Multi-party multi-objective evolution method for unmanned aerial vehicle path optimization

Opt-MPNDS3 algorithm combines non-dominant sorting and crowded distance to optimize the UAV path, and solves the multi-party and multi-objective optimization problem of efficiency and safety decision-making models in UAV path planning, achieving high-quality and efficient path planning.

CN118966493APending Publication Date: 2024-11-15XIAN UNIV OF TECH
View PDF 0 Cites 1 Cited by

Patent Information

Application Number
CN202411011927.5
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-26
Publication Date
2024-11-15

AI Technical Summary

Technical Problem

The existing technology is difficult to effectively solve the multi-party and multi-objective optimization problem of efficiency and safety decision-making models in drone path planning. Traditional methods cannot meet the needs of both parties in complex environments, resulting in the solution results that do not meet the high-quality and efficient path planning requirements.

Method used

Multi-party and multi-objective evolution method is adopted to initialize populations, activate, clonize, cross-mutation and offspring selection through the OptMPNDS3 algorithm, and combine non-dominant sorting and crowded distance to optimize the drone path to ensure the efficiency and safety of the path.

Benefits of technology

It improves the diversity and convergence of drone path planning, significantly improves the quality and efficiency of solutions, can effectively avoid high-risk areas, ensure path continuity, and provide high-quality and efficient flight paths.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure BDA0004964630240000031
    Figure BDA0004964630240000031
  • Figure BDA0004964630240000032
    Figure BDA0004964630240000032
  • Figure BDA0004964630240000041
    Figure BDA0004964630240000041
Patent Text Reader

Abstract

The invention discloses a multi-party multi-objective evolution method for unmanned aerial vehicle path optimization, and the method is specifically implemented according to the following steps: 1, initializing parameters, obtaining population initialization P0 and an objective function f0, and setting the number t of iterations as 0; step 2, performing a multi-party multi-target selection evaluation algorithm on the population P0 and the target function f0 to perform offspring selection; step 3, for the latest generation of population after t generations, activating the activation scale nA to obtain a plurality of non-dominated sorting layers AF of activated individuals and activated individuals A; 4, cloning according to the plurality of non-dominated sorting layers AF of the activated individuals, the activated individuals A and the population cloning scale nC to obtain cloned individuals C; 5, obtaining an individual O according to cross mutation of the cloned individual C and the activated individual A; step 6, evaluating the individual O to obtain a target function fO, and obtaining an intersection of the offspring Pt and the individual O to obtain an offspring Pt + 1; 7, offspring selection is carried out. According to the invention, a high-quality and high-efficiency flight path of the unmanned aerial vehicle can be provided.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention belongs to the technical field of evolutionary computing methods, and specifically relates to a multi-party and multi-objective evolutionary method for unmanned aerial vehicle path optimization. Background Art

[0002] Drones have been widely used in our daily life, for aerial image acquisition, logistics delivery services, infrastructure inspection, environmental monitoring and many other operations. Drone path planning needs to consider two different decision models. One is the efficiency decision model. For the flight efficiency of drones, at least the flight time and flight energy consumption of drones need to be considered. To minimize the flight time of drones, multiple goals such as the length and height of the drone route and the terrain angle can be optimized. On the other hand, it is a safety decision model, which takes into account the noise pollution, fatal risks and urban property safety of drones to minimize multiple goals, so as to achieve high safety. Such problems involve complex environmental changes and uncertain factors in our daily lives.

[0003] Based on the above problems, it is not feasible to obtain the optimal solution for UAV path planning by simply using a single decision maker model and using a multi-objective evolutionary optimization algorithm to approach the approximate frontier. Therefore, how to deal with these multiple decision makers, how to deal with environmental changes, and choose a path solution with better quality and shorter flight time as much as possible is a difficult problem. Because both parties are considered and the different needs of multiple decision makers need to be considered, ordinary multi-objective optimization algorithms cannot effectively distinguish this situation in such a complex environment, which ultimately means that the solution result cannot meet the needs of the biparty multi-objective UAV path planning strategy. If all targets are regarded as a single decision maker, the optimal solution set with two or more decision makers is only a very small part of the optimal solution set obtained. By using a single decision maker optimization algorithm to solve the UAV path planning strategy problem, a subset of solutions can be selected from the solution set where the safety and efficiency decision makers make joint decisions. This method cannot incorporate selection pressure into the algorithm's search process. Summary of the invention

[0004] The purpose of the present invention is to provide a multi-party and multi-objective evolutionary method for UAV path optimization, which can provide high-quality and efficient UAV flight paths.

[0005] The technical solution adopted by the present invention is a multi-party multi-objective evolutionary method for UAV path optimization, which is specifically implemented according to the following steps:

[0006] Step 1: Initialize parameters to get population initialization P 0 , and the objective function f 0 , and the number of iterations t is set to 0;

[0007] Step 2: For population P 0 And the objective function f 0 Conduct multi-party and multi-objective selection evaluation algorithm for offspring selection;

[0008] Step 3, performing activation operation on the latest generation population after t generations and the activation size nA to obtain multiple non-dominated sorting layers AF of activated individuals and activated individuals A;

[0009] Step 4: clone the cloned individual C according to multiple non-dominated sorting layers AF of the activated individual, the activated individual A and the population clone size nC;

[0010] Step 5: Obtain individual O through cross mutation of cloned individual C and activated individual A;

[0011] Step 6: Evaluate individual O to obtain the objective function f O , the offspring P t Take the intersection with individual O to get the offspring P t+1 ;

[0012] Step 7: Perform offspring selection to select the next generation with better performance.

[0013] The present invention is also characterized in that:

[0014] In step 1, the initialization parameters include the population clone size nC, the activation size nA, the upper limit ub of the decision variable and the lower limit lb of the decision variable.

[0015] In step 1, the OptMPNDS3 algorithm is used to sort the individuals and activate the individuals in the front part of the sorting results. All individuals whose sorting level is less than the activation scale nA will be activated:

[0016] A=P sorted (1: nA,:) (1)

[0017] Where P sorted is the set of individuals of population P sorted by the OptMPNDS3 algorithm, and nA is the preset number of activated individuals.

[0018] Step 2 is implemented as follows:

[0019] Step 2.1, the individual allocation of the current clone is determined by multiple non-dominated sorting layers and crowding information;

[0020] Step 2.2: In the calculation of the number of clones, the number of layers is used as the indicator of resource positioning. The convergence indicator calculation formula is as follows:

[0021]

[0022] Where p i As the convergence indicator of the i-th activated individual, AF i Represents multiple non-dominated sorting layers of the i-th activated individual, where the individual with a larger convergence index has better convergence. Based on the above convergence index, the calculation formula for the number of clones of each individual is obtained:

[0023]

[0024] Among them, Nums(i) is the number of clones of the i-th activated individual, cd(·) is the function formula for calculating the crowding distance, and cd(a i ) is the crowding distance calculation function for calculating the i-th activated individual.

[0025] Step 3 is implemented as follows:

[0026] Step 3.1: For a given path, the i-th discrete trajectory point can be expressed as The i-th trajectory segment can be expressed as

[0027] Step 3.2: The calculation formula of the objective function of the drone path length is as follows:

[0028]

[0029] Among them, g i is the i-th trajectory segment;

[0030] Step 3.3: The flight energy consumption of the drone depends on the flight state, hovering power consumption and the consumption of overcoming gravity. The energy consumption calculation formula is as follows:

[0031]

[0032] Where, parameter W is the battery weight of the UAV, ρ i is the density of the air fluid, G is the gravity, Δ is the increase in height, n is the number of rotors, ||g i || is the length of the aircraft's flight trajectory, V is the flight speed of the drone, is the area of ​​the rotating blade of the aircraft, and the function of the air fluid density is as follows:

[0033]

[0034] Step 3.4: If fewer aircraft altitude changes during the UAV mission means fewer climb and descent phases, the aircraft power system life will be extended. The path altitude changes are as follows:

[0035]

[0036] Step 5 is implemented according to the following steps:

[0037] Step 5.1: In the retained population operation, first use the OptMPNDS3 algorithm to sort and stratify the individuals;

[0038] Step 5.2: According to the number of sorted layers, add individuals to the next generation population layer by layer until the population size exceeds the number of individuals that can be retained;

[0039] Step 5.3: Use the crowding distance to select some individuals in this layer to join the next generation population, discard individuals with too small crowding distance, and select individuals with better crowding distance.

[0040] In step 6, the updated offspring P of the tth generation t Take the intersection with the individual O produced after mutation.

[0041] The beneficial effects of the present invention are:

[0042] The multi-party multi-objective evolutionary method for UAV path optimization of the present invention provides an innovative two-party multi-objective perspective for solving the UAV aircraft path planning problem. It introduces a two-way multi-objective optimization model into the UAV aircraft path planning for the first time. The OptMPNDS3 algorithm has significant advantages in dealing with two-party multi-objective UAV path planning problems, especially in avoiding high-risk areas, ensuring path continuity and improving optimization performance. Compared with the traditional NSGA-II algorithm, the present invention not only improves the diversity and convergence of solutions, but also makes significant progress in the quality and efficiency of solutions. DETAILED DESCRIPTION

[0043] The present invention is described in detail with reference to the following specific embodiments.

[0044] Example 1

[0045] The multi-party multi-objective evolutionary method for drone path optimization of the present invention is specifically implemented according to the following steps:

[0046] Step 1: Initialize parameters, including population clone size nC, activation size nA, upper limit ub of decision variables and lower limit lb of decision variables; initialize population clone size nC and upper and lower limits ub and lb of decision variables to obtain population initialization P 0 , and the objective function f0 , and the number of iterations t is set to 0;

[0047] Step 2: For the population P in the previous step 0 And the objective function f 0 The multi-party multi-objective selection evaluation algorithm OptMPNDS3 algorithm is used to select offspring;

[0048] Step 3, performing activation operation on the latest generation population after t generations and the activation size nA to obtain multiple non-dominated sorting layers AF of activated individuals and activated individuals A;

[0049] Step 4: Perform a cloning operation according to multiple non-dominated sorting layers AF of the activated individuals, the activated individuals A and the population clone size nC to obtain the cloned individuals C;

[0050] Step 5, the individual O generated after the operation is obtained according to the cross mutation of the cloned individual C and the activated individual A;

[0051] Step 6: Evaluate the individual O obtained in the previous step to obtain the objective function f O , the updated offspring P of the tth generation t The intersection of the individual O produced after the mutation is obtained to obtain the offspring P t+1 ;

[0052] Step 7: Use the evaluation algorithm 0ptMPNDS3 algorithm to select offspring and select the next generation with better performance.

[0053] Example 2

[0054] Multi-party and multi-objective evolutionary method for UAV path optimization, in which the OptMPNDS3 algorithm is used in step 1 to sort the individuals and activate the first part of the sorted results. All individuals with a sorting level less than the activation scale nA will be activated:

[0055] A=P sorted (1: nA,:) (1)

[0056] Where P sorted is the set of individuals of population P sorted by PptMPNDS3 algorithm, and nA is the preset number of activated individuals.

[0057] Step 2 is implemented as follows:

[0058] Step 2.1, the individual allocation of the current clone is determined by multiple non-dominated sorting layers and crowding information;

[0059] Step 2.2: In the calculation of the number of clones, the number of layers is used as the indicator of resource positioning. The convergence indicator calculation formula is as follows:

[0060]

[0061] Where p i As the convergence indicator of the i-th activated individual, AF i Represents multiple non-dominated sorting layers of the i-th activated individual, where the individual with a larger convergence index has better convergence. Based on the above convergence index, the calculation formula for the number of clones of each individual is obtained:

[0062]

[0063] Among them, Nums (i) is the number of clones of the activated individual, where cd(·) is the function formula for calculating the crowding distance, cf(a i ) is the crowding distance calculation function for calculating the i-th activated individual.

[0064] Step 3 is implemented as follows:

[0065] Step 3.1: From the perspective of efficiency decision makers, maximizing the flight efficiency of the drone path planning problem is the core requirement of this patent. This goal is mainly solved by minimizing the length of the drone path, minimizing the change in flight altitude, minimizing the flight energy consumption, and minimizing the total distance between the mission hovering points. For a given path, the i-th discrete trajectory point can be expressed as The i-th trajectory segment can be expressed as

[0066] Step 3.2, Path length When the UAV performs a mission, the path length will directly affect the running time of the UAV. User decision makers who are concerned about efficiency often want to choose the shortest path to reduce the running time. The calculation formula of the UAV path length objective function is as follows:

[0067]

[0068] Among them, g i is the i-th trajectory segment;

[0069] Step 3.3, flight energy consumption needs to be considered that the flight energy consumption of the drone mainly depends on the following factors: flight state, hovering power consumption and consumption of overcoming gravity. The energy consumption calculation formula is as follows:

[0070]

[0071] In the above formula, parameter W is the battery weight of the UAV aircraft, ρ i is the density of the air fluid, G is the gravity, Δ is the increase in height, n is the number of rotors, ||g i|| is the length of the aircraft's flight trajectory, V is the flight speed of the drone, is the area of ​​the rotating blade of the aircraft, and the function of the air fluid density is as follows:

[0072]

[0073] Step 3.4: If fewer aircraft altitude changes during the UAV mission means fewer climb and descent phases, the aircraft power system life will be extended. The path altitude changes are as follows:

[0074]

[0075] Example 3

[0076] The multi-party multi-objective evolutionary method for the optimization of the UAV path of the present invention, wherein, in the cloning operation part of step 4, the UAV brings uncontrollable third-party risks to people, vehicles and buildings in urban areas, such as: risk of death, property loss, noise pollution, etc. For safety decision makers, the core requirement is to ensure that the risk of UAV aircraft operation to urban residents and property is as small as possible, and the main goal of safety decision-making is to minimize the risk to pedestrians, vehicles, property and noise pollution;

[0077] Reduce the risk of endangering the safety of third parties, including pedestrian fatalities and vehicle fatalities. The fatality risk cost associated with pedestrian fatalities depends on the reliability of the drone itself, including the reliability of its hardware and software, as well as the size of the drone's impact zone.

[0078] Property risk: When a drone is operating at high altitude and collides with a high-rise building in a city, especially in a city with a high building density, the flight altitude is an important factor affecting property risk, so the property risk index needs to be considered;

[0079] Noise pollution: When drones are operated, if they significantly affect the lives and production of urban residents, complaints may be filed. However, as the altitude of the drones increases, the impact on ground personnel will gradually decrease, and noise beyond a certain distance will be considered not to cause noise pollution to third parties.

[0080] Step 5 is implemented according to the following steps:

[0081] Step 5.1: In the retained population operation, first use the OptMPNDS3 algorithm to sort and stratify the individuals;

[0082] Step 5.2: According to the number of sorted layers, add individuals to the next generation population layer by layer until the population size exceeds the number of individuals that can be retained;

[0083] Step 5.3: Use the crowding distance to select some individuals in this layer to join the next generation population, discard individuals with too small crowding distance, and select individuals with better crowding distance.

[0084] In step 6, the updated offspring P of the tth generation t Take the intersection with the individual O produced after mutation.

[0085] The present invention is used to solve the UAV aircraft path planning problem. It introduces a bidirectional multi-objective optimization model into the UAV aircraft path planning for the first time. The OptMPNDS3 algorithm has significant advantages in dealing with the bidirectional multi-objective UAV path planning problem, especially in avoiding high-risk areas, ensuring path continuity and improving optimization performance. Compared with the traditional NSGA-II algorithm, the bidirectional multi-objective algorithm proposed in the present invention not only improves the diversity and convergence of the solution, but also makes significant progress in the quality and efficiency of the solution.

[0086] By improving the generation of offspring, combining multi-party multi-objective algorithms, and using cloning strategies to increase the diverse nature of selection, the two decision-makers, efficiency decision-makers and safety decision-makers, consider their respective goals and apply multi-party multi-objective algorithms to the UAV path planning strategy. Different decision-makers are allowed to reach a consensus based on their evaluation strategies to obtain high-quality and efficient UAV flight paths. The multi-party multi-objective optimization model is directly applied to UAV path planning, thereby improving the stability and efficiency of path planning.

Claims

1. A multi-party and multi-objective evolutionary method for UAV path optimization, characterized by: Follow the steps below to implement it: Step 1: Initialize parameters to get population initialization P 0 , and the objective function f 0 , and the number of iterations t is set to 0; Step 2: For population P 0 And the objective function f 0 Conduct multi-party and multi-objective selection evaluation algorithm for offspring selection; Step 3, performing activation operation on the latest generation population after t generations and the activation size nA to obtain multiple non-dominated sorting layers AF of activated individuals and activated individuals A; Step 4: clone the cloned individual C according to multiple non-dominated sorting layers AF of the activated individual, the activated individual A and the population clone size nC; Step 5: Obtain individual O through cross mutation of cloned individual C and activated individual A; Step 6: Evaluate individual O to obtain the objective function f O , the offspring P t Take the intersection with individual O to get the offspring P t+1 ; Step 7: Perform offspring selection to select the next generation with better performance.

2. The multi-party multi-objective evolutionary method for UAV path optimization according to claim 1 is characterized in that: In step 1, the initialization parameters include the population clone size nC, the activation size nA, and the upper limit ub and lower limit lb of the decision variable.

3. The multi-party multi-objective evolutionary method for UAV path optimization according to claim 1 is characterized in that: In step 1, the OptMPNDS3 algorithm is used to sort the individuals and activate the first part of the individuals in the sorting result. All individuals whose sorting level is less than the activation scale nA will be activated: A=P sorted (1:nA,:) (1) Where P sorted is the set of individuals of the population P sorted by the OptMPNDS3 algorithm, and nA is the preset number of activated individuals.

4. The multi-party multi-objective evolutionary method for UAV path optimization according to claim 1 is characterized in that: The step 2 is specifically implemented according to the following steps: Step 2.1, the individual allocation of the current clone is determined by multiple non-dominated sorting layers and crowding information; Step 2.2: In the calculation of the number of clones, the number of layers is used as the indicator of resource positioning. The convergence indicator calculation formula is as follows: Where p i As the convergence indicator of the i-th activated individual, AF i Represents multiple non-dominated sorting layers of the i-th activated individual, where the individual with a larger convergence index has better convergence. Based on the above convergence index, the calculation formula for the number of clones of each individual is obtained: Among them, Nums (i) is the number of clones of the activated individual, where cd(·) is the function formula for calculating the crowding distance, cd(a i ) is the crowding distance calculation function for calculating the i-th activated individual.

5. The multi-party multi-objective evolutionary method for UAV path optimization according to claim 1 is characterized in that: The step 3 is specifically implemented according to the following steps: Step 3.1: For a given path, the i-th discrete trajectory point can be expressed as The i-th trajectory segment can be expressed as Step 3.2: The calculation formula of the objective function of the drone path length is as follows: Among them, g i is the i-th trajectory segment; Step 3.3: The flight energy consumption of the drone depends on the flight state, hovering power consumption and the consumption of overcoming gravity. The energy consumption calculation formula is as follows: Where, parameter W is the battery weight of the UAV, ρ i is the density of the air fluid, G is the gravity, Δ is the increase in height, n is the number of rotors, ||g i || is the length of the aircraft's flight trajectory, V is the flight speed of the drone, is the area of ​​the rotating blade of the aircraft, and the function of the air fluid density is as follows: Step 3.4: If fewer aircraft altitude changes during the UAV mission means fewer climb and descent phases, the aircraft power system life will be extended. The path altitude changes are as follows:

6. The multi-party multi-objective evolutionary method for UAV path optimization according to claim 1 is characterized in that: The step 5 is specifically implemented according to the following steps: Step 5.1: In the retained population operation, first use the OptMPNDS3 algorithm to sort and stratify the individuals; Step 5.2: According to the number of sorted layers, add individuals to the next generation population layer by layer until the population size exceeds the number of individuals that can be retained; Step 5.3: Use crowding distance to select some individuals in this layer to join the next generation population, discard individuals with too small crowding distance, and select individuals with better crowding distance.

7. The multi-party multi-objective evolutionary method for UAV path optimization according to claim 1 is characterized in that: In step 6, the updated offspring P of the tth generation t Take the intersection with the individual O produced after mutation.

Citation Information

Cited By

  • Multi-party multi-target unmanned aerial vehicle path planning method based on multi-department balance strategy

    CN119962793A