation system may be integrated into a user’s 888 respective device 808. In yet another example, the device 808 or vehicle 818 may execute a mapping application that provides maps or directions, identifies a user’s location, etc.

[0133] Any use of sensor data, e.g., air quality sensor data and / or location information of a user 888 is authorized by the respective user. For example, the user 888 may provide authorization to an application(s) for determining air quality metrics, cleaner routes, urban planning, etc. by setting certain permissions for the application(s). The authorization may be for the application(s) to access one or more databases or sub-databases in the memory of the device, vehicle, remote server, etc. According to one example, the user may select specificGOOGLE-4087 sub-databases to which the application is granted access. For instance, the user may grant access to the location history database but not the calendar archive database.

[0134] Vehicles 818 may include a computing device (not shown). The computing device may include one or more components similar to devices 808, such as one or more processors, memory, data, instructions, a display, a user input, etc. According to some examples, vehicles 818 may include a perception system for detecting and performing analysis on objects external to the vehicle such as other vehicles, obstacles in the roadway, traffic signals, signs, trees, etc. Additionally or alternatively, the perception system may determine whether the vehicle is in motion or parked. For example, the perception system may include lasers, sonar, radar, one or more cameras, or any other detection devices which record data which may be processed by a computing device (not shown) within vehicles 818. In the case where the vehicle is a small passenger vehicle such as a car, the car may include a laser mounted on the roof or other convenient locations as well as other sensors such as cameras, radars, sonars, air quality sensors, and additional lasers (not shown).

[0135] Storage system 840 may store various types of information. For instance, the storage system 840 may store data or information related to air quality data, historical air quality metrics, a user’s location information, geographic region delineations, available locations, navigational routes, etc.

[0136] According to some examples, storage system 840 may store data or information related to a user’s sensor data, e.g., air quality data and / or location information, after receiving authorization from the user 888. The authorization may be, for example, provided by setting permissions for the system to access sensor data, including air quality data and location information. For example, a user may be provided with controls allowing the user to make an election as to both if and when systems, programs, or features described herein may enable collection of location information, and if the user is sent content or communications from a server. In addition, certain data may be treated in one or more ways before it is stored or used, so that personally identifiable information is removed. The user may have control over what information is collected about the user, how that information is used, and what information is provided to the user.

[0137] While Figures 8A and 8B illustrate a single user 888 and their respective device(s) 808 and vehicle 818, it should be understood that there may be multiple users and their respective devices and vehicles. Each respective user provides authorization for an application to access their sensor data. Example Methods

[0138] Figure 9 is a flow diagram for an example method of determining an air quality metric for a geographic region. The following operations do not have to be performed in the precise order described below. Rather, various operations can be handled in a different order or simultaneously, and operations may be added or omitted.

[0139] In block 910, sensor data is received from a plurality of sensors. The plurality of sensors includes at least one air quality sensor. One or more of the plurality of sensors are within a geographic region. The geographic region may be determined based on a location of a user requesting the air quality metric. In some examples, the geographic region may be determined based on a location provided as input as part of the requestGOOGLE-4087 for determining the air quality metric. In some examples, the geographic region may be determined based on a threshold distance from the current location of the user and / or location provided as input. In another example, the geographic region may be determined based on a neighborhood, city, state, region, etc. encompassing the current location of the user and / or location provided as input. At least a portion of the sensor data is for one or more pollutants. The pollutant may be any solid or liquid that is suspended in the air. For example, the pollutant may be smoke, dust, dander, soot, pollen, sea salt, spores, water vapor (e.g., humidity), fumes, combustion byproducts (e.g., exhaust from a vehicle), or the like.

[0140] In block 920, at least a portion of the sensor data may be calibrated. Calibrating the sensor data may include translating the sensor data for the one or more pollutants from particle counts of the one or more pollutants to mass per volume of the one or more pollutants. For example, each type of pollutant has its own mass such that the count of that pollutant determined by the sensors can be converted to mass based on the specific, or typical, mass associated with that pollutant. The translated, or calibrated sensor data may be measured in micrograms per cubic meter (µg / m3).

[0141] According to some examples, one or more data outliers may be identified based on the calibrated sensor data. Identifying the one or more data outliers may include, for example, comparing the calibrated sensor data for a given air quality sensor in the geographic region to average calibrated sensor data for the geographic region. In examples where the calibrated sensor data for the given air quality sensor is more than a threshold amount above the average calibrated sensor data, the calibrated sensor data may be identified as an outlier. In examples where the calibrated sensor data for the given air quality sensor is more than a threshold amount below the average calibrated sensor data, the calibrated sensor data may be identified as an outlier.

[0142] In block 930, the sensor data including at least he portion of the calibrated sensor data is received as input into an artificial intelligence (AI) model trained to predict air quality metrics. The AI model may correspond to a virtual sensor for the geographic region. According to some examples, the virtual sensor may be generated based on an algorithm. The virtual sensor, e.g., AI model, may comprise one or more layers. Each layer may be executed simultaneously, semi-concurrently, or sequentially. Each layer may correspond to a type of sensor data. In some examples, each layer is a model trained to predict a preliminary air quality metric based on the type of sensor data.

[0143] In block 940, an air quality metric for the geographic region may be determined based on the virtual sensor. The air quality metric may be a current air quality metric. The current air quality metric may be based on the sensor data received at or near the time the air quality metric is determined such that the current air quality metric is determined in substantially real time.

[0144] According to some examples, a future air quality metric for the geographic region may be predicted. For example, an AI model, such as the air quality prediction model 300, may be executed and used to predict a future air quality metric for the geographic region. In some examples, the determined air quality metric for the geographic region, at least one of a time or a day in the future, and one or more of a future time of day, current weather for the geographic region, predicted weather forecast for the geographic region, environmentalGOOGLE-4087 factors may be received as input into the AI model. The environmental factors may include one or more of wind speed, wind direction, fires within a radius of the geographic region, earthquakes within the radius of the geographic region, and tornadoes within the radius of the geographic region. The future air quality metric may be provided as output.

[0145] According to some examples, a navigation request may be received. The navigation request may include a starting location and a destination location. A first navigational route from the starting location to the destination location may be identified. The first navigational route may have a first accumulated exposure metric. A second navigational route from the starting location to the destination location may be identified. The second route may have a second accumulated exposure metric. Determining the first accumulated exposure metric or the second accumulated exposure metric may include, for example, aggregating one or more air quality metrics for regions along the respective first or second navigational routes. The one or more air quality metrics for the regions along the respective first or second navigational routes may be current air quality metrics, predicted air quality metrics, and / or historical air quality metrics. For example, the predicted air quality metrics may correspond to the time at which the user would navigate through that region if the user took a given navigational route. Historical air quality metrics may be, for example, air quality metrics from the same day of the week and / or time of the day the user would navigate through that region if the user took a given navigational route.

[0146] The first accumulated exposure metric may be compared with the second accumulated exposure metric to determine which of the first route or the second route represents a cleaner route. In some examples, the cleaner route may be the route with the lower, better, and / or cleaner accumulated exposure metric. The navigation instructions relating to the cleaner route may be provided.

[0147] According to some examples, a geographic planning request for a geographic region may be received. The geographic planning request may be for an addition to the geographic region, such as a new building, a new business, a new transportation route, or the like. The geographic planning request may include one or more planning criteria. The planning criteria may include, for example, the intended use, whether the land is for sale or rent, the price of the property, the property size, zoning requirements and / or restrictions, parking requirements and / or restrictions, or the like.

[0148] Available locations within the geographic region may be identified. The available locations that are identified may be locations that fulfill the planning criteria. The geographic region may be further divided into discrete regions. The discrete regions may be, for example, smaller regions within the larger geographic region. The air quality metrics for the discrete regions may be determined.

[0149] The air quality metrics of the discrete regions may be compared. A suggested available location may be identified based on the comparison. For example, each region may have an associated air quality metric. The air quality metric may be a numerical value, such as an integer, where the lower the numerical value the better the air quality. In some examples, the air quality metric may be a level of concern, e.g., good, moderate, poor, unhealthy, hazardous, or the like. A lower numerical value of the air quality may, in some examples,GOOGLE-4087 correspond to a better level of concern. The discrete region associated with the suggested available location may have a lower numerical value for the air quality metric as compared to the numerical value of the air quality metric for the other discrete regions within the geographic region.

[0150] Aspects of this disclosure can be implemented in digital electronic circuitry, in tangibly-embodied computer software or firmware, and / or in computer hardware, such as the structure disclosed herein, their structural equivalents, or combinations thereof. Aspects of this disclosure can further be implemented as one or more computer programs, such as one or more modules of computer program instructions encoded on a tangible non-transitory computer storage medium for execution by, or to control the operation of, one or more data processing apparatus. The computer storage medium can be a machine-readable storage device, a machine-readable storage substrate, a random or serial access memory device, or combinations thereof. The computer program instructions can be encoded on an artificially generated propagated signal, such as a machine-generated electrical, optical, or electromagnetic signal, that is generated to encode information for transmission to suitable receiver apparatus for execution by a data processing apparatus.

[0151] The term “configured” is used herein in connection with systems and computer program components. For a system of one or more computers to be configured to perform particular operations or actions means that the system has installed on it software, firmware, hardware, or a combination thereof that cause the system to perform the operations or actions. For one or more computer programs to be configured to perform particular operations or actions means that the one or more programs include instructions that, when executed by one or more data processing apparatus, cause the apparatus to perform the operations or actions.

[0152] The term “data processing apparatus” refers to data processing hardware and encompasses various apparatus, devices, and machines for processing data, including programmable processors, a computer, or combinations thereof. The data processing apparatus can include special purpose logic circuitry, such as a field programmable gate array (FPGA) or an application specific integrated circuit (ASIC). The data processing apparatus can include code that creates an execution environment for computer programs, such as code that constitutes processor firmware, a protocol stack, a database management system, an operating system, or combinations thereof.

[0153] The data processing apparatus can include special-purpose hardware accelerator units for implementing machine learning models to process common and compute-intensive parts of machine learning training or production, such as inference or workloads. Machine learning models can be implemented and deployed using one or more machine learning frameworks.

[0154] The term “computer program” refers to a program, software, a software application, an app, a module, a software module, a script, or code. The computer program can be written in any form of programming language, including compiled, interpreted, declarative, or procedural languages, or combinations thereof. The computer program can be deployed in any form, including as a stand alone program or as a module, component, subroutine, or other unit suitable for use in a computing environment. The computer program can correspond to a file in a file system and can be stored in a portion of a file that holds other programs or data, such as oneGOOGLE-4087 or more scripts stored in a markup language document, in a single file dedicated to the program in question, or in multiple coordinated files, such as files that store one or more modules, sub programs, or portions of code. The computer program can be executed on one computer or on multiple computers that are located at one site or distributed across multiple sites and interconnected by a data communication network.

[0155] The term “database” refers to any collection of data. The data can be unstructured or structured in any manner. The data can be stored on one or more storage devices in one or more locations. For example, an index database can include multiple collections of data, each of which may be organized and accessed differently.

[0156] The term “engine” refers to a software-based system, subsystem, or process that is programmed to perform one or more specific functions. The engine can be implemented as one or more software modules or components, or can be installed on one or more computers in one or more locations. A particular engine can have one or more computers dedicated thereto, or multiple engines can be installed and running on the same computer or computers.

[0157] The processes and logic flows described herein can be performed by one or more computers executing one or more computer programs to perform functions by operating on input data and generating output data. The processes and logic flows can also be performed by special purpose logic circuitry, or by a combination of special purpose logic circuitry and one or more computers.

[0158] A computer or special purposes logic circuitry executing the one or more computer programs can include a central processing unit, including general or special purpose microprocessors, for performing or executing instructions and one or more memory devices for storing the instructions and data. The central processing unit can receive instructions and data from the one or more memory devices, such as read only memory, random access memory, or combinations thereof, and can perform or execute the instructions. The computer or special purpose logic circuitry can also include, or be operatively coupled to, one or more storage devices for storing data, such as magnetic, magneto optical disks, or optical disks, for receiving data from or transferring data to. The computer or special purpose logic circuitry can be embedded in another device, such as a mobile phone, a personal digital assistant (PDA), a mobile audio or video player, a game console, a Global Positioning System (GPS), or a portable storage device, e.g., a universal serial bus (USB) flash drive, as examples.

[0159] Computer readable media suitable for storing the one or more computer programs can include any form of volatile or non-volatile memory, media, or memory devices. Examples include semiconductor memory devices, e.g., EPROM, EEPROM, or flash memory devices, magnetic disks, e.g., internal hard disks or removable disks, magneto optical disks, CD-ROM disks, DVD-ROM disks, or combinations thereof.

[0160] Aspects of the disclosure can be implemented in a computing system that includes a back end component, e.g., as a data server, a middleware component, e.g., an application server, or a front end component, e.g., a client computer having a graphical user interface, a web browser, or an app, or any combination thereof. The components of the system can be interconnected by any form or medium of digitalGOOGLE-4087 data communication, such as a communication network. Examples of communication networks include a local area network (LAN) and a wide area network (WAN), e.g., the Internet.

[0161] The computing system can include clients and servers. A client and server can be remote from each other and interact through a communication network. The relationship of client and server arises by virtue of the computer programs running on the respective computers and having a client-server relationship to each other. For example, a server can transmit data, e.g., an HTML page, to a client device, e.g., for purposes of displaying data to and receiving user input from a user interacting with the client device. Data generated at the client device, e.g., a result of the user interaction, can be received at the server from the client device.

[0162] Unless otherwise stated, the foregoing alternative examples are not mutually exclusive, but may be implemented in various combinations to achieve unique advantages. As these and other variations and combinations of the features discussed above can be utilized without departing from the subject matter defined by the claims, the foregoing description of the examples should be taken by way of illustration rather than by way of limitation of the subject matter defined by the claims. In addition, the provision of the examples described herein, as well as clauses phrased as “such as,” “including” and the like, should not be interpreted as limiting the subject matter of the claims to the specific examples; rather, the examples are intended to illustrate only one of many possible implementations. Further, the same reference numbers in different drawings can identify the same or similar elements.

Claims

GOOGLE-4087 CLAIMS 1. A method, comprising: receiving, by one or more processors from a plurality of sensors, sensor data, wherein: the plurality of sensors includes at least one air quality sensor, one or more of the plurality of sensors are within a geographic region, and at least a portion of the sensor data is for one or more pollutants; calibrating, by the one or more processors, at least a portion of the sensor data; receiving, by the one or more processors, the sensor data including at least the portion of the calibrated sensor data as input into an artificial intelligence (AI) model trained to predict air quality metrics, wherein the AI model corresponds to a virtual sensor for the geographic region; and determining, by the one or more processors based on the virtual sensor, an air quality metric for the geographic region.

2. The method of claim 1, wherein calibrating the sensor data includes translating the sensor data for the one or more pollutants from particle counts of the one or more pollutants to mass per volume of the one or more pollutants.

3. The method of claim 1 or 2, wherein: the virtual sensor comprises one or more layers, each layer of the one or more layers corresponds to a type of sensor data, and each layer of the one or more layers is a model trained to predict a preliminary air quality metric based on the type of sensor data.

4. The method of any preceding claim, further comprising identifying, by the one or more processors based on the calibrated sensor data, one or more data outliers.

5. The method of claim 4, wherein when identifying the one or more data outliers, the method further comprises: comparing, by the one or more processors, the calibrated sensor data for a given air quality sensor in the geographic region to an average calibrated sensor data for the geographic region, when the calibrated sensor data for the given air quality sensor is more than a threshold amount above the average calibrated sensor data, the calibrated sensor data is identified as an outlier, and when the calibrated sensor data for the given air quality sensor is more than a threshold amount below the average calibrated sensor data, the calibrated sensor data is identified as an outlier.GOOGLE-4087 6. The method of any preceding claim, further comprising predicting, by the one or more processors executing a second artificial intelligence (AI) model, a future air quality metric for the geographic region.

7. The method of claim 6, further comprising: receiving, as input into the second AI model, the determined air quality metric for the geographic region, at least one of a time or a day in the future, and one or more of current weather for the geographic region, predicted weather forecast for the geographic region, and environmental factors; and providing as output, by the one or more processors executing the second AI model, the future air quality metric.

8. The method of claim 7, wherein the environmental factors include one or more of wind speed, wind direction, fires within a radius of the geographic region, earthquakes within the radius of the geographic region, and tornadoes within the radius of the geographic region.

9. The method of any preceding claim, further comprising: receiving, by the one or more processors, a navigation request including a starting location and a destination location; identifying, by the one or more processors, a first navigational route from the starting location to the destination location, the first navigational route having a first accumulated exposure metric; identifying, by the one or more processors, a second navigational routes from the starting location to the destination location, the second navigational route having a second accumulated exposure metric; comparing, by the one or more processors, the first accumulated exposure metric with the second accumulated exposure metric to determine which of the first navigational route or the second navigational route represents a cleaner route; and providing, by the one or more processors, navigation instructions relating to the cleaner route.

10. The method of claim 9, wherein determining the first accumulated exposure metric or the second accumulated exposure metric comprises aggregating, by the one or more processors, one or more air quality metrics for regions along the respective first or second navigational routes.

11. The method of claim 10, wherein aggregating the one or more air quality metrics comprises: determining, by the one or more processors, a respective air quality index for each type of pollutant along the respective first or second navigational routes; determining, by the one or more processors, a worse one of the respective air quality index among the types of pollutants along the respective first or second navigational routes; andGOOGLE-4087 identifying, by the one or more processors, the worse one of the respective air quality indexes for the respective first or second navigational routes as the respective first or second accumulated exposure metric.

12. The method of any preceding claim, further comprising: receiving, by the one or more processors, a geographic planning request for a geographic region; identifying, by the one or more processors based on the geographic planning request, available locations within the geographic region; determining, by the one or more processors, air quality metrics of one or more discrete regions within the geographic region, wherein the available locations are within at least one of the discrete regions; comparing, by the one or more processors, the air quality metrics of the one or more discrete region; and identifying, by the one or more processors based on the comparison, a suggested available location, wherein the discrete region associated with the suggested available location has a lower numerical value for the air quality metric as compared to other discrete regions.

13. The method of claim 12, wherein the geographic planning request includes one more or more planning criteria.

14. The method of claim 13, wherein the one or more planning criteria include at least one of intended use, rent, sale, price, property size, zoning requirements, and parking requirements.

15. A system, comprising: one or more processors, the one or more processors configured to: receive, from a plurality of sensors within a geographic region, sensor data , wherein: the plurality of sensors includes at least one air quality sensor, one or more of the plurality of sensors are within a geographic region, and at least a portion of the sensor data is for one or more pollutants; calibrate at least a portion of the sensor data; receive the sensor data including at least the portion of the calibrated sensor data as input into an artificial intelligence (AI) model trained to predict air quality metrics, wherein the AI model corresponds to ; and determine, based on the virtual sensor, an air quality metric for the geographic region.

16. The system of claim 15, wherein when calibrating the sensor data, the one or more processors are further configured to translate the sensor data for the one or more pollutants from particle counts of the one or more pollutants to mass per volume of the one or more pollutants.GOOGLE-4087 17. The system of claim 15 or 16, wherein: the virtual sensor comprises one or more layers, each layer of the one or more layers corresponds to a type of sensor data, and each layer of the one or more layers is a model trained to predict a preliminary air quality metric based on the type of sensor data.

18. The system of any of claims 15 to 17, wherein the one or more processors are further configured to identify, based on the calibrated sensor data, one or more data outliers.

19. The system of claim 18, wherein when identifying the one or more data outliers, the one or more processors are further configured to: compare the calibrated sensor data for a given air quality sensor in the geographic region to an average calibrated sensor data for the geographic region, when the calibrated sensor data for the given air quality sensor is more than a threshold amount above the average calibrated sensor data, the calibrated sensor data is identified as an outlier, and when the calibrated sensor data for the given air quality sensor is more than a threshold amount below the average calibrated sensor data, the calibrated sensor data is identified as an outlier.

20. The system of any of claims 15 to 19, wherein the one or more processors are further configured to predict, by executing a second artificial intelligence (AI) model, a future air quality metric for the geographic region.

21. The system of claim 20, wherein the one or more processors are further configured to: receive, as input into the second AI model, the determined air quality metric for the geographic region, at least one of a time or a day in the future, and one or more of current weather for the geographic region, predicted weather forecast for the geographic region, and environmental factors; and provide as output, by executing the second AI model, the future air quality metric.

22. The system of claim 21, wherein the environmental factors include one or more of wind speed, wind direction, fires within a radius of the geographic region, earthquakes within the radius of the geographic region, and tornadoes within the radius of the geographic region.

23. The system of any of claims 15 to 22, wherein the one or more processors are further configured to: receive a navigation request including a starting location and a destination location;GOOGLE-4087 identify a first navigational route from the starting location to the destination location, the first navigational route having a first accumulated exposure metric; identify a second navigational routes from the starting location to the destination location, the second navigational route having a second accumulated exposure metric; compare the first accumulated exposure metric with the second accumulated exposure metric to determine which of the first navigational route or the second navigational route represents a cleaner route; and provide navigation instructions relating to the cleaner route.

24. The system of claim 23, wherein when determining the first accumulated exposure metric or the second accumulated exposure metric the one or more processors are further configured to aggregate one or more air quality metrics for regions along the respective first or second navigational routes.

25. The system of claim 24, wherein aggregating the one or more air quality metrics comprises: determining, by the one or more processors, a respective air quality index for each type of pollutant along the respective first or second navigational routes; determining, by the one or more processors, a worse one of the respective air quality index among the types of pollutants along the respective first or second navigational routes; and identifying, by the one or more processors, the worse one of the respective air quality indexes for the respective first or second navigational routes as the respective first or second accumulated exposure metric.

26. The system of any of claims 15 to 25, wherein the one or more processors are further configured to: receive a geographic planning request for a geographic region; identify, based on the geographic planning request, available locations within the geographic region; determine air quality metrics of one or more discrete regions within the geographic region, wherein the available locations are within at least one of the discrete regions; compare the air quality metrics of the one or more discrete region; and identify, based on the comparison, a suggested available location, wherein the discrete region associated with the suggested available location has a lower numerical value for the air quality metric as compared to other discrete regions.

27. The system of claim 26, wherein the geographic planning request includes one more or more planning criteria.

28. The system of claim 27, wherein the one or more planning criteria include at least one of intended use, rent, sale, price, property size, zoning requirements, and parking requirements.GOOGLE-4087 29. One or more non-transitory computer-readable storage media encoding instructions that, when executed by one or more processors, cause the one or more processors to perform operations comprising: receiving, from a plurality of sensors within a geographic region, sensor data , wherein: the plurality of sensors includes at least one air quality sensor, one or more of the plurality of sensors are within a geographic region, and at least a portion of the sensor data is for one or more pollutants; calibrating, by the one or more processors, at least a portion of the sensor data; receiving the sensor data including at least the portion of the calibrated sensor data as input into an artificial intelligence (AI) model trained to predict air quality metrics, wherein the AI model corresponds to a virtual sensor for the geographic region; and determining, based on the virtual sensor, an air quality metric for the geographic region.

30. The one or more non-transitory computer-readable storage media of claim 29, wherein calibrating the sensor data includes translating the sensor data for the one or more pollutants from particle counts of the one or more pollutants to mass per volume of the one or more pollutants.

31. The one or more non-transitory computer-readable storage media of claim 29 or 30, wherein: the virtual sensor comprises one or more layers, each layer of the one or more layers corresponds to a type of sensor data, and each layer of the one or more layers is a model trained to predict a preliminary air quality metric based on the type of sensor data.

32. The one or more non-transitory computer-readable storage media of any of claims 29 to 31, wherein the operations further comprise identifying, based on the calibrated sensor data, one or more data outliers.

33. The one or more non-transitory computer-readable storage media of claim 32, wherein when identifying the one or more data outliers, the operations further comprise: comparing the calibrated sensor data for a given air quality sensor in the geographic region to an average calibrated sensor data for the geographic region, when the calibrated sensor data for the given air quality sensor is more than a threshold amount above the average calibrated sensor data, the calibrated sensor data is identified as an outlier, and when the calibrated sensor data for the given air quality sensor is more than a threshold amount below the average calibrated sensor data, the calibrated sensor data is identified as an outlier.GOOGLE-4087 34. The one or more non-transitory computer-readable storage media of any of claims 29 to 33, wherein the operations further comprise predicting, by the one or more processors executing a second artificial intelligence (AI) model, a future air quality metric for the geographic region.

35. The one or more non-transitory computer-readable storage media of claim 34, wherein the operations further comprise: receiving, as input into the second AI model, the determined air quality metric for the geographic region, at least one of a time or a day in the future, and one or more of current weather for the geographic region, predicted weather forecast for the geographic region, and environmental factors; and providing as output, by executing the second AI model, the future air quality metric.

36. The one or more non-transitory computer-readable storage media of claim 35, wherein the environmental factors include one or more of wind speed, wind direction, fires within a radius of the geographic region, earthquakes within the radius of the geographic region, and tornadoes within the radius of the geographic region.

37. The one or more non-transitory computer-readable storage media of any of claims 29 to 36, wherein the operations further comprise: receiving a navigation request including a starting location and a destination location; identifying a first navigational route from the starting location to the destination location, the first navigational route having a first accumulated exposure metric; identifying a second navigational routes from the starting location to the destination location, the second navigational route having a second accumulated exposure metric; comparing the first accumulated exposure metric with the second accumulated exposure metric to determine which of the first navigational route or the second navigational route represents a cleaner route; and providing, by the one or more processors, navigation instructions relating to the cleaner route.

38. The one or more non-transitory computer-readable storage media of claim 37, wherein when determining the first accumulated exposure metric or the second accumulated exposure metric the operations further comprise aggregating one or more air quality metrics for regions along the respective first or second navigational routes.

39. The one or more non-transitory computer-readable storage media of claim 38, wherein when aggregating the one or more air quality metrics the operations further comprise: determining a respective air quality index for each type of pollutant along the respective first or second navigational routes;GOOGLE-4087 determining a worse one of the respective air quality index among the types of pollutants along the respective first or second navigational routes; and identifying the worse one of the respective air quality indexes for the respective first or second navigational routes as the respective first or second accumulated exposure metric.

40. The one or more non-transitory computer-readable storage media of any of claims 29 to 39, wherein the operations further comprise: receiving a geographic planning request for a geographic region; identifying, based on the geographic planning request, available locations within the geographic region; determining air quality metrics of one or more discrete regions within the geographic region, wherein the available locations are within at least one of the discrete regions; comparing the air quality metrics of the one or more discrete region; and identifying, based on the comparison, a suggested available location, wherein the discrete region associated with the suggested available location has a lower numerical value for the air quality metric as compared to other discrete regions.

41. The one or more non-transitory computer-readable storage media of claim 40, wherein the geographic planning request includes one more or more planning criteria.

42. The one or more non-transitory computer-readable storage media of claim 41, wherein the one or more planning criteria include at least one of intended use, rent, sale, price, property size, zoning requirements, and parking requirements.

Images