System and method for proxy model of network quality

Abstract

A machine learning-based proxy model may be generated to determine a quality of user experience in a vehicle-based communication network in the absence of direct feedback from a user regarding the user experience. The machine learning-based proxy model may include supervised and / or unsupervised machine learning models, including, for example, k-means clustering algorithms and / or artificial neural networks. Once generated, the proxy model may be applied to observed operating parameters of the vehicle-based communication network to quantify a user experience of any user for any duration of the vehicle-based communication network.

Description

[0001] Cross-reference to related applications This application claims priority and benefit of U.S. Provisional Patent Application No. 63 / 493982, filed April 3, 2023, entitled “System and Method for Agent Models for Network Quality,” the entire disclosure of which is incorporated herein by reference. Technical Field

[0002] This disclosure relates to communication systems, and more specifically, to techniques for measuring the quality of experience at a user device during use of a communication network. Background Technology

[0003] Vehicles, including but not limited to aircraft, can establish one or more satellite- and / or terrestrial communication links to receive and / or transmit information to and from the vehicle. Vehicle-based communication systems are typically implemented via various onboard communication components, including, for example, one or more aircraft-mounted antennas and other components that can be implemented as, for example, onboard Line Replaceable Units (LRUs). The operation of a vehicle-based communication system enables an onboard communication network that, for example, allows passenger devices to receive live media content (e.g., web browsing, sports events, live news) on passenger electronic devices, or enables two-way live communication to and from passenger devices (e.g., internet browsing, cellular calls, etc.). Additionally or alternatively, such communication links can enable the vehicle to communicate with the ground to support necessary operations of vehicle instruments and / or crew (e.g., aircraft navigation systems or crew communication).

[0004] A fundamental goal of any vehicle-based communication system is to operate so that users of the system are satisfied, i.e., to provide a satisfactory onboard network experience from the user’s perspective for the duration of network usage (e.g., for the duration of a flight). Various technical factors can impact the ability of a communication system provider to provide a satisfactory network experience. These factors can include, for example, the availability and / or performance of components of the communication system through which the network is provided, and / or the resilience of software / hardware / firmware elements to mitigate or resolve problems encountered in real time. Other technical limitations of the communication system can also play a role in the user experience, such limitations not always being under the control of the provider of the onboard system. Such limitations can include, for example, limited bandwidth of an air-to-ground (ATG) or satellite link (and / or of the larger communication network on which the onboard network is dependent (e.g., of a larger communication network of cell towers in a large ATG network)), or latency associated with components in the larger communication network. Moreover, regardless of the level of technical resources available, and even when the system provider operates the onboard network in a manner that the system provider believes provides the best possible experience to users (e.g., with the network performing “well” from the system provider’s perspective by considering tradeoffs of application / feature availability, bandwidth, latency, etc.), the service parameters considered by the system provider can not match the service parameters that are most strongly correlated to a given user’s own evaluation of the user experience.

[0005] In view of these factors and limitations, vehicle communication system providers have sought models for assessing user satisfaction with the network experience (or “user experience”) provided to users during any given transit or across multiple transits. Models established by vehicle communication system providers to assess user experience have typically relied on direct, explicit feedback from users, e.g., rating their own experience and / or identifying specific factors observed that impacted their own experience. To this end, the system provider can provide a survey to users, e.g., delivered to the users’ respective devices during or immediately upon conclusion of the onboard network’s usage. However, the present disclosure identifies that for certain types of transit routes and services, the rate of feedback to user surveys is typically low, or non-existent. Moreover, survey-based feedback, even when received, can not specifically identify factors that positively and / or negatively impacted the user experience. SUMMARY

[0006] The present disclosure describes systems and methods for generating and applying a proxy model to assess user experience in a vehicle-based communication network, which will be used without direct feedback from network users. The proxy model, operating as a proxy for direct user feedback, can identify and model a relationship(s) between user experience and each of a plurality of parameters associated with operation of the onboard communication network.

[0007] These operational parameters can include, for example, identifications and / or characteristics of hardware components included in the vehicle-based communication system (e.g., identifications of onboard LRUs, or ages of LRUs). Additionally or alternatively, the operational parameters can include parameters associated with a state of the onboard network and / or a larger communication network in which the onboard network operates (e.g., an amount of time that the vehicle is within or outside of coverage of the larger communication network, or a total transmission power available to one or more cell towers of an ATG network via which the vehicle can communicate). Still additionally or alternatively, the operational parameters of the proxy model can include behavioral parameters indicative of behavior of the user(s) using the onboard network (e.g., usage of particular websites, applications, data consumption rates, etc.). Yet additionally or alternatively, in some embodiments, the operational parameters of the proxy model can include service parameters associated with providing onboard network access to the user(s) (e.g., network usage with respect to limitations on access limitations between a service provider and the respective user(s)).

[0008] The present disclosure will describe various techniques for generating the proxy model. Generally, generating the proxy model can involve machine learning techniques, for example, supervised and / or unsupervised machine learning using at least the operational parameters described above as input. For example, in some embodiments, a dataset including operational parameters in observed situations is provided as input to an unsupervised clustering algorithm to generate a plurality of clusters corresponding to respective quantifications of user experience (e.g., a k-means clustering algorithm can create a plurality of clusters k, for example, corresponding to integers on a 0-10 scale rating user experience). As another example, in some embodiments, a training dataset including observed operational parameters and recorded user experience scores from the same observed situations is provided as input to an artificial neural network, which can be trained based on the training data to rate user experience (e.g., provide a score on a 0-10 integer scale as output). In various embodiments, other supervised and / or unsupervised machine learning techniques can also be employed to generate the proxy model, including combinations thereof.

[0009] Once the proxy model is generated (e.g., once the machine learning model(s) are trained), the proxy model can be applied to quantify a portion of any user’s user experience for any particular transportation or portion thereof (e.g., a portion or all of a user’s use of an in-flight communication network using one or more devices). In embodiments, the in-flight communication system provider (and / or providers of individual hardware, services, and / or applications therein) can leverage the user experience values and / or trends identified therein to identify and implement improvements to hardware, firmware, software, and / or processes for providing access to in-flight communication networks. Thus, even in situations where direct feedback from a portion of users is inaccessible, the generation and application of proxy models can enable in-flight communication system providers to quantify and leverage determinations of user experience for in-flight networks.

[0010] In embodiments, a computer-implemented method is provided that is performed via one or more processors. The computer-implemented method can include: (1) training a proxy model to analyze values of a plurality of operational parameters associated with a vehicle-based communication network that provides a vehicle, based on training data that includes a training set of operational parameter values associated with use of the vehicle-based communication network by users, to generate a measure of quality of user experience in the vehicle-based communication network; (2) obtaining a first observed set of operational parameter values corresponding to use of the vehicle-based communication network by a first user during a first instance of travel of the vehicle; and / or (3) analyzing the first observed set of operational parameter values via one or more trained proxy models to generate a first measure of quality of user experience in the vehicle-based communication network for the first user in the vehicle. The method can include additional, fewer, or alternative actions, including those described herein.

[0011] In another embodiment, one or more non-transitory computer-readable media are provided. The one or more non-transitory computer-readable media store non-transitory computer-executable instructions that, when executed by the one or more processors of one or more computing devices, cause the one or more computing devices to: (1) train, based on training data, a proxy model to analyze values of a plurality of operational parameters associated with a vehicle-based communication network that provides a vehicle to generate a measure of a quality of user experience in the vehicle-based communication network, the training data comprising a training set of operational parameter values associated with usage of the vehicle-based communication network by users; (2) obtain a first observed set of operational parameter values corresponding to usage of the vehicle-based communication network by a first user during a first trip of the vehicle; and / or (3) analyze, via one or more trained proxy models, the first observed set of operational parameter values to generate a first measure of a quality of user experience in the vehicle-based communication network for the first user in the vehicle. The one or more non-transitory computer-readable media can include additional, fewer, or alternative instructions, including those described herein.

[0012] In yet another embodiment, a computing system is provided. The computing system can include one or more processors and one or more memories storing instructions that, when executed via the one or more processors, cause the computing system to: (1) train, based on training data, a proxy model to analyze values of a plurality of operational parameters associated with a vehicle-based communication network that provides a vehicle to generate a measure of a quality of user experience in the vehicle-based communication network, the training data comprising a training set of operational parameter values associated with usage of the vehicle-based communication network by users; (2) obtain a first observed set of operational parameter values corresponding to usage of the vehicle-based communication network by a first user during a first trip of the vehicle; and / or (3) analyze, via one or more trained proxy models, the first observed set of operational parameter values to generate a first measure of a quality of user experience in the vehicle-based communication network for the first user in the vehicle. The computing system can include additional, fewer, or alternative computing elements and / or instructions, including those described herein. BRIEF DESCRIPTION OF DRAWINGS

[0013] The accompanying drawings, together with the detailed description below, are incorporated into and form part of the specification, and are included to further illustrate embodiments including the concepts of the claimed embodiments and to explain various principles and advantages of those embodiments. In the drawings, like reference numerals refer to like or functionally similar elements throughout the separate views.

[0014] Figure 1 An example computing environment in which the technology described herein can be implemented is shown in accordance with some embodiments; Figure 2A Elements of an example artificial neural network are shown in accordance with some embodiments; Figure 2B Elements of an example artificial neural network are shown in accordance with some embodiments; Figure 2A Figure 3 A block diagram of an example computing device that implements the technology of the present disclosure is shown in accordance with some embodiments; Figure 4A An example computer-implemented method associated with generating a proxy model of a user experience is shown in accordance with some embodiments; and Figure 4B An example computer-implemented method associated with a proxy model for a user experience is shown in accordance with some embodiments. DETAILED DESCRIPTION

[0015] While the examples provided herein will describe an exemplary aircraft-based network environment, it should be appreciated that the systems and methods herein can be applied to other types of vehicles in which an onboard communication network is provided (e.g., trains, ships, buses, etc.). Moreover, in still other embodiments, at least some of the technology herein can be applied to communication networks other than vehicle-based communication networks, including, for example, various other visitor-based communication networks (e.g., networks located in fixed environments such as businesses, homes, event spaces, transportation hubs, etc.).

[0016] While the following detailed description will disclose detailed descriptions of example methods, devices, and / or articles of manufacture, it should be understood that the legal scope of the property is defined by the literal words of the claims set forth at the end of this patent. Thus, the following detailed description will be interpreted only as examples and will not describe every possible example, as describing every possible example is impractical, if not impossible. Numerous alternatives will be apparent to those skilled in the art given the benefit of this disclosure. It is contemplated that such alternatives still fall within the scope of the claims.

[0017] Example communication system Figure 1 ​An example communication system 100 is shown, through which the techniques of this disclosure can be implemented. The communication system 100 includes an aircraft 102 having one or more antennas 104 mounted to it. Figure 1 In the illustrated embodiment, an antenna 104 is mounted to the belly of the aircraft 102 (e.g., to implement one or more air-to-ground (ATG) networks, such as via the S-band, cellular / LTE band, etc.). An antenna 104 is mounted to the top of the fuselage of the aircraft 102 (e.g., to implement one or more satellite-based networks, such as via K-band). u Bandwidth, K a (Band, L-band, S-band, etc.). In various embodiments, additional or fewer antennas 104 are possible. Although reference will be made to... Figure 1 Aircraft 102 is described, but other types of transportation (e.g., personal cars, buses, trains, ships, helicopters, emergency vehicles, etc.) are conceivable in various embodiments.

[0018] Communication system 100 uses one or more antennas 104 and other onboard networking equipment to provide an onboard communication network 106 (i.e., one or more communication networks) (e.g., Wi-Fi network, cellular network, Bluetooth®, one or more wired networks, etc.) within the cabin of aircraft 102. Communication network 106 enables personal electronic devices 110 of occupants on board aircraft 102 to send and receive data, such as browsing the internet or consuming media content (e.g., movies, television, news broadcasts, and / or other media that can be locally stored on aircraft 102 and / or transmitted from the ground). Although not shown in the illustration for clarity... Figure 1 As shown, however, aircraft 102 may include electronic systems, such as avionics systems (or equivalents of non-aircraft vehicles), such as communication systems, navigation systems, instruments, flight control systems, or collision avoidance systems, any one or more of which may additionally use communication network 106 to send communications to and / or receive communications from the ground.

[0019] Because many of the electronic systems of the aircraft 102 can require some degree of stability and / or secure attachment during transport, at least some of the other on-board network devices of the aircraft 102 can be included in line replaceable units (LRUs) 120 that are fixed or rigidly attached to the aircraft 102. The LRUs 120 can be composed of modular components that can be sealed units of the aircraft designed to be efficiently replaced without the use of specialized tools when the LRU 120 is being tested and repaired, thereby enabling the aircraft 102 to be quickly returned to service. Generally, the LRUs 120 are electronic assemblies that perform a particular function in the aircraft 102 and can be removed or replaced as a unit and repaired at a vehicle maintenance center. Some of the electronic systems of the aircraft 102 can not be included in the LRUs 120. For example, instead of being fixedly connected to the aircraft 102 via the LRUs 120, some of the electronic systems can be fixedly connected to the aircraft 102 using some other means, such as a bracket or other attachment device. In any case, the other on-board network devices of the aircraft 102 include one or more modems 122, a content delivery system 124, and a content library 126.

[0020] The one or more modems 122 can be configured to be compatible with a plurality of different communication standards utilized by the one or more ATG communication links 130 and / or the one or more satellite communication links. The one or more ATG communication links 130 can utilize a communication protocol associated with terrestrial communications (e.g., TDMA, GSM, CDMA, LTE, WiMAX, Wi-Fi, 4G, 5G, etc.) and / or frequencies in the K a band, the K u band, the L band, and / or any other suitable wireless communication band. Each of the one or more modems 122 can be connected (wired or wirelessly) to at least one respective antenna 104 (e.g., via a connection of a Tx port of the modem 122 to an Rx port of the antenna 104, and / or via a connection of an Rx port of the modem 122 to a Tx port of the antenna 104). The communication links 130 can utilize a communication protocol associated with terrestrial communications (e.g., TDMA, GSM, CDMA, LTE, WiMAX, Wi-Fi, 4G, 5G, etc.) and / or frequencies in the K a band, the K u band, the L band, and / or any other suitable wireless communication band.

[0021] Each modem 122 can operate on one or more frequency bands, and the aircraft 102 can receive data from or transmit data to the aircraft 102 with the modems 122. For example, the aircraft 102 can include one of the modems 122 thereon that is tuned to a frequency band allocated for direct communication between the aircraft 102 and a ground station, or that supports a direct air-to-ground (ATG) communication link (e.g., 849-851 MHz and 894-896 MHz). The aircraft 102 can additionally or alternatively include one of the modems 122 thereon that is tuned to a frequency band allocated for satellite communication, such as the L-band (40 to 60 GHz or 1 to 2 GHz), the K u band (12 to 18 GHz), the K a band (26.5 to 40 GHz), and / or other spectrum allocated for satellite communication. Moreover, each modem 122 can operate according to certain communication protocols. For example, at least one modem 122 can operate according to a 4G communication protocol, and at least one modem 122 can operate according to a 5G communication protocol. The ATG communication link(s) 130 (and / or one or more satellite-based communication links) can connect the aircraft 102 to one or more land-based stations 140, which in turn can be connected to a ground network 150. The ground network can include various elements, including, for example, one or more data centers that provide ground-based services to elements of the communication system 100, and / or various sources and / or destinations of content transmitted to or received from the aircraft 102 (e.g., the Internet). In particular, the ground network 150 can include one or more agent model modules 160 that can be configured to perform at least a portion of the techniques of the agent model of the present disclosure (e.g., generating an agent model based on flight data and other operational parameters obtained from one or more aircraft 102, and / or applying the agent model to subsequent operations of the network 106).

[0022] Returning to the LRU 120, the content delivery system 124 is configured to deliver content from the content library 126 to the electronic devices 110 via the communication network 106, among other things. The electronic devices 110 can include mobile computing devices of passengers and crew members, such as smartphones, tablets, laptops, personal digital assistants, e-readers, smart glasses, smart watches, and / or any other mobile computing devices capable of wireless communication. In embodiments, the electronic devices 110 can include seatback-mounted units configured to provide content to users at respective seats of the aircraft 102. The electronic devices 110 can initiate requests for content, for example, in response to user input. The content delivery system 124 can deliver content to the electronic devices 110 regardless of their connectivity to external networks, such as the Internet 160. The content delivery system 124 can be configured to provide an interface to the electronic devices 110 via the communication network 106 to view a list of content, select content, view content, download content, or purchase content or access content, for example, via the Internet 160 or the content library 126. The content delivery system 124 can communicate with the electronic devices 110, for example, via one or more wired access points and / or wireless access points (WAPs) placed around the cabin of the aircraft 102. The communication network 106 can include additional networking equipment, such as routers, hubs, switches, repeaters, bridges, and / or gateway devices. Some networking equipment can utilize a spread-spectrum paradigm and / or one or more RF bands (e.g., ISM bands, such as 900 MHz bands, 2.4 GHz bands, or 5 GHz bands) to facilitate communication.

[0023] In response to one or more requests for delivery of content to one or more electronic devices 110, for example, from the one or more electronic devices 110, content provided via the content delivery system 124 can be rendered at a user interface of the one or more electronic devices 110. Some examples of media content include movies, television programs, songs, video games, digital magazines, news feeds, web data, applications, messages, or any other content involving textual, audio, and / or visual presentation. Content can also include software, configuration data, files, etc., which can be installed at the aircraft 102 or systems of the aircraft 102 (e.g., the LRU 120), applied to the aircraft 102 or systems of the aircraft 102 (e.g., the LRU 120), made available via the aircraft 102 or systems of the aircraft 102 (e.g., the LRU 120) (e.g., as a file server), for example. In particular, software content provided by the content delivery system 124 to the one or more electronic devices 110 can include an antenna installation assistance application (and / or data or functionality associated therewith) provided to one or more electronic devices 110 of operators or maintenance personnel associated with the aircraft 102.

[0024] Generally, the agent model module(s) 160 can include hardware and / or software elements configured to generate and / or apply agent models for evaluating the user experience of the network 106 of the aircraft 102, or more generally, the user experience of the respective onboard communication network in each of a plurality of vehicles (e.g., a fleet of vehicles operated by a communication system provider). Generating and applying agent models at a high level can involve the agent model module(s) 160 accessing and analyzing various operational parameters associated with the communication system 100 and its operation, such as the identification of components included in the system 100, operational data of the LRUs 120, and / or network traffic information associated with the use of the electronic devices 110 in the onboard network 106. The agent model module(s) 160 on the ground can obtain such operational parameters, for example, when the aircraft 102 is in flight or after the aircraft 102 has completed a given flight, via the ATG link 130 and / or satellite link(s) implemented through the aircraft 102. Additionally or alternatively, in some embodiments, the agent model module(s) 160 can receive physical storage media indicative of operational parameters associated with the system 100, such as by transferring or copying one or more physical storage media from the LRUs 120 to the agent model module(s) 160.

[0025] In some embodiments, in addition to or as an alternative to being implemented within the ground network 150, portions of the agent model module(s) 160 can be implemented within the aircraft 102 itself (e.g., within the LRUs). For example, in some embodiments, the agent model module(s) 160 within the LRUs 120 can be configured to generate or re-generate (e.g., re-train) onboard agent models based on real-time operational parameters obtained during operation of the onboard network 106 (e.g., during or immediately before / after a trip of the aircraft 102). As another example, in some embodiments, the agent model module(s) 160 within the LRUs 120 are configured to receive instructions of agent models generated within the ground network 150 and subsequently apply the agent models to operational parameters of any particular trip of the aircraft 102 in real-time. Generally, however, generating agent models on the ground (i.e., within the ground network 150) can have some advantages over generating agent models on the aircraft 102. For example, the ground network 150 can have access to a larger pool of computing resources (e.g., processing power), easier access to operational parameters of multiple aircraft 102 rather than just one aircraft 102, and closer proximity to human personnel trained to facilitate the generation / application of agent models.

[0026] Example operational parameters associated with an onboard network Vehicle-based communication system (e.g., from Figure 1The system 100) can supply various operational parameters to potentially serve as a basis for evaluating a proxy model for assessing user experience in an onboard network. These operational parameters can include discrete variables, continuous variables, and / or categorical variables. Numerous examples of these operational parameters will be provided below, spanning categories of the parameters including at least (1) system hardware component parameters, (2) network state parameters, (3) user behavior parameters, and / or (4) service parameters.

[0027] While the present disclosure suggests that certain of the parameters discussed herein can have a particular impact on user experience, it should be appreciated that the purpose of the present disclosure is to utilize a proxy model to determine whether and to what extent any given operational parameter of a system / network tends to positively or negatively impact user experience in an onboard network. For example, while it can be suggested that loss of ATG network coverage can negatively impact a user's experience due to the user's inability to send / receive content from the Internet, training of the proxy model can reveal that the negative impact on user experience is eliminated or at least mitigated for some users if the user only used the network for onboard, locally stored content prior to loss of coverage, if the user sent / received less than a threshold amount of data to / from the ground, and / or if the user previously only used ground-based communications for services that had a viable alternative provided locally on the vehicle.

[0028] Turning to the first category of operational parameters suggested above, system hardware component parameters include various identifiers and / or characteristics of hardware elements of the vehicle-based communication system. These hardware elements include, for example, one or more onboard units (e.g., the LRU(s) 120) that implement the onboard network in the vehicle's cabin. These system elements can also include other onboard hardware in the communication path from the LRU(s) to user devices (e.g., wireless access points, installed seatback units, etc.), and / or hardware in the communication path between the LRU(s) to the antenna(s) (up to and including the antenna itself). Figure 1

[0029] ​In particular, system hardware component parameters can include identifiers of one or more LRUs in the vehicle. Different LRUs, which can be manufactured by the same or different system providers, can provide varying capabilities and functionality. For example, different LRUs can support different (or additional) communication protocols, higher / lower bandwidths, higher / lower maximum numbers of user devices supported simultaneously, different supported applications, and / or different effective service altitudes. Thus, the identifier of an LRU itself can at least indirectly provide various operational parameters for a given vehicle, such as the communication protocol(s) used to support user usage of the onboard network, the number of devices using the network simultaneously, and / or what application(s) the user can have used in the onboard network (e.g., voice, text, web browsing, video streaming, audio streaming, email, VPN, etc.). There can be additional (or alternative) inherent differences in LRU configuration and functionality, including, for example, configured ground-based support capabilities, system fault prevention or response protocols (e.g., power cycling), or service customization capabilities. Moreover, while there are the aforementioned differences in capabilities / functionality even when different LRUs are each operating properly, actual operation of different LRUs can reveal software errors and / or hardware defects specific to each respective LRU. For example, one particular LRU can provide unexpected latency under certain conditions, or can inadvertently limit the number of devices that can be connected to the onboard network simultaneously, thereby potentially completely preventing additional users from utilizing the onboard network. In short, given any of the aforementioned factors and / or additional factors not considered above, different LRUs often prove to provide different levels of quality of user experience to users of the onboard network.

[0030] Among the operational parameters provided to the agent model, there can also be included still other aspects of the onboard hardware. These parameters can include, for example, an age of the vehicle and / or particular components onboard the vehicle, or an indication of a last maintenance time of the vehicle and / or particular component(s) thereof.

[0031] Moving to the second category of operational parameters described above, the quality of the user experience can be affected by a number of characteristics or states of the onboard network and / or the larger communication network supporting the onboard network. First and perhaps most importantly, the amount of time spent outside of the coverage area of the communication system provider can impact the user experience. For example, the communication system provider can only operate in a selected set of countries, states, provinces, etc., and thus, when the vehicle is outside of the coverage area, providing ground-based content to the vehicle can not be possible. "Dead" zones can also exist in such areas (e.g., in an ATG network, areas not within line of sight of any terrestrial base station). In these cases, the user can not have access to ground-based communications and thus can be limited to using the onboard network locally to the vehicle for services / content (e.g., consuming locally stored video / audio content). An operational parameter can include the total or percentage of time spent by the vehicle outside of the coverage area or outside of the optimal coverage area during the transit.

[0032] Operational parameters that can potentially impact the user experience can include various aspects of the vehicle-based network as a whole. For example, the user experience can be impacted by: the maximum number of onboard devices using wired and / or wireless services of the onboard network during the transit, the average number of devices using wired / wireless services of the onboard network during the transit, the total or average data throughput of user devices during the transit, the available communication bandwidth to onboard devices during the transit, the available bandwidth to particular user devices during the transit, and / or the average, maximum, minimum, or standard deviation of data throughput speed available to particular user devices of the traffic. Additionally or alternatively, operational parameters impacting the user experience can include aspects of the larger communication network outside of the vehicle (e.g., including satellites, base stations, and connections thereto). For example, operational parameters impacting the user experience of a given vehicle can include an indication of congestion of the larger communication network (e.g., average / maximum / minimum number of other vehicles in the same coverage area as the vehicle, particular communication sockets inhibited or blocked to the vehicle during the transit, and / or the amount of availability of ground-based networks during the transit, e.g., as determined based on the number or proportion of successful "heartbeat" pings to ground-based networks during the transit).

[0033] Moving to the third category of operational parameters, operational parameters that can potentially affect the user experience quality for a given user can include indicators of the user's behavior while using the onboard network. These parameters can include, for example, the total data consumed by the user in the course of the trip, a comparison of the user's data usage to an average or benchmark user (e.g., indicating a "high usage user" that uses 20%, 30%, 50%, or 100% more than the average user, or that uses more different electronic devices than the normal number at the same time), data used by the user for various individual services / applications (e.g., web browsing, texting, calling, video streaming, email, etc.), a split between wired and wireless data usage, and / or usage of background applications on the user device(s).

[0034] Additionally or alternatively, operational parameters related to user behavior can include behavior indicative of potential errors or disruptions in the onboard network. For example, a user's access to the network support pages of the onboard service provider can be indicative of the user having a problem in traffic. To provide other examples, such operational parameters can include the number or frequency of page refreshes by the user, and / or the number of lost browsing sessions, lost calls, lost packets, failed SMS messages, etc. As yet another example, a parameter can indicate a support case raised by the user during or shortly after the trip. In particular, a relevant support case can include a support case raised shortly after an initial installation, reinstallation, or modification of the configuration of the hardware and / or software onboard the vehicle. Further, a particularly relevant support case can include a support case raised by the user with reference to a time when the communication system provider determined that the onboard network was not operating nominally or optimally. In some embodiments, defining operational parameters related to user behavior can include refining the initial set of obtained operational parameters by analyzing the initial operational parameters to determine whether the user behavior does in fact correspond to an error / disruption in the onboard network (or alternatively, to a cause originating from the user(s)).

[0035] Moving to the fourth category of operational parameters potentially relevant to user experience, the operational parameters can include service parameters associated with providing in-flight network access to users. Such service parameters can include, for example, an indication of whether a user subscribes to a particular individual service within an in-flight network experience (e.g., subscribes to a particular data rate, data cap, in-flight video / audio streaming service, ground-based video / audio streaming service, etc.). Similarly, service parameters can include a user's device limitations (e.g., indicating a total number of devices or cooperating devices that a user can connect to an in-flight network at a particular transportation). Service parameters can additionally or alternatively include an indication of a user's frequency of use and / or total amount of use of in-flight networks (i.e., outside of a single given transportation that can be evaluated for user experience). Yet additionally or alternatively, service parameters can include an indication of whether a vehicle has suspended or interrupted communication services, e.g., due to maintenance of the vehicle, thereby preventing use of the in-flight network services of the vehicle that includes the vehicle.

[0036] Generally speaking, observed values of operational parameters (e.g., any combination of the above-described parameters) can be provided as one or more inputs to one or more machine learning models, which can include supervised and / or unsupervised models, examples of which will be provided in this disclosure. Different ones of these parameters can prove to be strongly, weakly, or not at all correlated to the quality of user experience of an in-flight network when applied to one or more machine learning models described herein. Moreover, different parameters can still arise that are relevant to the quality of user experience, potentially including parameters that are directly or indirectly observable from the parameters described herein.

[0037] Example machine learning techniques for proxy models Generally speaking, the proxy models of this disclosure for evaluating user experience in in-flight networks can be machine learning-based models, i.e., incorporating one or more machine learning techniques. Examples of machine learning techniques for generating and applying proxy models will be provided herein. However, it should be appreciated that additional and / or machine learning techniques can be applied in various embodiments.

[0038] Techniques for generating proxy models can utilize supervised and / or unsupervised machine learning. In supervised machine learning, a processing element (i.e., one or more processing elements) is provided with labeled training data including example inputs and their associated outputs, and can generally seek to discover a logic that maps inputs to outputs such that, when a new input is subsequently provided to the trained machine learning model, the trained machine learning model accurately predicts the correct output (or at least increases the likelihood of a correct prediction of the output). In unsupervised machine learning, the processing element can be required to find its own structure in unlabeled training data including example inputs without associated outputs.

[0039] Accordingly, in supervised and / or unsupervised machine learning models for evaluating user experience, the training data provided to the machine learning model can include values of operational parameters associated with operation of the vehicle-based communication system / network, including, for example, any of the operational parameters discussed in the foregoing sections. In unsupervised machine learning models (e.g., a k-means clustering algorithm, which will be described in this section), the processing element can be provided with operational parameter inputs without associated outputs (e.g., a set of operational parameter inputs for a corresponding session in the onboard network without a corresponding user experience score provided by the user themselves). The unsupervised model can be trained (i.e., generated) to identify structure in the input operational parameters, e.g., to determine logic, relationships, and / or trends between the input(s). On the other hand, in supervised machine learning models (e.g., an artificial neural network), the processing element can be provided with operational parameter inputs and their associated outputs (e.g., a set of operational parameter inputs for a corresponding session in the onboard network with a corresponding user experience score provided by the user themselves, e.g., through a survey response). The supervised model can be trained to accurately predict the correct user experience score based on new operational parameter inputs for a new session in the network (e.g., to predict an integer score on a 0-10 scale that represents an increasing level of satisfaction with using the onboard network). The correctness of the generated supervised machine learning model can be verified and / or updated by testing the generated model against new operational parameters with associated user experience score outputs, enabling further refinement of the supervised model (e.g., adapting the model to new network environments, different hardware, different user tendencies, changes in service offerings, etc.).

[0040] The techniques described herein can be applied at various levels of granularity with respect to vehicles and onboard networks managed by a vehicle-based communication system provider. For example, in some embodiments, the generation and application of machine learning models can be isolated to the onboard network of a single vehicle, and to the extent that a system operator can operate multiple vehicles, different proxy models (including different machine learning models) can be generated and applied for each respective vehicle. In some embodiments, proxy models can be generated using aggregated data across each of a plurality of similar or comparable vehicles (and onboard systems / networks therein) operated by the communication system provider.

[0041] Example neural network for evaluating user experience Figure 2A Elements of an example artificial neural network 200 are shown, i.e., a supervised machine learning model that can be trained to evaluate user experience in an onboard communication network (e.g., to generate a score representing the quality of user experience or user satisfaction of any given user in any given vehicle). The neural network 200 can be trained, for example, via Figure 1by the agent model module(s) 160 and / or other suitable computing devices of the present disclosure.

[0042] The neural network 200 includes a plurality of neurons arranged in a plurality of layers. These layers include an input layer 202, one or more intermediate or "hidden" layers 204-1 to 204-n, and an output layer 206. The respective inputs x1 to x a may correspond to operating parameters associated with the provision of the onboard communication network (e.g., any operating parameter of the present disclosure, or a combination thereof). In various embodiments, the inputs in the input layer can include discrete, continuous, and / or categorical parameters.

[0043] The outputs y1 in the output layer 206 can correspond to a representation of a user experience or satisfaction of any particular user (e.g., a passenger) for any particular transportation or a portion thereof. For example, the output can be a fraction on an integer scale from zero to ten, where zero represents the least satisfactory user experience and ten represents the most satisfactory user experience. In embodiments, additional or alternative outputs can be possible. For example, rather than providing only one output indicative of a composite user experience score, the neural network 200 can provide several outputs indicative of respective probabilities of each of several possible user experience scores for a given scenario. Additionally or alternatively, the output of the neural network 200 can include an indication of an action that can be taken to address an unsatisfactory user experience and / or replicate a satisfactory user experience. Such an action can include, for example, a modification to software, firmware, and / or hardware supporting the onboard network. Further examples of such actions will be provided in subsequent sections of the present disclosure.

[0044] The number of hidden layers n can be any integer greater than or equal to one. Each of the hidden layers 204-1 to 204-n can have the same or different number of neurons, where b represents the number of neurons in the hidden layer 204-1 and c represents the number of neurons in the hidden layer 204-n. The number of neurons in each hidden layer can be different from the number of inputs in the input layer 202 and / or the number of outputs in the output layer 206. The input to each neuron in each hidden layer can operate on one or more inputs from the input layer 202 and / or one or more outputs from other layers (e.g., preceding intermediate layers). In some embodiments, the output of the neural network 200 can be obtained not only from the output layer 206, but also from one or more of the intermediate layers 204-1 to 204-n. Various other configurations of the neural network 200 are possible.

[0045] Training of the neural network 200 can involve providing sets of operational parameters as input, each set of input corresponding to a particular traffic (or "trip") usage of an onboard communication network (or "session" therein). Each set of input can be labeled with a corresponding true user experience score, which serves as the "correct" output for that set of input. Thus, as the training of the neural network 200 identifies relationships between inputs, outputs, and / or other neurons, the hidden layers 204-1 through 204-n can emerge and change. The training of the neural network 200 adjusts the neural network 200 to provide correct outputs based on new sets of input data, e.g., to provide user experience scores that match (or at least are likely to match) scores explicitly provided by users regarding their onboard network experience (e.g., via survey data).

[0046] In some embodiments, the neural network 200 can be a recurrent neural network, in which outputs from at least one layer of the neural network 200 are fed back to previous layer(s) during training to provide an indication of the importance of a particular input or intermediate layer output in determining a particular decision or calculation. Further, as a result of training, the number of inputs in the input layer 202 can change, e.g., when certain inputs are identified as being less important to the output of the neural network 200.

[0047] Figure 2B An example neuron 220 is shown, which can correspond to a neuron in the neural network 200 of Figure 2A x1through x i are provided to the neuron 220. The inputs can be, for example, one of the inputs from the input layer 202 of the neural network 200, or an output of a neuron of another layer of the neural network 200 (e.g., a neuron of a hidden layer). Each input can be assigned a respective weight w1through w i , and the respective weights of each input can be determined and adjusted during training of the neural network. The weights w1through w1may include the same and / or different weights. A larger weight corresponds to an increased relevance of the input to the decision or calculation of the neuron 220, while inputs determined to be less important or completely irrelevant to the decision or calculation can be assigned a weight of zero or near zero. The weighted inputs x i through x n may be provided to a summation node 222, which is configured to provide a sum j1of the weighted inputs to a function 224. The function 224 operates on the sum j1to produce one or more outputs, including one or more final outputs of the neural network and / or one or more values to be provided to a neuron of another hidden layer of the neural network 200.

[0048] Example K-means clustering for evaluating user experience As an addition or alternative to neural network models, the agent model for assessing user experience in an on-board network can comprise a k-means clustering model and / or other forms of clustering models and / or other unsupervised machine learning models.

[0049] In particular, the proposed k-means clustering model generally seeks to classify input points into a predefined number of similar clusters k, where each different cluster of the k clusters can correspond to a different level of quality of user experience using the on-board network. For example, the model can be defined with three clusters, under this concept the respective clusters can correspond to a satisfactory, neutral and unsatisfactory user experience. In different embodiments, eleven clusters can be envisaged, under this concept each cluster can correspond to a user experience score on an integer scale from zero (worst user experience) to ten (best user experience).

[0050] The training input to the k-means clustering model comprises a plurality of data points, where each data point can for example represent a usage session of the on-board network by a respective user of the respective traffic. Each respective data point can comprise the values of one, two, three, four, five or more operational parameters measured or otherwise observed for the respective usage session. The data space from which the clustering model will identify the k clusters can thus be a one-dimensional data space, a two-dimensional space, a three-dimensional, four-dimensional or five-dimensional data space, etc.

[0051] The training of the k-means clustering model comprises iteratively defining the center and the boundaries of each of the plurality of clusters (numbered k) in the data space. At the first iteration, a respective centroid or “mean” point is defined for each of the k clusters (e.g. by randomly generating the centroid points). Each respective training input data point of the training input data points can then be associated with the cluster having the closest centroid. The k-means clustering algorithm then re-computes the new centroid of each cluster as the mean of each of the training input data points associated with the respective cluster. Each training input data point can then be re-associated with a cluster (i.e. the same or a different cluster) based on the determination of the closest centroid of the new k centroids. This iteration is repeated at any successive redefinition of the centroids and re-association of the data points until the centroids and the cluster membership no longer change. The attempt of clustering the training input is thus completed, where each training input data point has been assigned to a respective cluster of the predefined number of clusters k.

[0052] In a k-means clustering model, the assignment of data points to clusters is sensitive to the location of the initially defined centroids. In other words, if the k centroids are initially placed at different locations in the data space, the final locations of the clusters and the assignment of data points can be different. For this reason, it is often advantageous to attempt a k-means clustering algorithm two, three, four, five, or more times, where each execution produces a different set of k clusters. From these attempts, the k-means clustering model can select the "best" clustering attempt, e.g., by selecting the clustering attempt with the smallest total within-cluster variance (e.g., based on the least-squares distance between data points and their respective centroids) among all k clusters.

[0053] By selecting the best clustering attempt, a final cluster is assigned for each training input data point. The centroid of each final cluster can be the mean of its constituent data points, and for each respective final cluster, the boundary of the respective cluster contains all points in the data space for which the nearest defined centroid is the centroid of the respective cluster. Thus, any new data point thereafter can be associated with one of the final clusters defined by the k-means clustering model.

[0054] Notably, because the k-means model is unsupervised, the final defined clusters can not inherently be associated with any particular assessment of user experience quality. Rather, the respective clusters define similar types of user experiences based on the parameters used for the input to the model. Further techniques can be needed to assign a user experience value (e.g., a score) to each cluster. In some embodiments, a human user assesses the defined clusters and rates or ranks the defined clusters according to a perceived quality of user experience (e.g., by assigning an integer score on a scale of zero to ten to each cluster), e.g., based on a comparison of the labeled operational parameters with user experience scores provided by corresponding users with other sessions. In some embodiments, the labeled operational parameters with corresponding user-provided experience scores can be plotted on the trained k-means clustering model, and the user experience score for each respective cluster can be determined based on the scores of the labeled data falling within each cluster (e.g., the centroid of the respective cluster is the average of the scores of the labeled data points for which the nearest centroid is one of the k centroids).

[0055] Additionally or alternatively, in some embodiments, the artificial neural network described above can determine a user experience value for each of the k clusters. For each respective cluster, the operational parameter value of the respective centroid (and / or other points within the respective cluster) can be provided as a new input to the trained neural network model, and the output produced by the neural network (e.g., a score from zero to ten) can be used as the user experience value for the respective cluster.

[0056] Other techniques can be envisioned that combine the k-means clustering model and the artificial neural network model. For example, in embodiments, particular operational parameters used to define the data space of the k-means clustering model are selected based on knowledge generated by the neural network model. That is, operational parameters determined by the neural network model to be highly correlated with user experience can be used to define the k-means clustering data space, while operational parameters determined to be not highly correlated with user experience can not be used to define the k-means clustering data space. Additionally or alternatively, in some embodiments, the operational parameters used to define the k-means clustering data space include not only the raw operational parameters described herein, but also one or more outputs generated by the neural network model (e.g., Figure 2A the output of the hidden layer of the neural network model 200).

[0057] Yet additional or alternative machine learning models can be used for the proxy model of the present disclosure. In some embodiments, the proxy model can include additional or alternative supervised machine learning models, such as decision trees, linear regression, random forest algorithms, Bayesian classifiers, supervised clustering algorithms, and the like. In some embodiments, the proxy model can include additional or alternative unsupervised machine learning models, such as clustering algorithms different from the clustering algorithms described above.

[0058] Application of the model to improve system / network implementation In accordance with the techniques of the present disclosure, the proxy model, including one or more trained machine learning models, facilitates determining a user experience quality of any given user for any given transportation or portion thereof without explicit feedback provided by the user themselves during or after the transportation. The ability to assess user experience in this way can enable a variety of improvements to the hardware, software, and / or firmware associated with providing the onboard network to the user. These improvements can include improvements made to the onboard communication system / network to better serve subsequent user sessions. In some embodiments, the improvements to the onboard system / network can be made while the vehicle is in transportation. This section of the present disclosure will provide some examples of such improvements, but it should be understood that additional and / or alternative techniques can be envisioned in embodiments.

[0059] In embodiments, one or more computing devices, onboard or non-onboard the vehicle, can be configured to apply the trained proxy model to compute a user experience value for a user during or after a usage session in the onboard network, analyze the proxy model output to identify adjustments to the vehicle-based system / network based on the proxy model output, and / or implement the identified adjustments to the onboard system / network (e.g., during a usage session, and / or between usage sessions of the onboard network, such as between transportations or during maintenance downtime of the vehicle). For example, returning to Figure 1LRU 120 can store at least a portion of the agent model (or otherwise communicate with the agent model in the ground network 150) and identify one or more adjustments to network / service parameters during the flight of aircraft 102, thereby attempting to improve the user experience of one or more measures based on an evaluation of the user experience from previous segments of the same flight. As another example, while aircraft 102 is in flight, one or more computing devices operating as part of the ground network 150 can communicate with aircraft 102 to enable LRU 120 to make real-time modifications to network or service parameters. In embodiments, combinations of the above implementations can be implemented. That is, applying the agent model to evaluate the user experience, analyzing the agent model output to identify system / network adjustments, and / or implementing such adjustments can combine onboard and offboard computing elements of the vehicle.

[0060] In some embodiments, the identified modifications to a vehicle-based system / network may include the modification or replacement of one or more hardware components of the system implementing the vehicle-based communication network. While it may be difficult or impossible to modify or replace many hardware components when a vehicle is in use and / or in traffic, certain system hardware components of the vehicle may be flagged for maintenance and / or replacement based on a user experience score calculated by an agent model for one or more users from one or more traffic flows of the vehicle. More specifically, system hardware components of the vehicle may be identified based on specific operating parameters that are taken into account in the user experience score calculated by the agent model. For example, if the user experience in one or more sessions in an airborne network is negatively affected by a connectivity interruption or communication path loss between the vehicle hardware (e.g., antennas) and the terrestrial network, one or more antennas of the vehicle may be flagged for replacement and / or other issues such as improper wiring, cross-modulation, physical damage, and / or other problems with other airborne hardware may be checked. As another example, if the user experience is repeatedly affected by latency or reduced per-device bandwidth in multiple sessions across an onboard network spanning one or more traffic flows, the vehicle's installed LRUs can be replaced with new LRUs with greater communication throughput capabilities and / or local storage content provisioning to alleviate congestion in future use of the onboard network. If a negative user experience is identified in one or more sessions that are connected to the user device itself rather than to the hardware installed on the vehicle, the system provider can push one or more notifications to the user device to encourage remedial measures, such as application updates, operating system updates, replacement of the user device, and / or measures that can improve the experience of subsequent use of the onboard network.

[0061] In some embodiments, the identified adjustments to the vehicle-based system / network can include adjustments or replacements to one or software elements associated with the implementation of the onboard network. For example, the onboard LRUs can implement one or more updates or restorations to the LRU system software, and / or implement updates / restorations to respective applications related to any unsatisfactory user experience in the onboard network. Additionally or alternatively, in embodiments where the unsatisfactory user experience of one or more users is related to the use of a particular application (e.g., a video streaming application in communication with the ground), the LRU can push a notification to the one or more users, encouraging the one or more users to switch to a comparable equivalent of the application(s) related to the unsatisfactory experience (e.g., encouraging the user to switch to an application streaming of locally stored video content, rather than from the ground, as the former has less demand on the communication path between the vehicle and the ground).

[0062] In some embodiments, the identified adjustments to the vehicle-based system / network can include adjustments to service parameters, policies, and / or protocols of the implementation of the onboard network. As an example, if it is determined that the onboard network is not able to provide an optimal experience to a particular service, application, and / or website, the provider of the onboard communication network can encourage, discourage, and / or block the use of the particular service, application, and / or website (e.g., favoring services provided locally). Conversely, if a service, application, or website has been previously blocked, but would prove to improve the user experience if only partially serviceable, the network provider can unblock the service, application, or website, thereby improving the user experience of subsequent users. As yet another example of service-based adjustments, at least some mitigation of unsatisfactory user experience can be achieved by alerting users (e.g., via push notifications) of impending events that are found to negatively impact the user experience (e.g., alerting users of an impending loss of connection to a ground-based network, which can mitigate the inconvenience associated with attempting to access a ground-based network when the ground-based network is unavailable).

[0063] Yet additional and / or alternative adjustments to the vehicle-based system / network are possible based on the information identified through the application of the agent model of the present disclosure. In various embodiments, in response to identifying unsatisfactory user experiences in a transportation of one vehicle, in multiple transportations of one vehicle, or in multiple transportations of multiple vehicles, a human staff can intervene to investigate and / or modify aspects of the onboard network.

[0064] In some embodiments, the proxy model of the present disclosure can be used to measure the positive or negative impact of one or more system / network adjustments identified above. That is, a plurality of user experience scores for a respective user / session can be computed before and after any given adjustment to the system / network, e.g., to identify positive changes achieved via the adjustment, and replicate that adjustment across one or more additional vehicles / networks operated by the communication system provider.

[0065] Example computing device Figure 3 An example computing device 300 in which techniques of the present disclosure can be implemented in some embodiments is shown. Computing device 300 may, for example, be an onboard or offboard computing device (e.g., a server) configured to generate and / or apply a proxy model for user experience (e.g., as part of proxy model module 160 of FIG. 1). Figure 1

[0066] Computing device 300 includes a processor 302 (i.e., one or more processors), such as a microprocessor, controller, and / or other suitable type of processor. Computing device 300 also includes memory 304 (i.e., one or more computer memories), which can include volatile memory and / or non-volatile memory, containing computer-executable instructions accessible by processor 302 to cause computing device 300 to perform the actions described herein. Computing device 300 also includes a network interface 306 (i.e., one or more network communication interfaces). In some embodiments, computing device 300 includes an input / output (I / O) interface 308 (i.e., one or more input and / or output interfaces). The components of computing device 300 are operatively coupled via a computing bus 312.

[0067] ​Network interface 306 may enable computing device 300 to communicate with one or more other devices, such as base stations, other elements of terrestrial networks, and / or components of one or more vehicles. Network interface 306 may include any suitable type(s) of communication interface, such as wired and / or wireless interfaces configured to operate according to any suitable(s) of communication protocol(s). Example network interface 306 includes a TCP / IP interface, a Wi-Fi transceiver (e.g., according to the IEEE 802.11x standard family), an Ethernet transceiver, cellular network radio, satellite network radio, or any other suitable interface based on any other suitable communication protocol or standard. I / O interface 308 may include, for example, a Bluetooth® interface, a Near Field Communication (NFC) interface, a Universal Serial Bus (USB) interface, a serial interface, an infrared interface, etc., to enable receiving user input (e.g., touchscreen, keyboard, mouse, touchpad, joystick, trackball, microphone, button, etc.) and transmitting output data to the user (e.g., via a display, speaker, printer, etc.).

[0068] Returning to memory 304, the non-transitory portion of memory 304 may include agent model generation / application module 316 (i.e., one or more modules) configured to generate and / or apply agent models for user experience in an onboard network in accordance with the techniques of this disclosure. Module 316 may include non-transitory instructions executable via processor 302 to enable computing device 300 to perform actions related to generating / applying agent models.

[0069] In various embodiments, computing device 300 may include additional, fewer, and / or alternative components, including those described in this detailed description.

[0070] Example computer implementation method Figure 4A and Figure 4B Block diagrams of example computer implementations of methods 400 and 440 according to some embodiments are shown respectively. Methods 400 and / or 440 can be executed, for example, via non-transitory instructions stored in one or more computer memories. Figure 1 Components of the ground network 150 (e.g., components of the proxy model module 160) and / or Figure 3 This is achieved using a computing device 300.

[0071] from Figure 4AInitially, the method 400 includes training a proxy model to analyze values of a plurality of operational parameters associated with a vehicle-based communication network that provides a vehicle (402). The training of the proxy model trains the proxy model to generate a measure of quality of user experience in the vehicle-based communication network. The training data can include a training set of operational parameter values associated with a user’s use of the vehicle-based communication network, which in some embodiments can be labeled with a user experience measure provided by the respective user during or after use of the vehicle-based communication network (e.g., a fraction on a zero-to-ten integer scale). The proxy model can include a k-means clustering model, an artificial neural network, and / or another machine learning-based model described in the present disclosure.

[0072] The method 400 also includes obtaining a first observed set of operational parameter values corresponding to the first user’s use of the vehicle-based communication network during a first trip of the vehicle (404). The operational parameter values can include values of any of the operational parameters described in the present disclosure (e.g., hardware identification / configuration information, network status information, user behavior, etc.).

[0073] The method 400 also includes analyzing the first observed set of operational parameters via the trained proxy model (406). The analysis of the first observed set of operational parameters generates a first measure of quality of user experience in the vehicle-based communication network for the first user in the vehicle.

[0074] Moving to Figure 4B , the method 440 includes obtaining a first observed set of operational parameter values corresponding to the first user’s use of the vehicle-based communication network of the vehicle during a first trip of the vehicle (442). The operational parameter values can include values of any of the operational parameters described in the present disclosure (e.g., hardware identification / configuration information, network status information, user behavior, etc.).

[0075] The method 440 also includes analyzing the first observed set of operational parameters via the trained proxy model (444). The analysis of the first observed set of operational parameters generates a first measure of quality of user experience in the vehicle-based communication network for the first user in the vehicle. The proxy model can include a k-means clustering model, an artificial neural network, and / or another machine learning-based model described in the present disclosure.

[0076] The method 440 yet further includes causing one or more adjustments to the vehicle-based communication network based on the first measure of user experience quality (446). The one or more adjustments are for improving user experience in subsequent use of the vehicle-based communication network by the first user or one or more other users. The one or more adjustments can for example include (1) replacement, maintenance, or reconfiguration of one or more hardware components in the vehicle-based communication network, (2) replacement, update, restoration, or reconfiguration of one or more software elements in the vehicle-based communication network, (3) adjustment to availability of one or more services, applications, or websites via the vehicle-based communication network, and / or another or more techniques described in the present disclosure.

[0077] It will be appreciated that each of the methods 400 and / or 440 can include yet additional, fewer, and / or alternative actions to those shown in FIGS. 4 and 5, including any suitable actions described in this detailed description. In particular, aspects of the methods 400 and 440 can be combined with each other in various embodiments. Moreover, the order of the actions in the methods 400 and / or 440 can be different than shown in the figures in some embodiments. Figure 4A and Figure 4B It will be appreciated that each of the methods 400 and / or 440 can include yet additional, fewer, and / or alternative actions to those shown in FIGS. 4 and 5, including any suitable actions described in this detailed description. In particular, aspects of the methods 400 and 440 can be combined with each other in various embodiments. Moreover, the order of the actions in the methods 400 and / or 440 can be different than shown in the figures in some embodiments.

[0078] Additional Considerations Throughout this specification, plural instances can implement components, operations, or structures described as a single instance. Although individual operations of one or more methods are illustrated and described as separate operations, one or more of the individual operations can be performed concurrently, and nothing requires that the operations be performed in the order illustrated. Structures and functionality presented as separate components in example configurations can be implemented as a combined structure or component. Similarly, structures and functionality presented as a single component can be implemented as separate components. These and other variations, modifications, additions, and improvements fall within the scope of the subject matter of this document.

[0079] As used herein, any reference to “one embodiment” or “an embodiment” means that a particular element, feature, structure, or characteristic described in connection with the embodiment is included in at least one embodiment. The appearances of the phrase “in one embodiment” in various places in the specification are not necessarily all referring to the same embodiment.

[0080] Some embodiments can be described using the expression “coupled” and “connected” along with their derivatives. For example, some embodiments can be described using the term “coupled” to refer to two or more elements that are in direct physical or electrical contact. However, the term “coupled” can also mean that two or more elements are not in direct contact with each other, but yet still co-operate or interact with each other. Embodiments are not limited to this context.

[0081] As used herein, the terms “comprises,” “comprising,” “includes,” “including,” “has,” “having,” or any other variation thereof, are intended to cover a non-exclusive inclusion. For example, a process, method, article, or apparatus that comprises a list of elements is not necessarily limited to only those elements but can include other elements not expressly listed or inherent to such process, method, article, or apparatus. Further, unless expressly stated to the contrary, “or” refers to an inclusive or and not to an exclusive or. For example, a condition A or B is satisfied by any one of the following: A is true (or present) and B is false (or not present), A is false (or not present) and B is true (or present), and both A and B are true (or present).

[0082] Also, the use of “a” or “an” is intended to be synonymous with “one,” “one or more,” and “at least one.” As used herein in the context of describing structures and elements of the embodiments, such use is merely for convenience and to give a general sense of the referents discussed.

[0083] When implemented, any of the methods and techniques described herein, or portions thereof, can be implemented by executing software stored in one or more non-transitory, tangible, computer-readable storage media or memory, such as disk, laser disc, optical disk, semiconductor memory, biological memory, other memory device, or other storage medium, etc., in a computer or processor.

[0084] This detailed description should be construed to cover any alternatives that fall within the scope of the claims, as described below. Many alternatives not explicitly described or only implicitly described can be implemented without departing from the scope of the claims. For example, one of ordinary skill in the art will recognize that the disclosed aspects can be embodied not only as methods, but also as data structures, systems, and computer program products. Accordingly, the disclosure is not limited to purely hardware implementations, but also includes hybrid implementations that include software and hardware components. A1. A computer-implemented method implemented via one or more processors, the method comprising: training a proxy model to analyze values of a plurality of operational parameters associated with a vehicle-based communication network that provides a vehicle to generate a measure of a quality of user experience in the vehicle-based communication network based on training data, the training data comprising a training set of operational parameter values associated with usage of the vehicle-based communication network by users; obtaining a first observed set of operational parameter values corresponding to usage of the vehicle-based communication network by a first user during a first instance of travel of the vehicle; and / or analyzing the first observed set of operational parameter values via one or more trained proxy models to generate a first measure of a quality of user experience in the vehicle-based communication network for the first user in the vehicle.

[0085] A2. The computer-implemented method of aspect Al, wherein the vehicle is an airplane.

[0086] A3. The computer-implemented method of aspect Al or A2, wherein the proxy model comprises one or more unsupervised machine learning models.

[0087] A4. The computer-implemented method of aspect A3, wherein the one or more unsupervised machine learning models comprise one or more unsupervised clustering models.

[0088] A5. The computer-implemented method of aspect A4, wherein the one or more unsupervised clustering models comprise a k-means clustering model, wherein training the proxy model comprises training the k-means clustering model to classify a respective set of operational parameters for a respective user into one of a plurality of clusters associated with different levels of quality of user experience in the vehicle-based communication network.

[0089] A6. The computer-implemented method of any of aspects Al to A5, wherein the proxy model comprises one or more supervised machine learning models.

[0090] A7. The computer-implemented method of aspect A6, wherein the one or more supervised machine learning models comprise one or more artificial neural networks, wherein the training data comprises labeled data comprising a respective labeled measure of user experience for each of the training set of operational parameter values, and wherein training the proxy model comprises training the one or more artificial neural networks to generate correct measures of user experience based on the training data.

[0091] A8. The computer-implemented method of aspect A7, wherein the one or more artificial neural networks comprise one or more recurrent neural networks.

[0092] A9. The computer-implemented method of any one of aspects Al to A8, wherein the training of the proxy model comprises training one or more ground-based processing elements that are external to the vehicle.

[0093] A10. The computer-implemented method of any one of aspects Al to A9, wherein analyzing the first observation set of operational parameter values comprises analyzing the first observation set of operational parameter values via one or more processors disposed within the vehicle.

[0094] A11. The computer-implemented method of any one of aspects Al to A9, wherein analyzing the first observation set of operational parameter values comprises analyzing the first observation set of operational parameter values via one or more processors of one or more ground-based computing devices that are external to the vehicle.

[0095] A12. The computer-implemented method of any one of aspects Al to Al l, wherein the analyzing of the first observation set of operational parameter values is performed during travel of the vehicle.

[0096] A13. The computer-implemented method of any one of aspects Al to A12, wherein the plurality of operational parameters comprises one or more parameters indicative of a configuration of one or more hardware components of the vehicle-based communication network or one or more hardware elements of the vehicle-based communication network.

[0097] A14. The computer-implemented method of any one of aspects Al to A13, wherein the plurality of operational parameters comprises one or more parameters indicative of a status, availability, or capability of one or more elements of the vehicle-based communication network.

[0098] A15. The computer-implemented method of aspect A14, wherein the one or more elements of the vehicle-based communication network comprise one or more elements that are external to the vehicle.

[0099] A16. The computer-implemented method of any one of aspects Al to A15, wherein the plurality of operational parameters comprises one or more parameters indicative of a usage behavior of the user of the vehicle-based communication network.

[0100] A17. The computer-implemented method of any of aspects Al to A16, wherein the plurality of operational parameters includes one or more parameters indicative of a user’s subscription to one or more terms of service of the vehicle-based communication network.

[0101] A18. The computer-implemented method of any of aspects Al to A17, in combination with any other suitable aspect of aspects Al to A17.

[0102] A19. One or more non-transitory computer-readable media comprising instructions that, when executed via one or more processors, cause one or more computing devices to: train a proxy model to analyze values of a plurality of operational parameters associated with a vehicle-based communication network that provides vehicles based on training data to generate a measure of a quality of user experience in the vehicle-based communication network, the training data comprising a training set of operational parameter values associated with usage of the vehicle-based communication network by users; obtain a first observed set of operational parameter values corresponding to usage of the vehicle-based communication network by a first user during a first instance of travel of the vehicle; and / or analyze the first observed set of operational parameter values via one or more trained proxy models to generate a first measure of a quality of user experience in the vehicle-based communication network for the first user in the vehicle.

[0103] A20. The one or more non-transitory computer-readable media of aspect A19, wherein the vehicle is an airplane.

[0104] A21. The one or more non-transitory computer-readable media of aspect A19 or A20, wherein the proxy model comprises one or more unsupervised machine learning models.

[0105] A22. The one or more non-transitory computer-readable media of aspect A21, wherein the one or more unsupervised machine learning models comprise one or more unsupervised clustering models.

[0106] A23. The one or more non-transitory computer-readable media of aspect A22, wherein the one or more unsupervised clustering models comprise a k-means clustering model, wherein the instructions to train the proxy model comprise instructions to train the k-means clustering model to classify a respective set of operational parameters for a respective user into one of a plurality of clusters associated with different levels of quality of user experience in the vehicle-based communication network.

[0107] A24. The one or more non-transitory computer-readable media of any of aspects A19-A23, wherein the proxy model comprises one or more supervised machine learning models.

[0108] A25. The one or more non-transitory computer-readable media of aspect A24, wherein the one or more supervised machine learning models comprise one or more artificial neural networks, wherein the training data comprises labeled data comprising a respective labeled user experience measurement for each of the training set of operational parameter values, and wherein the instructions to train the proxy model comprise instructions to train the one or more artificial neural networks to generate correct measurements of user experience based on the training data.

[0109] A26. The one or more non-transitory computer-readable media of aspect A25, wherein the one or more artificial neural networks comprise one or more recurrent neural networks.

[0110] A27. The one or more non-transitory computer-readable media of any of aspects A19-A26, wherein the instructions to train the proxy model comprise instructions to train one or more ground-based processing elements external to the vehicle.

[0111] A28. The one or more non-transitory computer-readable media of any of aspects A19-A27, wherein the instructions to analyze the first observed set of operational parameter values comprise instructions to analyze the first observed set of operational parameter values via one or more processors disposed within the vehicle.

[0112] A29. The one or more non-transitory computer-readable media of any of aspects A19-A27, wherein the instructions to analyze the first observed set of operational parameter values comprise instructions to analyze the first observed set of operational parameter values via one or more processors of one or more ground-based computing devices external to the vehicle.

[0113] A30. The one or more non-transitory computer-readable media of any of aspects A19-A29, wherein the analysis of the first observed set of operational parameter values is performed during travel of the vehicle.

[0114] A31. The one or more non-transitory computer-readable media of any of aspects A19 to A30, wherein the plurality of operational parameters comprises one or more parameters indicative of a configuration of one or more hardware components of the vehicle-based communication network or one or more hardware elements of the vehicle-based communication network.

[0115] A32. The one or more non-transitory computer-readable media of any of aspects A19 to A31, wherein the plurality of operational parameters comprises one or more parameters indicative of a status, availability, or capability of one or more elements of the vehicle-based communication network.

[0116] A33. The one or more non-transitory computer-readable media of aspect A32, wherein the one or more elements of the vehicle-based communication network comprise one or more elements external to the vehicle.

[0117] A34. The one or more non-transitory computer-readable media of any of aspects A19 to A33, wherein the plurality of operational parameters comprises one or more parameters indicative of a usage behavior of the user of the vehicle-based communication network.

[0118] A35. The one or more non-transitory computer-readable media of any of aspects A19 to A34, wherein the plurality of operational parameters comprises one or more parameters indicative of a subscription of a user to one or more terms of service of the vehicle-based communication network.

[0119] A36. The one or more non-transitory computer-readable media of any of aspects A19 to A35, in combination with any other suitable aspect of aspects A19 to A35.

[0120] A37. The one or more non-transitory computer-readable media of any of aspects A19 to A36, comprising instructions to perform the method of any suitable aspect of aspects Al to A18.

[0121] A38. A computing system comprising: one or more processors; and one or more memories storing non-transitory instructions that, when executed via the one or more processors, cause the computing system to: train a proxy model to analyze values of a plurality of operational parameters associated with a vehicle-based communication network that provides vehicles to generate a measure of quality of user experience in the vehicle-based communication network based on training data comprising a training set of operational parameter values associated with usage of the vehicle-based communication network by users; obtain a first observed set of operational parameter values corresponding to usage of the vehicle-based communication network by a first user during a first instance of travel of the vehicle; and / or analyze the first observed set of operational parameter values via one or more trained proxy models to generate a first measure of quality of user experience in the vehicle-based communication network for the first user in the vehicle.

[0122] A39. The computing system of aspect A38, wherein the vehicles are airplanes.

[0123] A40. The computing system of aspect A38 or A39, wherein the proxy model comprises one or more unsupervised machine learning models.

[0124] A41. The computing system of aspect A40, wherein the one or more unsupervised machine learning models comprise one or more unsupervised clustering models.

[0125] A42. The computing system of aspect A41, wherein the one or more unsupervised clustering models comprise a k-means clustering model, wherein the instructions to train the proxy model comprise instructions to train the k-means clustering model to classify a respective set of operational parameters for a respective user into one of a plurality of clusters associated with different levels of quality of user experience in the vehicle-based communication network.

[0126] A43. The computing system of any of aspects A38 to A42, wherein the proxy model comprises one or more supervised machine learning models.

[0127] A44. The computing system of aspect A43, wherein the one or more supervised machine learning models comprise one or more artificial neural networks, wherein the training data comprises labeled data comprising a respective labeled measure of user experience for each of the training set of operational parameter values, and wherein the instructions to train the proxy model comprise instructions to train the one or more artificial neural networks to generate correct measures of user experience based on the training data.

[0128] A45. The computing system of aspect A44, wherein the one or more artificial neural networks comprise one or more recurrent neural networks.

[0129] A46. The computing system of any one of aspects A38-A45, wherein the instructions to train the proxy model comprise instructions to train one or more ground-based processing elements that are external to the vehicle.

[0130] A47. The computing system of any one of aspects A38-A46, wherein the instructions to train the proxy model comprise instructions to train one or more ground-based processing elements that are external to the vehicle.

[0131] A48. The computing system of any one of aspects A38-A46, wherein the instructions to analyze the first set of observations of operational parameter values comprise instructions to analyze the first set of observations of operational parameter values via one or more processors of one or ground-based computing devices that are external to the vehicle.

[0132] A49. The computing system of any one of aspects A38-A48, wherein the analysis of the first set of observations of operational parameter values is performed during travel of the vehicle.

[0133] A50. The computing system of any one of aspects A38-A49, wherein the plurality of operational parameters comprise one or more parameters indicative of a configuration of one or more hardware components of the vehicle-based communication network or one or more hardware elements of the vehicle-based communication network.

[0134] A51. The computing system of any one of aspects A38-A50, wherein the plurality of operational parameters comprise one or more parameters indicative of a status, availability, or capability of one or more elements of the vehicle-based communication network.

[0135] A52. The computing system of aspect A51, wherein the one or more elements of the vehicle-based communication network comprise one or more elements that are external to the vehicle.

[0136] A53. The computing system of any one of aspects A38-A52, wherein the plurality of operational parameters comprise one or more parameters indicative of a usage behavior of the user of the vehicle-based communication network.

[0137] A54. The computing system of any one of aspects A38-A53, wherein the plurality of operational parameters comprises one or more parameters indicative of a user’s subscription to one or more terms of service of the vehicle-based communication network.

[0138] A55. The computing system of any one of aspects A38-A54, in combination with any other suitable aspect of aspects A38-A54.

[0139] A56. The computing system of any one of aspects A38-A55, configured to execute instructions of one or more non-transitory computer-readable media according to any suitable aspect of aspects A19-A37.

[0140] A57. The computing system of any one of aspects A38-A56, configured to implement a method according to any suitable aspect of aspects Al-A18.

[0141] A58. Any one of aspects Al-A57 in combination with any other suitable aspect of aspects Al-A57.

[0142] B1. A computer-implemented method implemented via one or more processors, the method comprising: obtaining a first observed set of values of operational parameters corresponding to usage of a vehicle-based communication network of a vehicle by a first user during a first instance of travel of the vehicle; analyzing, via a trained proxy model, the first observed set of values of operational parameters to generate a first measure of a user experience quality of the first user in the vehicle-based communication network; and / or causing one or more adjustments to the vehicle-based communication network based on the first measure of user experience quality to improve a quality of subsequent usage of the vehicle-based communication network by the first user or by one or more other users.

[0143] B2. The computer-implemented method of aspect Bl, wherein the vehicle is an airplane.

[0144] B3. The computer-implemented method of aspect Bl or B2, wherein the proxy model comprises one or more unsupervised machine learning models.

[0145] B4. The computer-implemented method of aspect B3, wherein the one or more unsupervised machine learning models comprise one or more unsupervised clustering models.

[0146] B5. The computer-implemented method of aspect B4, wherein the one or more unsupervised clustering models comprise a k-means clustering model.

[0147] B6. The computer-implemented method of any one of aspects B1-B5, wherein the proxy model comprises one or more supervised machine learning models.

[0148] B7. The computer-implemented method of aspect B6, wherein the one or more supervised machine learning models comprise one or more artificial neural networks.

[0149] B8. The computer-implemented method of aspect B7, wherein the one or more artificial neural networks comprise one or more recurrent neural networks.

[0150] B9. The computer-implemented method of any one of aspects B1-B8, wherein the one or more processors comprise one or more processors on-board the vehicle.

[0151] B10. The computer-implemented method of any one of aspects B1-B9, wherein the one or more processors comprise one or more processors in a ground-based network external to the vehicle.

[0152] B11. The computer-implemented method of any one of aspects B1-B10, wherein the analysis of the first set of observations of operational parameter values is performed during a trip of the vehicle, and wherein the first set of observations of operational parameter values corresponds to a first portion of the trip of the vehicle.

[0153] B12. The computer-implemented method of aspect B11, wherein the causing of the one or more adjustments to the vehicle-based communication network is performed during the trip of the vehicle.

[0154] B13. The computer-implemented method of aspect B12, further comprising, after the causing of the one or more adjustments to the vehicle-based communication network: obtaining a second set of observations of operational parameter values corresponding to usage of the vehicle-based communication network by the first user during a second portion of the first trip of the vehicle; analyzing, via the trained proxy model, the second set of observations of operational parameter values to generate a second measure of a quality of user experience of the first user in the vehicle-based communication network; and / or comparing the first and second measures of the quality of user experience of the first user to determine an efficacy of the one or more adjustments to the vehicle-based communication network.

[0155] B14. The computer-implemented method of any of aspects Bl to B13, further comprising, after the causing of the one or more adjustments to the vehicle-based communication network: obtaining a further one or more observed sets of values of operational parameters corresponding to usage of the vehicle-based communication network by the first user or another one or more users during a second trip of the vehicle; analyzing, via the trained proxy model, the further one or more observed sets of values of operational parameters to generate a further one or more measures of quality of user experience in the vehicle-based communication network by the first user or the another one or more users; and / or comparing the first measure to the further one or more measures to determine efficacy of the one or more adjustments to the vehicle-based communication network.

[0156] B15. The computer-implemented method of any of aspects Bl to B14, wherein the one or more adjustments comprise replacement, maintenance, or reconfiguration of one or more hardware components in the vehicle-based communication network.

[0157] B16. The computer-implemented method of any of aspects Bl to B15, wherein the one or more adjustments comprise replacement, update, restoration, or reconfiguration of one or more software elements in the vehicle-based communication network.

[0158] B17. The computer-implemented method of any of aspects Bl to B16, wherein the one or more adjustments comprise an adjustment to availability of one or more services, applications, or websites via the vehicle-based communication network.

[0159] B18. The computer-implemented method of any of aspects Bl to B17, in combination with any other suitable aspect of aspects Bl to B17.

[0160] B19. One or more non-transitory computer-readable media comprising instructions that, when executed via one or more processors, cause one or more computing devices to: obtain a first observed set of values of operational parameters corresponding to usage of a vehicle-based communication network of a vehicle by a first user during a first trip of the vehicle; analyze, via a trained proxy model, the first observed set of values of operational parameters to generate a first measure of quality of user experience in the vehicle-based communication network by the first user; and / or cause, based on the first measure of quality of user experience, one or more adjustments to the vehicle-based communication network to improve quality of subsequent usage of the vehicle-based communication network by the first user or one or more other users.

[0161] B20. The one or more non-transitory computer-readable media of aspect B19, wherein the vehicle is an aircraft.

[0162] B21. The one or more non-transitory computer-readable media of aspect B19 or B20, wherein the proxy model comprises one or more unsupervised machine learning models.

[0163] B22. The one or more non-transitory computer-readable media of aspect B21, wherein the one or more unsupervised machine learning models comprise one or more unsupervised clustering models.

[0164] B23. The one or more non-transitory computer-readable media of aspect B22, wherein the one or more unsupervised clustering models comprise a k-means clustering model.

[0165] B24. The one or more non-transitory computer-readable media of any of aspects B19-B23, wherein the proxy model comprises one or more supervised machine learning models.

[0166] B25. The one or more non-transitory computer-readable media of aspect B24, wherein the one or more supervised machine learning models comprise one or more artificial neural networks.

[0167] B26. The one or more non-transitory computer-readable media of aspect B25, wherein the one or more artificial neural networks comprise one or more recurrent neural networks.

[0168] B27. The one or more non-transitory computer-readable media of any of aspects B19-B26, wherein the one or more processors comprise one or more processors onboard the vehicle.

[0169] B28. The one or more non-transitory computer-readable media of any of aspects B19-B27, wherein the one or more processors comprise one or more processors in a ground-based network external to the vehicle.

[0170] B29. The one or more non-transitory computer-readable media of any of aspects B19-B28, wherein the analysis of the first observed set of operational parameter values is performed during a trip of the vehicle, and wherein the first observed set of operational parameter values corresponds to a first portion of the trip of the vehicle.

[0171] B30. The one or more non-transitory computer-readable media of aspect B29, wherein the causing of the one or more adjustments to the vehicle-based communication network is performed during the first instance of travel of the vehicle.

[0172] B31. The one or more non-transitory computer-readable media of aspect B30, wherein the instructions, when executed via the one or more processors, further cause the one or more computing devices to: obtain a second set of observations of operational parameter values corresponding to usage of the vehicle-based communication network by the first user during a second portion of the first instance of travel of the vehicle; analyze, via the trained proxy model, the second set of observations of operational parameter values to generate a second measure of user experience quality in the vehicle-based communication network by the first user; and / or compare the first measure and the second measure of user experience quality by the first user to determine efficacy of the one or more adjustments to the vehicle-based communication network.

[0173] B32. The one or more non-transitory computer-readable media of any one of aspects B19-B31, wherein the instructions, when executed via the one or more processors, further cause the one or more computing devices to: obtain one or more additional sets of observations of operational parameter values corresponding to usage of the vehicle-based communication network by the first user or another one or more users during a second instance of travel of the vehicle; analyze, via the trained proxy model, the one or more additional sets of observations of operational parameter values to generate one or more additional measures of user experience quality in the vehicle-based communication network by the first user or the another one or more users; and / or compare the first measure and the one or more additional measures to determine efficacy of the one or more adjustments to the vehicle-based communication network.

[0174] B33. The one or more non-transitory computer-readable media of any one of aspects B16-B32, wherein the one or more adjustments comprise replacement, maintenance, or reconfiguration of one or more hardware components in the vehicle-based communication network.

[0175] B34. The one or more non-transitory computer-readable media of any one of aspects B16-B33, wherein the one or more adjustments comprise replacement, update, restoration, or reconfiguration of one or more software elements in the vehicle-based communication network.

[0176] B35. The computer-implemented method of any of aspects B16 through B34, wherein the one or more adjustments comprise an adjustment to availability of one or more services, applications, or websites via the vehicle-based communication network.

[0177] B36. The one or more non-transitory computer-readable media of any of aspects B16 through B35, in combination with any other suitable aspect of aspects B19 through B35.

[0178] B37. The one or more non-transitory computer-readable media of any of aspects B19 through B36, comprising instructions to perform the method of any suitable aspect of aspects B1 through B18.

[0179] B38. A computing system comprising: one or more processors; and one or more memories storing non-transitory instructions that, when executed via the one or more processors, cause the computing system to: obtain a first set of observations of values of operational parameters corresponding to usage of a vehicle-based communication network of a vehicle by a first user during a first instance of travel of the vehicle; analyze, via a trained proxy model, the first set of observations of values of operational parameters to generate a first measure of a quality of user experience in the vehicle-based communication network by the first user; and / or cause one or more adjustments to the vehicle-based communication network based on the first measure of quality of user experience to improve a quality of subsequent usage of the vehicle-based communication network by the first user or by one or more other users.

[0180] B39. The computing system of aspect B38, wherein the vehicle is an airplane.

[0181] B40. The computing system of aspect B38 or B39, wherein the proxy model comprises one or more unsupervised machine learning models.

[0182] B41. The computing system of aspect B40, wherein the one or more unsupervised machine learning models comprise one or more unsupervised clustering models.

[0183] B42. The computing system of aspect B41, wherein the one or more unsupervised clustering models comprise a k-means clustering model.

[0184] B43. The computing system of any of aspects B38 through B42, wherein the proxy model comprises one or more supervised machine learning models.

[0185] B44. The computing system of aspect B43, wherein the one or more supervised machine learning models comprise one or more artificial neural networks.

[0186] B45. The computing system of aspect B44, wherein the one or more artificial neural networks comprise one or more recurrent neural networks.

[0187] B46. The computing system of any one of aspects B38-B45, wherein the one or more processors comprise one or more processors on-board the vehicle.

[0188] B47. The computing system of any one of aspects B38-B46, wherein the one or more processors comprise one or more processors in a ground-based network external to the vehicle.

[0189] B48. The computing system of any one of aspects B38-B47, wherein the analysis of the first set of observations of operational parameter values is performed during a trip of the vehicle, and wherein the first set of observations of operational parameter values correspond to a first portion of the trip of the vehicle.

[0190] B49. The computing system of aspect B48, wherein the causing of the one or more adjustments to the vehicle-based communication network is performed during the trip of the vehicle.

[0191] B50. The computing system of aspect B49, wherein the instructions, when executed via the one or more processors, further cause the computing system to: obtain a second set of observations of operational parameter values corresponding to usage of the vehicle-based communication network by the first user during a second portion of the first trip of the vehicle; analyze, via the trained agent model, the second set of observations of operational parameter values to generate a second measure of user experience quality of the first user in the vehicle-based communication network; and / or compare the first measure and the second measure of user experience quality of the first user to determine efficacy of the one or more adjustments to the vehicle-based communication network.

[0192] B51. The computing system of any one of aspects B38-B50, wherein, when executed by the one or more processors, the instructions further cause the computing system to: obtain a further one or more observed sets of operational parameter values corresponding to usage of the vehicle-based communication network by the first user or another user or users during a second instance of travel by the vehicle; analyze, via the trained proxy model, the further one or more observed sets of operational parameter values to generate a further one or more measures of user experience quality of the first user or the other user or users in the vehicle-based communication network; and / or compare the first measure to the further one or more measures to determine efficacy of the one or more adjustments to the vehicle-based communication network.

[0193] B52. The computing system of any one of aspects B16-B51, wherein the one or more adjustments comprise replacement, maintenance, or reconfiguration of one or more hardware components in the vehicle-based communication network.

[0194] B53. The computing system of any one of aspects B16-B52, wherein the one or more adjustments comprise replacement, update, restoration, or reconfiguration of one or more software elements in the vehicle-based communication network.

[0195] B54. The computer-implemented method of any one of aspects B16-B53, wherein the one or more adjustments comprise an adjustment to availability of one or more services, applications, or websites via the vehicle-based communication network.

[0196] B55. The computing system of any one of aspects B16-B54, in combination with any other suitable aspect of aspects B38-B54.

[0197] B56. The computing system of any one of aspects B38-B55, configured to execute the instructions of one or more non-transitory computer-readable media of any suitable aspect of aspects B19-B37.

[0198] B57. The computing system of any one of aspects B38-B56, configured to implement the method of any suitable aspect of aspects B1-B18.

[0199] B58. Any one of aspects B1-B57 in combination with any other suitable aspect of aspects B1-B57.

[0200] C1. Any of aspects A1-A58 or B1-B58, in combination with any other suitable aspect of aspects A1-A58 or B1-B58.

[0201] Thus, numerous modifications and variations are possible in light of the above teachings. Therefore, the method and apparatus described herein are meant to be illustrative only and not limiting, as the scope of the application is to be given by the claims.

Claims

1. A computer-implemented method implemented via one or more processors, the method comprising: The agent model is trained based on training data to analyze the values ​​of multiple operational parameters associated with a transportation-based communication network that provides transportation, thereby generating a measurement of the quality of user experience in the transportation-based communication network. The training data includes a training set of operational parameter values ​​associated with user use of the transportation-based communication network. Obtain a first set of observations corresponding to the use of the vehicle-based communication network by the first user during the first traffic trip of the vehicle; and The first set of observations, which analyzes the values ​​of operating parameters via one or more trained agent models, is used to generate a first measurement of the user experience quality of the first user in the vehicle within the vehicle-based communication network.

2. The computer-implemented method according to claim 1, wherein, The means of transportation mentioned is an airplane.

3. The computer-implemented method according to claim 1 or claim 2, wherein, The proxy model includes a k-means clustering model, wherein training the proxy model includes training the k-means clustering model to classify the corresponding user's corresponding set of operational parameters into one of a plurality of clusters associated with different user experience quality levels in the vehicle-based communication network.

4. The computer-implemented method according to any one of claims 1-3, wherein, The proxy model includes one or more artificial neural networks, wherein the training data includes labeled data, the labeled data including a corresponding labeled user experience measurement of one of the operating parameter values ​​in the training set, and wherein training the proxy model includes training the one or more artificial neural networks to generate a correct measurement of the user experience based on the training data.

5. The computer-implemented method according to any one of claims 1-4, wherein, Training the agent model includes training one or more ground-based processing units outside the vehicle.

6. The computer-implemented method according to any one of claims 1-5, wherein, The first set of observations for analyzing operating parameter values ​​includes the first set of observations for analyzing operating parameter values ​​via one or more processors disposed within the vehicle.

7. The computer-implemented method according to any one of claims 1-6, wherein, The first set of observations for analyzing operating parameter values ​​includes the first set of observations for analyzing operating parameter values ​​via one or more processors of one or more ground-based computing devices external to the vehicle.

8. The computer-implemented method according to any one of claims 1-7, wherein, The plurality of operating parameters include one or more parameters that indicate the configuration of one or more hardware components of the vehicle-based communication network or the one or more hardware elements of the vehicle-based communication network.

9. The computer-implemented method according to any one of claims 1-8, wherein, The plurality of operating parameters include one or more parameters that indicate the status, availability, or capability of one or more elements of the vehicle-based communication network.

10. The computer-implemented method according to any one of claims 1-9, wherein, The plurality of operating parameters include one or more parameters that indicate the user's usage behavior of the vehicle-based communication network.

11. The computer-implemented method according to any one of claims 1-10, wherein, The plurality of operating parameters include one or more parameters that indicate a user's subscription to one or more terms of service of the vehicle-based communication network.

12. One or more non-transitory computer-readable media, including instructions that, when executed via one or more processors, cause one or more computing devices to: The agent model is trained based on training data to analyze the values ​​of multiple operational parameters associated with a transportation-based communication network that provides transportation, thereby generating a measurement of the quality of user experience in the transportation-based communication network. The training data includes a training set of operational parameter values ​​associated with user use of the transportation-based communication network. Obtain a first set of observations corresponding to the use of the vehicle-based communication network by the first user during the first traffic trip of the vehicle; and The first set of observations, which analyzes the values ​​of operating parameters via one or more trained agent models, is used to generate a first measurement of the user experience quality of the first user in the vehicle within the vehicle-based communication network.

13. One or more non-transitory computer-readable media according to claim 12, wherein, The means of transportation mentioned is an airplane.

14. One or more non-transitory computer-readable media according to claim 12 or 13, wherein, The proxy model includes a k-means clustering model, wherein the instructions for training the proxy model include instructions for training the k-means clustering model to classify the corresponding user's set of operation parameters into one of a plurality of clusters associated with different user experience quality levels in the vehicle-based communication network.

15. One or more non-transitory computer-readable media according to any one of claims 12-14, wherein, The proxy model includes one or more artificial neural networks, wherein the training data includes labeled data, the labeled data including corresponding labeled user experience measurements for each of the values ​​of the operating parameters in the training set, and wherein the instructions for training the proxy model include instructions for training the one or more artificial neural networks to generate correct measurements of the user experience based on the training data.

16. One or more non-transitory computer-readable media according to any one of claims 12-15, wherein, The instructions used to train the agent model include instructions used to train one or more ground-based processing elements outside the vehicle.

17. One or more non-transitory computer-readable media according to any one of claims 12-16, wherein, The instructions for analyzing the first set of observations of operating parameter values ​​include instructions for analyzing the first set of observations of operating parameter values ​​via one or more processors disposed within the vehicle.

18. One or more non-transitory computer-readable media according to any one of claims 12-17, wherein, The instructions for analyzing the first set of observations of operating parameter values ​​include instructions for analyzing the first set of observations of operating parameter values ​​via one or more processors of one or more ground-based computing devices outside the vehicle.

19. One or more non-transitory computer-readable media according to any one of claims 12-18, wherein, The plurality of operating parameters include one or more parameters that indicate the configuration of one or more hardware components of the vehicle-based communication network or the one or more hardware elements of the vehicle-based communication network.

20. One or more non-transitory computer-readable media according to any one of claims 12-19, wherein, The plurality of operating parameters include one or more parameters that indicate the status, availability, or capability of one or more elements of the vehicle-based communication network.

21. One or more non-transitory computer-readable media according to any one of claims 12-20, wherein, The plurality of operating parameters include one or more parameters that indicate the user behavior of the vehicle-based communication network.

22. One or more non-transitory computer-readable media according to any one of claims 12-21, wherein, The plurality of operational parameters include one or more parameters that instruct a user to subscribe to one or more terms of service of the vehicle-based communication network.

23. A computing system, comprising: One or more processors; and One or more memories storing non-transitory instructions, which, when executed via one or more processors, cause the computing system to: The agent model is trained based on training data to analyze the values ​​of multiple operational parameters associated with a transportation-based communication network that provides transportation, thereby generating a measurement of the quality of user experience in the transportation-based communication network. The training data includes a training set of operational parameter values ​​associated with user use of the transportation-based communication network. Obtain a first set of observations corresponding to the use of the vehicle-based communication network by the first user during the first traffic trip of the vehicle; and The first set of observations, which analyzes the values ​​of operating parameters via one or more trained agent models, is used to generate a first measurement of the user experience quality of the first user in the vehicle within the vehicle-based communication network.

Images