ACAS X anti-collision decision-making method, system and equipment based on probability prediction

Abstract

The invention provides an ACAS X anti-collision decision-making method, system and equipment based on probability prediction, and relates to the technical field of aviation safety, and the method comprises the steps: when a local machine detects a target machine, setting the relative motion state information of the two machines at an initial moment as an observation value at the initial moment, setting the variance at the initial moment as 0, and setting the variance at the initial moment as 0; constructing a state vector at an initial moment; establishing a discrete time system state equation for describing time evolution of the state vector; for each action in the candidate action set, on the basis of the state vector at the initial moment and a system state equation, online predicting future probability distribution of the state vector after the local machine executes the current action; based on the future probability distribution of the state vector, calculating the conflict probability of the local machine and the target machine in the future after the local machine executes the current action; and the action with the minimum conflict probability is determined as the optimal anti-collision decision instruction from the candidate action set, so that the problems of poor flexibility, difficulty in updating and insufficient real-time performance of an existing anti-collision system are solved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of aviation safety technology, and particularly relates to an ACAS X anti-collision decision method, system and equipment based on probability prediction. BACKGROUND

[0002] Air traffic management system (ATM) and air collision avoidance system (TCAS) play a vital role in ensuring flight safety. During flight, the aircraft may face the approach of other target aircraft, and if appropriate horizontal maneuvering measures cannot be taken in time, near-space collision may occur. In order to reduce the risk of collision, the effect of action under different flight states needs to be evaluated, and effective collision avoidance strategies need to be developed.

[0003] The traditional rule-based air collision avoidance system (such as TCAS II) is prone to produce excessive alarms in complex scenarios, and even reverse or enhanced phenomena, increasing the operating burden of pilots. Therefore, ACAS Xa proposes an optimization method based on partially observable Markov decision process (POMDP), which improves the robustness and optimality of anti-collision decision through offline cost table and online table lookup. However, the generation of offline cost table is usually based on high-dimensional simulation, which is computationally expensive and difficult to update dynamically. SUMMARY

[0004] The technical problem to be solved by the present application is the poor flexibility, difficulty in updating and insufficient real-time performance of the existing ACAS Xa anti-collision system due to the dependence on offline precomputed cost table. The present application provides an ACAS X anti-collision decision method, system and equipment based on probability prediction, which solves the above problems.

[0005] The present application is implemented by the following technical solutions: In a first aspect, the present application provides an ACAS X anti-collision decision method based on probability prediction, comprising: When the local machine detects a target machine, the relative motion state information of the two machines at the initial time is set as the observation value at the initial time, the variance at the initial time is set as 0, and the state vector at the initial time is constructed; the relative motion state information includes the relative height of the two machines, the horizontal distance of the two machines, the vertical speed of the local machine and the vertical speed of the target machine; A discrete-time system state equation is established for describing the evolution of the state vector over time; For each action in the candidate action set, the future probability distribution of the state vector after the local machine executes the current action is predicted online based on the state vector at the initial time and the system state equation; the candidate action set includes ascending action, descending action and level flight action; Based on the future probability distribution of the state vector, calculate the probability of a future conflict between the local machine and the target machine after the local machine performs the current action; From the set of candidate actions, the action with the lowest probability of conflict is determined as the optimal collision avoidance decision instruction.

[0006] Optionally, the system state equation is expressed as: x(t+Δt) = A(Δt)x(t) + G(Δt,x(t))w(t) + B(Δt)u(t,x(t)) Where x(t) is the state vector at time t; x(t+Δt) is the state vector at time t+Δt; A(Δt) represents the state transition matrix; G(Δt,x(t)) represents the noise gain matrix at time t; w(t) represents the system noise vector at time t; B(Δt) is the control input matrix; and u(t,x(t)) represents the control vector at time t, whose values ​​correspond to different actions in the candidate action set.

[0007] Optionally, the approximate value of the control vector is:

[0008]

[0009] in, Let r' be the vertical acceleration of the machine at time t, and r' be the rate of change of the horizontal distance between the two machines. S is the acceleration due to gravity; RA Indicates a release instruction; Let be the vertical speed of the machine at time t.

[0010] Optionally, for each action in the candidate action set, based on the state vector at the initial moment and the system state equation, predicting online the future probability distribution of the state vector after the local machine executes the current action includes: For each action in the candidate action set, perform the following operation: Based on the probability distribution of the vertical velocity of the machine in the state vector at the initial moment, calculate the delay probability of the machine executing the current action; Based on the delay probability, the statistical properties of the control vector are calculated; the statistical properties include the expected value and the variance. Substituting the statistical characteristics of the control vector into the system state equation, and solving it recursively, the future probability distribution of the state vector after the machine performs the current action is obtained.

[0011] Optionally, the delay probability of the aircraft performing the current action is calculated based on the probability distribution of the vertical speed of the aircraft in the state vector at the initial time, and the delay probability of the aircraft performing the current action comprises: determining a probability density function of the vertical speed of the aircraft based on the state vector at the initial time; if the current action is a descending action, integrating the probability density function of the vertical speed of the aircraft on the interval [-1500, ∞), and taking the integral result as the delay probability of the aircraft performing the descending action; if the current action is an ascending action, integrating the probability density function of the vertical speed of the aircraft on the interval (-∞, +1500], and taking the integral result as the delay probability of the aircraft performing the ascending action; if the current action is a level flight action, the delay probability of the aircraft performing the level flight action is 0.

[0012] Optionally, the conflict probability of the aircraft after performing the current action with the target aircraft in the future is calculated based on the future probability distribution of the state vector, and the conflict probability of the aircraft after performing the current action with the target aircraft in the future comprises: extracting a mean value and a variance of the relative height of the two aircraft from the future probability distribution of the state vector; based on the mean value and the variance of the relative height, calculating a probability that the relative height is in a conflict interval by using a Gaussian distribution density function; taking the probability that the relative height is in the conflict interval as the conflict probability of the aircraft after performing the current action with the target aircraft in the future.

[0013] Optionally, the mean value and the variance of the relative height of the two aircraft are extracted from the future probability distribution of the state vector, and the method comprises: recording horizontal distances of the two aircraft at each time point within a prediction time range of the future probability distribution of the state vector; determining a time point corresponding to the minimum horizontal distance as a closest time of the two aircraft; extracting a mean value and a variance of the relative height of the two aircraft at the closest time from the future probability distribution of the state vector.

[0014] Optionally, the conflict probability is calculated according to the following formula:

[0015] wherein, P conflict is the conflict probability; h is the relative height of the two aircraft at the closest time; μ is the mean value of the relative height of the two aircraft at the closest time; σ is the variance of the relative height of the two aircraft at the closest time; h min is the minimum value of the conflict interval; h maxis the maximum value of the conflict interval.

[0016] In a second aspect, the present application provides a conflict probability prediction based ACAS X collision avoidance decision system, comprising: a state perception module, configured to model relative motion state information of two aircrafts at an initial time as random variables when a local aircraft detects a target aircraft, and to construct a state vector at the initial time; the relative motion state information includes relative altitude of the two aircrafts, horizontal distance of the two aircrafts, vertical speed of the local aircraft and vertical speed of the target aircraft; a prediction module, configured to establish a discrete-time system state equation for describing evolution of the state vector over time; for each action in a candidate action set, based on the state vector at the initial time and the system state equation, to predict a future probability distribution of the state vector after the local aircraft executes a current action; the candidate action set includes an ascending action, a descending action and a level flight action; a probability evaluation module, configured to calculate a future conflict probability of the local aircraft and the target aircraft after the local aircraft executes the current action based on the future probability distribution of the state vector; a decision generation module, configured to determine an action with the minimum conflict probability as an optimal collision avoidance decision instruction from the candidate action set.

[0017] In a third aspect, the present application provides an electronic device, the computer device comprising a memory and a processor, the memory storing a computer program, and the processor executing the computer program to realize the ACAS X collision avoidance decision method based on probability prediction according to any one of the first aspect.

[0018] Compared with the prior art, the present application has the following advantages and beneficial effects: The present application provides an ACAS X collision avoidance decision method based on probability prediction, which models flight states (relative altitude, horizontal distance, vertical speed) as random variables, constructs a state equation and performs probability prediction online, calculates a future conflict probability of the local aircraft and the target aircraft after the local aircraft executes different actions through the probability distribution of the state vector, and selects a decision with the lowest risk based thereon. This method discards the mode of ACAS Xa relying on a large, fixed offline cost table, and directly generates a collision avoidance decision according to real-time perceived flight states, which enables the system to immediately respond to rapidly changing airspace environments, complex multi-aircraft interaction scenarios and sudden changes in flight states, solves the problems of poor flexibility, difficult updating and insufficient real-time performance of offline lookup methods, and significantly improves the adaptability and reliability of the system in real complex environments. Therefore, the present application provides an ACAS X collision avoidance decision method based on probability prediction, which is used to dynamically generate a collision avoidance decision in combination with real-time states during flight to improve adaptability and real-time performance. BRIEF DESCRIPTION OF DRAWINGS

[0019] In order to more clearly illustrate the technical solutions of the exemplary embodiments of the present application, the following will briefly introduce the drawings needed to be used in the embodiments. It should be understood that the following drawings only show some of the embodiments of the present application, and therefore should not be considered as a limitation to the scope. For those skilled in the art, other related drawings can also be obtained without creative effort based on these drawings. In the drawings: Figure 1 A flowchart of a method for ACAS X collision avoidance decision based on probability prediction provided by an embodiment of the present application; Figure 2 A structure diagram of a system for ACAS X collision avoidance decision based on probability prediction provided by an embodiment of the present application. DETAILED DESCRIPTION

[0020] In order to make the objectives, technical solutions and advantages of the present application clearer, the following will further describe the present application in detail with reference to the embodiments and drawings, the exemplary embodiments of the present application and the descriptions thereof are only used to explain the present application, and should not be considered as a limitation to the present application.

[0021] In order to solve the problems of poor flexibility, difficulty in updating and insufficient real-time performance of the existing ACAS Xa collision avoidance system due to the dependence on offline pre-computed cost table, an embodiment of the present application provides a method, system and device for ACAS X collision avoidance decision based on probability prediction.

[0022] Please refer to Figure 1 A flowchart of a method for ACAS X collision avoidance decision based on probability prediction provided by an embodiment of the present application, the following Figure 1 will introduce the method for ACAS X collision avoidance decision based on probability prediction shown in the flowchart.

[0023] S101, when the local aircraft detects a target aircraft, the relative motion state information of the two aircrafts at the initial time is set as the observation value at the initial time, the variance at the initial time is set as 0, and the state vector at the initial time is constructed.

[0024] In the specific implementation process, when the local air collision avoidance system (TCAS) detects that there is a target aircraft in the nearby airspace, the local aircraft can obtain the relative motion state information of the two aircrafts (including the local aircraft and the target aircraft) at the initial time from multiple data sources. The relative motion state information includes the relative height of the two aircrafts, the horizontal distance of the two aircrafts, the vertical speed of the local aircraft and the vertical speed of the target aircraft. The relative height of the two aircrafts refers to the difference between the barometric height of the target aircraft and the barometric height of the local aircraft, the horizontal distance of the two aircrafts refers to the projection distance of the target aircraft and the local aircraft in the horizontal plane, the vertical speed of the local aircraft refers to the speed of the local aircraft in the vertical direction, and the vertical speed of the target aircraft refers to the speed of the target aircraft in the vertical direction.

[0025] Further, considering the sensor measurement error, communication delay and target behavior uncertainty, these relative motion state information are modeled as random variables, rather than deterministic values, which together constitute the state vector at the initial time. The state vector is defined as follows: The state vector is defined as follows:

[0026] where x(t) is the state vector at time t, is the relative altitude of the two aircraft at time t, is the horizontal distance of the two aircraft at time t, is the vertical speed of the aircraft at time t, is the vertical speed of the target aircraft at time t.

[0027] Each component in the state vector x(t) is defined as a Gaussian random variable, which is completely determined by two parameters: mean and variance. The mean represents the best estimate of the state quantity, and at the initial time, the observed value can be directly taken as the mean. The variance represents the measure of uncertainty of the state quantity, and at the initial time, the variance is set to 0.

[0028] S102, a discrete-time system state equation for describing the evolution of the state vector over time is established.

[0029] In the specific implementation process, a discrete-time system state equation is used to describe the dynamic evolution process of the state vector. The equation combines the deterministic motion of the system, the control input and the random noise, and constitutes the basis for probability prediction.

[0030] The system state equation is represented as: x(t+Δt) = A(Δt)x(t) + G(Δt,x(t))w(t) + B(Δt)u(t,x(t)) where x(t) represents the state vector at time t; x(t+Δt) represents the state vector at time t+Δt; A(Δt) represents the state transition matrix; G(Δt,x(t)) represents the noise gain matrix at time t; w(t) represents the system noise vector at time t; B(Δt) is the control input matrix; u(t,x(t)) represents the control vector at time t, which takes values corresponding to different actions in the candidate action set.

[0031] The approximate value of the control vector is:

[0032] where r' is the rate of change of the horizontal distance of the two aircraft; is the vertical acceleration of the ownship at time t, when the pilot does not respond to the resolution advisory, is the vertical acceleration of the ownship at time t, when the pilot responds to the resolution advisory, is the 0.25g acceleration applied in the direction indicated by the resolution advisory, which can be expressed as:

[0033] wherein, is the gravitational acceleration; is the vertical speed of the ownship at time t; S RA is the resolution advisory, which is a mandatory operational advisory generated by the ACAS X system, explicitly instructing the pilot to maneuver the aircraft in a specific way to establish a safety separation.

[0034] if the resolution advisory is a descent advisory and h 1'(t) > -1500 ft / min, the ownship will perform a descent maneuver with a vertical acceleration of h 1'(t) < +1500 ft / min, the ownship will perform an ascent maneuver with a vertical acceleration of

[0035] In combination with the flight scenario, the system state equation can be further expressed as:

[0036] wherein, denotes the noise distribution of the vertical speeds of the two aircrafts; w 1(t) and w 2(t) are independent of each other. When the pilot responds to the resolution advisory (i.e., the ownship performs the current maneuver), w 1(t) is invalid, G(Δt, x(t)) is expressed as:

[0037] S103, for each action in the candidate action set, based on the state vector at the initial time and the system state equation, the future probability distribution of the state vector after the ownship performs the current action is predicted online.

[0038] In the specific implementation process, the candidate action set includes an ascent action, a descent action and a level flight action.

[0039] In one possible embodiment, for each action in the candidate action set, the following operations are performed: The delay probability of the aircraft executing the current action is calculated based on the probability distribution of the vertical speed of the aircraft in the state vector at the initial time; the statistical characteristics of the control vector are calculated based on the delay probability; the statistical characteristics include the expected value and the variance; the statistical characteristics of the control vector are substituted into the system state equation, and the future probability distribution of the state vector after the aircraft executes the current action is obtained by recursive solution.

[0040] The mean of x(t+Δt) is as follows:

[0041] The covariance of x(t+Δt) is as follows:

[0042] Wherein, E[ ] represents the mean; cov[ ] represents the covariance, R represents the noise covariance matrix, which describes the variance and correlation of the noise w(t). A represents the state transition matrix; A T represents the transpose matrix of A; G represents the noise gain matrix; G T represents the transpose matrix of G; B represents the control input matrix; B T represents the transpose matrix of B.

[0043] Based on the system state equation, the state mean μ(t) and variance σ(t) are approximately recursively calculated by the following formula:

[0044] Wherein, μ(t+Δt) is the mean of x(t+Δt); σ(t+Δt) is the covariance matrix of x(t+Δt); A(Δt) represents the state transition matrix; G(Δt,x(t)) represents the noise gain matrix at t; w(t) represents the system noise vector at t; B(Δt) is the control input matrix; u(t,μ(t)) is the control variable at t, which corresponds to different actions in the candidate action set; S is the approximate covariance matrix, which reflects the uncertainty of acceleration, and is expressed as:

[0045] is the variance of the vertical speed.

[0046] First, the delay probability of the pilot executing the release instruction is calculated, and then the mean and variance of the speed are calculated according to the delay probability. For the descent action, the mean μ d and the variance σ d are respectively:

[0047] Wherein, μ d(t) is the mean of the state vector after the local execution of the descent maneuver, σ d (t) is the variance of the state vector after the local execution of the descent maneuver; P d is the delay probability of the pilot executing the descent instruction (i.e. the local execution of the descent maneuver).

[0048] For the ascent maneuver, the mean μ c (t) and the variance σ c (t) are respectively:

[0049] wherein, μ c (t) is the mean of the state vector after the local execution of the ascent maneuver, σ c (t) is the variance of the state vector after the local execution of the ascent maneuver; P c is the delay probability of the pilot executing the ascent instruction (i.e. the local execution of the ascent maneuver).

[0050] For the level flight maneuver, since the local machine does not execute a vertical maneuver, it has no uncertainty in the vertical speed, so the variance of the vertical speed in the state vector: n (t) = 0.

[0051] In one possible embodiment, the step of calculating the delay probability of the local machine executing the current maneuver based on the probability distribution of the vertical speed of the local machine in the state vector at the initial time comprises: determining the probability density function of the vertical speed of the local machine based on the state vector at the initial time; if the current maneuver is a descent maneuver, integrating the probability density function of the vertical speed of the local machine over the interval [-1500, ∞) and taking the integral result as the delay probability of the local machine executing the descent maneuver; if the current maneuver is an ascent maneuver, integrating the probability density function of the vertical speed of the local machine over the interval (-∞, +1500] and taking the integral result as the delay probability of the local machine executing the ascent maneuver; if the current maneuver is a level flight maneuver, the delay probability of the local machine executing the level flight maneuver is 0.

[0052] In the specific implementation process, if the release instruction is a descent instruction, the current maneuver is a descent maneuver, and the delay probability P d is expressed as:

[0053] If the release instruction is an ascent instruction, the corresponding current maneuver is an ascent maneuver, and the delay probability P cis expressed as:

[0054] wherein, is the vertical velocity of the ownship is the probability density function of the future state vector. P d is the delay probability of the pilot executing the descent command (i.e. the ownship executing the descent action). P c is the delay probability of the pilot executing the climb command (i.e. the ownship executing the climb action).

[0055] If the release command is the level command, the corresponding current action is the level action. Since the level action does not require the pilot to perform a sharp vertical maneuver, its execution process is almost delay-free. Therefore, the delay probability of the ownship executing the level action is set to 0, indicating that the system defaults that the command can be responded and executed instantaneously.

[0056] S104, based on the future probability distribution of the state vector, calculate the conflict probability of the ownship after executing the current action and the target in the future.

[0057] In one possible embodiment, from the future probability distribution of the state vector, the mean and variance of the relative height of the two aircraft are extracted; based on the mean and variance of the relative height, the probability of the relative height being in the conflict interval is calculated using the Gaussian distribution density function; the probability of the relative height being in the conflict interval is taken as the conflict probability of the ownship after executing the current action and the target in the future.

[0058] In the specific implementation process, the future probability distribution of the state vector is the probability law of the state vector taking various possible values at future multiple time points. For each action in the candidate action set, the mean and variance of the relative height of the two aircraft are extracted from the future probability distribution of the state vector corresponding to the action, a conflict interval is defined, for example [-100ft, 100ft], and the probability of the relative height falling into the conflict interval is calculated using the properties of the Gaussian distribution, which is the conflict probability.

[0059] Traditional TCAS makes decision based on determined trajectory, the output is binary: conflict or no conflict, it cannot distinguish "just clear" (e.g. minimum separation 101 feet) and "safe clear" (e.g. minimum separation 1000 feet) which are essentially totally different in risk. In the embodiments of the present application, by calculating conflict probability, continuous and accurate measurement of risk is realized. For example, the conflict probability of action A is 0.1%, the conflict probability of action B is 5%, the system can know that action B is more dangerous than action A, and more know that the risk of action B is 50 times that of action A, the system can make more fine and optimized choices based on more abundant information, and actively avoid actions with risk of tens of times higher. And through the variance of relative height, the system can automatically perceive uncertainty, when the uncertainty is high, even if the mean value of relative height seems safe, the conflict probability may be high, thus triggering more conservative decision.

[0060] In a possible embodiment, the step of extracting the mean and variance of the relative height of the two aircraft from the future probability distribution of the state vector comprises: In the prediction time range of the future probability distribution of the state vector, the horizontal distance of the two aircraft at each time point is recorded; the time point corresponding to the minimum horizontal distance is determined as the closest time of the two aircraft; the mean and variance of the relative height of the two aircraft at the closest time are extracted from the future probability distribution of the state vector.

[0061] In the specific implementation process, according to the system state equation, the mean and variance of the state vector are updated every 1 second, and the horizontal distance r(t) of the two aircraft at each time point is recorded; the horizontal distance of all system states is traversed to find the time point corresponding to the minimum value, that is, the cpa time, which represents the time when the horizontal distance of the two aircraft is closest, and finally the mean and variance of the relative height of the two aircraft at the cpa time are extracted.

[0062] In the embodiments of the present application, the continuous probability distribution evaluation is reduced to the evaluation of a key time point (the time when the horizontal distance is closest), which greatly reduces the calculation load, so that the complex probability prediction algorithm can realize real-time running under the limited resources of the onboard computer. And the conflict risk reaches the highest when the two aircraft are closest, if the conflict probabilities of multiple time points are averaged or integrated, the high risk of the most dangerous time may be "diluted". Therefore, in the embodiments of the present application, the prediction is based on the closest time of the two aircraft, so that the finally generated collision avoidance decision instruction is more safe and conservative.

[0063] In a possible embodiment, the formula of the conflict probability is as follows:

[0064] Wherein, P conflictwhere h is the relative altitude of the two aircraft at the closest point in time, mu is the mean of the relative altitude of the two aircraft at the closest point in time, and sigma is the variance of the relative altitude of the two aircraft at the closest point in time. min where h is the relative altitude of the two aircraft at the closest point in time, mu is the mean of the relative altitude of the two aircraft at the closest point in time, and sigma is the variance of the relative altitude of the two aircraft at the closest point in time. max where h is the relative altitude of the two aircraft at the closest point in time, mu is the mean of the relative altitude of the two aircraft at the closest point in time, and sigma is the variance of the relative altitude of the two aircraft at the closest point in time.

[0065] S105, from the candidate action set, the action with the minimum conflict probability is determined as the optimal collision avoidance decision instruction.

[0066] In the specific implementation process, the conflict probabilities corresponding to all actions in the candidate action set are compared, and the action with the minimum conflict probability is selected as the optimal collision avoidance decision instruction.

[0067]

[0068] where argmin represents taking the minimum value, P conflict (a) represents the conflict probability of the host aircraft and the target aircraft in the future after the host aircraft performs the a-th action, A represents the candidate action set, represents the action with the minimum conflict probability.

[0069] In summary, the application also provides an ACAS X collision avoidance decision method based on probability prediction. Firstly, the relative altitude and vertical motion state information of the host aircraft and the intruder aircraft are modeled as Gaussian random variables. Then, the state evolution and uncertainty under each possible action are recursively calculated by combining the discrete-time recursive relationship of the flight state, to obtain the state distribution at the future time. Then, the probability of the relative altitude falling into the safe threshold interval is calculated by using the Gaussian distribution density function, and the conflict risk under different actions is quantified. Finally, the optimal action scheme is determined according to the conflict probability of each action, thereby supporting rapid real-time decision-making during flight. The method has the following beneficial effects: 1. The conflict probability is evaluated by probability integration, which can distinguish between high-risk and low-risk situations, thereby reducing false alarms and unnecessary warning times while ensuring safety.

[0070] 2. In the traditional method, small fluctuations in the state near the threshold value can cause the warning instruction to frequently switch between rising and falling. However, in the present application, the conflict probability changes smoothly, avoiding nonlinear jumps in decision-making, which can reduce the frequent switching of the warning instruction near the threshold value, effectively reducing the number of reversals and increases, and improving the continuity of pilot operation and system stability.

[0071] 3. The existing ACAS X offline cost table requires about one hundred megabytes of storage space, and occupies too much memory in actual embedded hardware. The present application generates collision avoidance decisions by analyzing and approximating, only needs to store limited parameters and state variables, greatly reducing the dependence on memory and hardware resources.

[0072] 4. Existing ACAS X offline cost table data is essentially discretely sampled. In practical use, interpolation calculations are required to cover the complete state space, which not only increases the computational load but may also introduce interpolation errors. This invention uses a continuous analytical probability approximation, which can directly calculate the conflict probability and cost value in any state without interpolation, resulting in more accurate results.

[0073] 5. The existing offline cost table of ACAS X requires sampling and numerical integration in a large-scale state-action space, with a huge sample size (about 20,000 samples) and long computation time. This application directly calculates the conflict probability and action cost in a two-dimensional vertical space, avoiding the high-dimensional integration of the POMDP method and greatly reducing the amount of computation.

[0074] Based on the same inventive concept, please refer to Figure 2 This application also provides an ACAS X collision avoidance decision system based on probability prediction, the system comprising: The state awareness module is used to set the relative motion state information of the two machines at the initial moment as the observation value at the initial moment when the local machine detects the target machine, set the variance at the initial moment to 0, and construct the state vector at the initial moment; the relative motion state information includes the relative height of the two machines, the horizontal distance between the two machines, the vertical speed of the local machine and the vertical speed of the target machine. The prediction module is used to establish the discrete-time system state equations that describe the evolution of the state vector over time; for each action in the candidate action set, based on the initial state vector and the system state equations, it predicts online the future probability distribution of the state vector after the machine executes the current action; the candidate action set includes ascent, descent, and level flight actions; The probability assessment module is used to calculate the probability of a future conflict between the local machine and the target machine after the current action is performed, based on the future probability distribution of the state vector. The decision generation module is used to determine the action with the lowest probability of conflict from the set of candidate actions as the optimal collision avoidance decision instruction.

[0075] Optionally, the system state equation can be expressed as: x(t+Δt) = A(Δt)x(t) + G(Δt,x(t))w(t) + B(Δt)u(t,x(t)) Where x(t) is the state vector at time t; x(t+Δt) is the state vector at time t+Δt; A(Δt) represents the state transition matrix; G(Δt,x(t)) represents the noise gain matrix at time t; w(t) represents the system noise vector at time t; B(Δt) is the control input matrix; and u(t,x(t)) represents the control vector at time t, whose values ​​correspond to different actions in the candidate action set.

[0076] Optionally, the approximation of the control vector is:

[0077]

[0078] wherein, is the vertical acceleration of the ego aircraft at time t, and r' is the rate of change of the horizontal distance between the two aircrafts; is the acceleration due to gravity; S RA represents a release instruction; is the vertical speed of the ego aircraft at time t.

[0079] Optionally, the prediction module is specifically configured to: for each action in the candidate action set, perform the following operations: based on the probability distribution of the vertical speed of the ego aircraft in the state vector at the initial time, calculate the delay probability of the ego aircraft performing the current action; based on the delay probability, calculate the statistical characteristics of the control vector; the statistical characteristics include the expected value and the variance; substitute the statistical characteristics of the control vector into the system state equation, and obtain the future probability distribution of the state vector of the ego aircraft after performing the current action by recursive solution.

[0080] Optionally, the prediction module is specifically configured to: based on the state vector at the initial time, determine the probability density function of the vertical speed of the ego aircraft; if the current action is a descending action, integrate the probability density function of the vertical speed of the ego aircraft on the interval [-1500, ∞), and take the integral result as the delay probability of the ego aircraft performing the descending action; if the current action is an ascending action, integrate the probability density function of the vertical speed of the ego aircraft on the interval (-∞, +1500], and take the integral result as the delay probability of the ego aircraft performing the ascending action; if the current action is a level flight action, the delay probability of the ego aircraft performing the level flight action is 0.

[0081] Optionally, the probability evaluation module is specifically configured to: from the future probability distribution of the state vector, extract the mean and variance of the relative height of the two aircrafts; based on the mean and variance of the relative height, calculate the probability that the relative height is in the conflict interval by using the Gaussian distribution density function; take the probability that the relative height is in the conflict interval as the future conflict probability of the ego aircraft performing the current action with the target aircraft.

[0082] Optionally, the probability evaluation module is specifically configured to: record the horizontal distance between the two aircrafts at each time point within the prediction time range of the future probability distribution of the state vector; determine the time point corresponding to the minimum horizontal distance as the closest time point of the two aircrafts; extract the mean and variance of the relative height of the two aircrafts at the closest time point from the future probability distribution of the state vector.

[0083] Optionally, the conflict probability is calculated according to the following formula:

[0084] wherein, P conflict is the conflict probability; h is the relative height of the two aircrafts at the closest time point; μ is the mean of the relative height of the two aircrafts at the closest time point; σ is the variance of the relative height of the two aircrafts at the closest time point; h min is the minimum value of the conflict interval; h max is the maximum value of the conflict interval.

[0085] It should be noted that the modules in the conflict probability prediction collision avoidance decision system in the embodiment are one-to-one corresponding to the steps in the ACAS X collision avoidance decision method based on probability prediction in the foregoing embodiment, and thus the specific implementation of the embodiment can refer to the implementation of the ACAS X collision avoidance decision method based on probability prediction in the foregoing embodiment, which will not be described herein again.

[0086] Based on the same inventive concept, the present application further provides a computer device, which comprises a processor, a memory, and a computer program stored in the memory, and the computer program is run by the processor to implement the ACAS X collision avoidance decision method based on probability prediction.

[0087] Based on the same inventive concept, the present application further provides a computer storage medium, which stores a computer program, and the computer program is run by the processor to implement the ACAS X collision avoidance decision method based on probability prediction.

[0088] In some embodiments, the computer readable storage medium can be a memory such as FRAM, ROM, PROM, EPROM, EEPROM, flash memory, magnetic surface memory, optical disc, or CD-ROM; or various devices comprising one or any combination of the above memories. The computer can be various computing devices including smart terminals and servers.

[0089] In some embodiments, the executable instructions can take the form of programs, software, software modules, scripts, or code, written in any form of programming language, including compiled or interpreted languages, or declarative or procedural languages; and they can be deployed in any form, including as a stand-alone program or as a module, component, subroutine, or other unit suitable for use in a computing environment.

[0090] By way of example, the executable instructions can, but need not, correspond to a file in a file system, can be stored in a portion of a file that holds other programs or data, e.g., one or more scripts stored in a markup language document, in a single file dedicated to the program in question, or in multiple coordinated files, e.g., files that store one or more modules, sub programs, or portions of code.

[0091] By way of example, the executable instructions can be stored on the one or more computer- readable storage media of a server and adapted to be transferred using, for example, one or more application

[0092] It is to be appreciated that the term "include," "comprise," or variations thereon, as used herein, is intended to cover the presence of one or more elements, steps, or components, but not the exclusion of any other elements, steps, or components. Nothing in the present disclosure is intended to be construed as limiting the scope of the disclosure.

[0093] The above-mentioned sequence numbers of the embodiments of the present application are only for description, and do not represent advantages or disadvantages of the embodiments.

[0094] The above detailed description merely describes exemplary embodiments of the application, and is not intended to limit the scope of the application. The description is not intended to limit the scope of the application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A probabilistic prediction-based ACAS X collision avoidance decision-making method, characterized in that, include: When the local machine detects the target machine, it sets the relative motion state information of the two machines at the initial moment as the observation value at the initial moment, sets the variance at the initial moment to 0, and constructs the state vector at the initial moment; the relative motion state information includes the relative height of the two machines, the horizontal distance between the two machines, the vertical speed of the local machine and the vertical speed of the target machine; Establish a discrete-time system state equation to describe the evolution of the state vector over time; For each action in the candidate action set, based on the state vector at the initial moment and the system state equation, the future probability distribution of the state vector after the local machine executes the current action is predicted online; the candidate action set includes ascent action, descent action, and level flight action. Based on the future probability distribution of the state vector, calculate the probability of a future conflict between the local machine and the target machine after the local machine performs the current action; From the set of candidate actions, the action with the lowest probability of conflict is determined as the optimal collision avoidance decision instruction.

2. The ACAS X collision avoidance decision method based on probability prediction according to claim 1, characterized in that, The system state equation is expressed as: x(t+Δt) = A(Δt)x(t) + G(Δt,x(t))w(t) + B(Δt)u(t,x(t)) Where x(t) is the state vector at time t; x(t+Δt) is the state vector at time t+Δt; A(Δt) represents the state transition matrix; G(Δt,x(t)) represents the noise gain matrix at time t; w(t) represents the system noise vector at time t; B(Δt) is the control input matrix; and u(t,x(t)) represents the control vector at time t, whose values ​​correspond to different actions in the candidate action set.

3. The ACAS X collision avoidance decision method based on probability prediction according to claim 2, characterized in that, The approximate value of the control vector is: ； ； in, Let r' be the vertical acceleration of the machine at time t, and r' be the rate of change of the horizontal distance between the two machines. S is the acceleration due to gravity; RA Indicates a release instruction; Let be the vertical speed of the machine at time t.

4. The ACAS X collision avoidance decision method based on probability prediction according to claim 1, characterized in that, For each action in the candidate action set, based on the state vector at the initial moment and the system state equation, the online prediction of the future probability distribution of the state vector after the local machine executes the current action includes: For each action in the candidate action set, perform the following operation: Based on the probability distribution of the vertical velocity of the machine in the state vector at the initial moment, calculate the delay probability of the machine executing the current action; Based on the delay probability, the statistical properties of the control vector are calculated; the statistical properties include the expected value and the variance. Substituting the statistical characteristics of the control vector into the system state equation, and solving it recursively, the future probability distribution of the state vector after the machine performs the current action is obtained.

5. The ACAS X collision avoidance decision method based on probability prediction according to claim 4, characterized in that, The calculation of the delay probability of the machine executing the current action based on the probability distribution of the vertical velocity of the machine in the state vector at the initial moment includes: Based on the state vector at the initial moment, determine the probability density function of the vertical speed of the machine; If the current action is a descent action, the probability density function of the vertical speed of the machine is integrated over the interval [-1500, ∞), and the integration result is used as the delay probability of the machine performing the descent action. If the current action is an ascending action, the probability density function of the vertical speed of the machine is integrated over the interval (-∞, +1500], and the integration result is used as the delay probability of the machine performing the ascending action. If the current action is a level flight action, the delay probability of the machine performing the level flight action is 0.

6. The ACAS X collision avoidance decision method based on probability prediction according to claim 1, characterized in that, The calculation of the probability of a future collision between the local machine and the target machine after the current action is performed, based on the future probability distribution of the state vector, includes: Extract the mean and variance of the relative heights of the two machines from the future probability distribution of the state vector; Based on the mean and variance of the relative height, the probability that the relative height is in the conflict zone is calculated using the Gaussian distribution density function; The probability that the relative altitude is in the conflict zone is taken as the probability of a future conflict between the local machine and the target machine after the local machine performs the current action.

7. The ACAS X collision avoidance decision method based on probability prediction according to claim 6, characterized in that, Extracting the mean and variance of the relative heights of the two machines from the future probability distribution of the state vector includes: Within the prediction time range of the future probability distribution of the state vector, record the horizontal distance between the two machines at each time point; The time point corresponding to the minimum horizontal distance is determined as the closest moment between the two machines; From the future probability distribution of the state vector, extract the mean and variance of the relative altitude of the two machines at the closest time.

8. The ACAS X collision avoidance decision method based on probability prediction according to claim 7, characterized in that, The formula for calculating the conflict probability is as follows: ； Among them, P conflict Let h be the collision probability; h be the relative altitude of the two aircraft at the closest moment; μ be the mean of the relative altitude of the two aircraft at the closest moment; σ be the variance of the relative altitude of the two aircraft at the closest moment; h min h is the minimum value of the conflict interval; max This represents the maximum value of the conflict interval.

9. An ACAS X collision avoidance decision system based on probability prediction, comprising: The state awareness module is used to set the relative motion state information of the two machines at the initial moment as the observation value at the initial moment when the local machine detects the target machine, set the variance at the initial moment to 0, and construct the state vector at the initial moment; the relative motion state information includes the relative height of the two machines, the horizontal distance between the two machines, the vertical speed of the local machine and the vertical speed of the target machine. The prediction module is used to establish a discrete-time system state equation describing the evolution of the state vector over time; for each action in the candidate action set, based on the state vector at the initial moment and the system state equation, it predicts online the future probability distribution of the state vector after the local machine executes the current action; the candidate action set includes ascent action, descent action, and level flight action; The probability assessment module is used to calculate the probability of a future conflict between the local machine and the target machine after the current action is performed, based on the future probability distribution of the state vector. The decision generation module is used to determine the action with the lowest conflict probability from the set of candidate actions as the optimal collision avoidance decision instruction.

10. An electronic device, characterized in that, The aircraft includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement a probability prediction-based ACAS X collision avoidance decision method as described in any one of claims 1-8.

Images