Abstract: The technology relates to real-time state estimation and sensor management. A method for real-time state estimation and sensor management may include receiving telemetry from a fleet of aerial vehicles, storing telemetry in a telemetry buffer, generating a real-time state estimate of an aerial vehicle in the fleet using a group of estimators, the estimators being of one or more types, such as a sensor management estimator, a bias and noise management estimator, and a physical modeling estimator, and providing the real-time state estimate of the aerial vehicle to a job in a fleet management system.

Abstract: The technology relates to aerial vehicle launch and land site selection. A method for determining beneficial launch and land sites may include computing a launch delay for a desired time period for each cell in a grid map with a target zone and an existing site located on the grid map, computing a flight time to target for a delay time that accounts for a launch delay, computing a launch time to target based on the launch delay and the flight time to target, receiving geographical restrictions data, and determining an efficiency benefit over the existing site based on the geographical restrictions data and a comparison of the launch time to target of each cell with the launch time to target of the existing site for the desired time period.

Abstract: The technology relates to health and lifetime estimation for a lighter-than-air (LTA) vehicle. An LTA vehicle health and lifetime estimation system may include a processor and a memory storing instructions executable by the processor to cause the processor to implement an estimation service for determining a remaining lifetime output and a simulator for simulating a terminal event based on the remaining lifetime output. The estimation service may include a thermal model configured to determine a gas temperature, a gas and air estimator configured to estimate a gas amount and an air amount remaining in a balloon of the LTA vehicle, a leak rate estimator configured to estimate a leak rate, and a zero pressure estimator configured to determine the remaining lifetime output based on the leak rate. The system also may include an air flow estimator configured to determine an air mass flow rate based on the air amount.

Abstract: The technology relates to health and lifetime estimation for a lighter-than-air vehicle. A method for vehicle health and lifetime estimation may include receiving flight data inputs associated with a vehicle, determining a gas temperature based on the flight data inputs, estimating a gas amount remaining in a balloon envelope of the vehicle, estimating a gas leak rate based on the gas amount, estimating a component lifetime representing an estimated lifetime of a component of the vehicle, and determining a remaining lifetime output based on one or both of the gas leak rate or the component lifetime, the remaining lifetime output indicating a remaining lifetime estimate for the vehicle. The method also may include simulating a terminal event and causing the vehicle to take an action based on the remaining lifetime estimate.

Abstract: The technology relates to health and lifetime estimation for a lighter-than-air vehicle. A method for vehicle health and lifetime estimation may include receiving flight data inputs associated with a vehicle, determining a gas temperature based on the flight data inputs, estimating a gas amount remaining in a balloon envelope of the vehicle, estimating a gas leak rate based on the gas amount, and determining a remaining lifetime output based on the gas leak rate, the remaining lifetime output indicating a remaining lifetime estimate for the vehicle. The method also may include simulating a terminal event and causing the vehicle to take an action based on the remaining lifetime estimate.

Abstract: The technology described here relates to LTA vehicle launch configuration and in-flight optimization. A method for optimizing for an objective of an LTA vehicle launch may include receiving a desired objective, receiving known parameters of the LTA vehicle, including a pressure threshold, performing probabilistic calculations based on the desired objective and the known parameters, the probabilistic calculations configured to model setup parameters and to output probabilities for the setup parameters, the output indicating probabilities that a simulated vehicles would achieve the desired objective. The method also includes selecting a setup parameter value based on a high probability indicated in the output. Also described is an LTA vehicle launch configuration system implementing a thermal model, a physics model, and a fill and ballast tool, including an altitude range estimator, a gas-air estimator, and a pre-flight ballast model.

Abstract: Methods and systems for controlling a group of aerial vehicles to meet a connectivity service objective are provided. A method may include causing aerial vehicles in the group to arrive at a target service area during a given arrival time window associated with the connectivity service objective, which may indicate a desired probability of service coverage of the target service area. The method further includes calculating a probability of service coverage of the target service area for the vehicles for a time period after the vehicles are expected to arrive at the target service area, determining whether the probability of service coverage meets a threshold, and causing the vehicles to operate according to a station seeking flight policy during the time period.

Abstract: The technology relates to an intelligent weather data service for providing optimized weather forecast data using real time weather observation data in response to weather forecast requests. A weather data system may include a scorer job configured to obtain weather observation data and weather forecast data from a plurality of weather data sources and to generate a plurality of error scores for the weather forecast data over a predetermined time period, each of the plurality of error scores corresponding to one of the plurality of weather data sources and indicating, for a cell on a grid map, an amount or a probability of error in the weather forecast data by a weather data source in relation to the weather observation data; an observations index configured to index the weather observation data; a data store configured to store the plurality of error scores; and a fusion layer configured to receive and respond to a weather data request using the plurality of error scores.

Abstract: The technology relates to dynamic risk assessment of a fleet of aircraft. A method for dynamic risk assessment of a flight in a fleet of aircraft includes fetching state data for a flight from a distributed database configured to store state data for flights being performed by the fleet of aircraft; running simulations, by risk modules, based on the state data for the flight, each of the risk modules configured to assess a type of risk and output a risk score, the simulations resulting in risk scores for the flight; merging the risk scores for the flight by stacking the risk scores for the flight beginning with a highest risk score until a merged risk threshold is exceeded or each of the risk scores have been stacked; providing merged risk score data to an automated system configured to generate an automated message to the flight in response to the merged risk score data.

Abstract: The technology relates to recovery and test dispatchers for fleet management and flight planning systems. A method for dispatching to a decommissioned vehicle zone includes receiving a first input comprising state data of an aerial vehicle and a second input comprising an objective; determining, using the first input, the aerial vehicle meets a set of criteria associated with the objective; selecting a dispatcher configured to dispatch a vehicle to a decommissioned vehicle zone; allocating the aerial vehicle to the dispatcher; and selecting, by the dispatcher, the decommissioned vehicle zone to which the aerial vehicle is to be dispatched, wherein the objective comprises a test objective or recovery objective, and wherein the dispatcher comprises a test dispatcher or recovery dispatcher.

Abstract: The technology relates to service dispatchers for fleet management and flight planning systems. A method for dispatching vehicles for service includes receiving, by a service dispatcher, a first input comprising a plurality of state data of a fleet of aerial vehicles and a second input comprising a service objective; determining, using the first input, that one or more aerial vehicles in the fleet meets a set of criteria associated with performing the service objective; submitting a bid on the one or more aerial vehicles, the bid configured to induce a vehicle allocator to allocate the one or more aerial vehicles to the service dispatcher; receiving an allocation of some or all of the one or more aerial vehicles; generating a plan to use the allocation to perform the service objective; and dispatching the allocation according to the plan.

Abstract: Methods and systems for controlling a group of aerial vehicles to meet a connectivity service objective are provided. A method may include causing aerial vehicles in the group to arrive at a target service area during a given arrival time window associated with the connectivity service objective, which may indicate a desired probability of service coverage of the target service area. The method further includes calculating a probability of service coverage of the target service area for the vehicles for a time period after the vehicles are expected to arrive at the target service area, determining whether the probability of service coverage meets a threshold, and causing the vehicles to operate according to a station seeking flight policy during the time period.

Abstract: The technology relates to controllers for lighter-than-air vehicles using deep reinforcement learning. A method for generating an objective-directed controller for an aerial vehicle includes defining an action space for an objective-directed controller, providing a set of feature vectors and the action space as input to a simulation module, training, by a learning module, the objective-directed controller according to a reward function correlated with the desired objective, evaluating the trained objective-directed controller, and storing high performing trained objective-directed controller.

Abstract: The technology relates to generating a flight map for an aircraft. For instance, this may include, running a plurality of simulations by placing a simulated aircraft within each cell of a grid representing areas of the earth and using predicted wind data. Each simulations identifies a cell in which each aircraft is located at the end of the simulation. A connection graph may be generated using any identified cells. The connection graph may be used to determine a flight map for an actual aircraft using a cost function and iterating from a destination cell to an initial cell. The flight map may be used to determine a route for the actual aircraft. In some examples, the flight map may be refined by running further simulations. The refined flight map may then be used to determine a route for the actual aircraft.

Abstract: A method for planning a high altitude platform-based communication network includes aggregating data from at least one data source, wherein the data includes environmental data. Based on the aggregated data, a plurality of network expansion potential scores are computed according to geographic location. A visual output is generated based on the computed plurality of network expansion potential scores.

Abstract: The technology relates to navigating aerial vehicles using deep reinforcement learning techniques to generate flight policies. An operational system for controlling flight of an aerial vehicle may include a computing system configured to process an input vector representing a state of the aerial vehicle and output an action, an operation-ready policies server configured to store a trained neural network encoding a learned flight policy, and a controller configured to control the aerial vehicle. The input vector may be processed using the trained neural network encoding the learned flight policy.

Abstract: The technology relates to navigating aerial vehicles using deep reinforcement learning techniques to generate flight policies. A computing system may include a simulator configured to produce simulations of a flight of the aerial vehicle in a region of an atmosphere, a replay buffer configured to store frames of the simulations, and a learning module having a deep reinforcement learning architecture configured to, by a reinforcement learning algorithm, process an input of a set of frames, and output a neural network encoding a learned flight policy. A meta-learning system may include stacks of learning systems, a coordinator configured to provide an instruction to the learning systems that includes a parameter and a start time, and an evaluation server configured to evaluate resulting rewards from learned flight policies generated by the learning systems.

Abstract: A system for controlling an aerial vehicle includes an aerial vehicle, a ballast coupled to the aerial vehicle, a server including a processor and a memory, and a wireless communication link that communicatively couples the aerial vehicle and the server. the memory stores instructions that, when executed by the processor, cause the server to receive weather data, determine, based on the weather data, that the aerial vehicle is experiencing, or is expected to experience, weather that satisfies a predetermined criterion, and cause the aerial vehicle to decouple at least a portion of the ballast based on a result of the determination.

Abstract: The technology relates to fleet management and flight planning for a fleet of aerial vehicles. For instance, a plurality of state messages may be received from the aerial vehicles. Each of the aerial vehicles may be allocated to one of a plurality of dispatcher software modules by allocating aerial vehicles according to a set of rules and after allocating aerial vehicles according to a set of rules, allocating any remaining aerial vehicles using a requesting process where one or more of the plurality of dispatcher software modules submits requests.

Abstract: The present disclosure relates to systems and methods for forecasting power usage of an aerial vehicle. An illustrative system includes an aerial vehicle including at least one component, and a computing device communicatively coupled to the aerial vehicle. The computing device includes a processor and a memory storing instructions which, when executed by the processor, cause the computing device to receive power consumption data corresponding to the at least one component, and generate a simulation model of power usage based on the power consumption data corresponding to the at least one component.