Flying vehicle path planning method capable of dynamically adjusting weight

Abstract

The invention provides a flying vehicle path planning method based on dynamic weight adjustment, and belongs to the technical field of aircraft planning. Comprising the following steps: acquiring planning data under different fixed weight values; processing planning data according to actual vehicle energy storage and task limiting time, and fitting weight mathematical representations under different task requirements by using a radial basis function neural network RBFNN (Radial Basis Function Neural Network); the method comprises the following steps: designing a weight A * algorithm of dynamic weight adjustment, dynamically adjusting weight values of different costs, searching a path node with the minimum comprehensive cost, reasonably switching different motion modes and planning a short-time energy-saving task path on the basis of weight mathematical representation and aiming at task requirements changing in real time. According to the method, reasonable switching of different motion modes is realized by searching the path node with the minimum comprehensive cost, and a short-time energy-saving task path is planned. Weight mathematical representations under different task requirements are constructed through the RBFNN, and weight values of different costs are dynamically adjusted to meet the task requirements changing in real time.

Description

TECHNICAL FIELD

[0001] The application provides a flight vehicle path planning method with dynamic weight adjustment, and belongs to the technical field of aircraft planning. BACKGROUND

[0002] Autonomous task execution of a flight vehicle requires technologies such as perception, planning and control. Path planning can provide the flight vehicle with an optimal task path without collision, helping the flight vehicle to efficiently complete various tasks. The flight vehicle has two movement modalities, ground driving and air flight. The flight vehicle can easily cross ground obstacles by air flight, reduce vehicle running distance and save task time. However, frequent takeoff will consume a large amount of vehicle energy, so in the area without ground obstacles, the vehicle mainly runs in the ground driving modality. In the path planning process, how to reasonably switch different movement modalities to save vehicle task time and not increase excessive energy consumption and obtain a short-time energy-saving task path becomes a technical difficulty.

[0003] A* algorithm is a common path planning algorithm. The improved A* algorithm can consider multiple costs such as vehicle movement energy consumption, task time and path safety, reasonably switch different movement modalities and plan a task route with minimum comprehensive cost. The above different costs have different weights, and the final path planning effect also differs when the weights are set differently. For example, when the vehicle has sufficient energy storage, the movement energy consumption cost is not considered, that is, the weight value of the energy consumption cost is small, and the final planned path will have frequent takeoff and obstacle crossing. At this time, the vehicle movement energy consumption increases, but the running distance is significantly shortened and the task time is saved. Through reasonable weight setting, a task path with comprehensive planning effect of low energy consumption, short time, high safety and the like can be obtained.

[0004] The air flight modality of the flight vehicle can effectively save the vehicle task time, and the ground driving modality can maintain a low energy consumption. The path planning of the flight vehicle needs to reasonably switch different movement modalities to save the vehicle task time and not increase excessive movement energy consumption and obtain a short-time energy-saving task path. At present, the related path planning technology is not perfect, and the above short-time energy-saving comprehensive planning effect is difficult to meet.

[0005] Part of the study carried out to consider the task time and motion energy consumption and other comprehensive optimization of multiple cost flight vehicle path planning technology exploration. The different costs usually have different weights, in the existing research, the weight value setting of different costs mostly adopts the fixed value method, that is, the fixed weight value is set in advance before path planning, and it does not change in the planning process. However, in the process of path planning, the task demand will change in real time, for example, as the task is executed, the vehicle may face energy storage shortage as the vehicle energy consumption is large, at this time, the original energy consumption cost weight value no longer meets the current task demand, the weight should be appropriately increased to reduce the subsequent energy consumption. The fixed value method in the current research obviously cannot cope with the real-time changes of the above task demand. SUMMARY

[0006] The present application proposes a flight vehicle path planning method with dynamic weight adjustment. This method mainly considers the comprehensive optimization of two costs of task time and motion energy consumption, but in subsequent applications, it is not limited to the above two costs.

[0007] The specific technical solutions are as follows:

[0008] A flight vehicle path planning method with dynamic weight adjustment, comprising the following steps:

[0009] Step 1: Collect planning data under different fixed weight values; the data collection process does not consider the task demand limit of the flight vehicle, that is, it is assumed that the vehicle energy storage is sufficient and the task limited time is sufficient.

[0010] Step 2: Construct weight mathematical representation; according to the actual vehicle energy storage and task limited time, process the above planning data, and further fit the weight mathematical representation under different task demands by using radial basis function neural network RBFNN.

[0011] Step 3: Task path planning; design a weighted A* algorithm with dynamic weight adjustment, based on the above weight mathematical representation, dynamically adjust the weight values of different costs according to the real-time changing task demand, search for the path node with the minimum comprehensive cost, including the mode switching path point, reasonably switch different motion modes, and plan a short-time energy-saving task path.

[0012] Further, the specific method of Step 1 is as follows:

[0013] Grid map, input start point and end point, set multiple different fixed weight values for task time cost and motion energy consumption cost, plan path and collect planning data. The planning data includes the fixed weight value of each cost, the task time after planning, and the vehicle energy consumption after planning.

[0014] Further, Step 2: Design task demand representation quantity SOT and SOE according to real-time changing task demand, and the specific definitions are as follows:

[0015]

[0016] In the formula, SOT is the current task time condition, T c is the current task time, T t is the actual task limit time. SOE is the current vehicle energy storage condition, E c is the current vehicle energy consumption, E t is the actual total vehicle energy storage.

[0017] Based on the above formula, the planning data collected in Step 1 is processed. In Step 1, the task demand limit of the flying vehicle is not considered, and the collected data may exceed the actual task demand limit. The data exceeding the limit should be removed, that is, to ensure that T c ≤T t , E c ≤E t . The SOT and SOE under different weight values are solved There are n groups of data.

[0018] Set the weight value of the task time cost as K T , and the weight value of the motion energy consumption cost as K E . Each group of [SOT i SOE i ] has a corresponding i=1, 2, …, n, as follows:

[0019]

[0020] In practical applications, the current K T will be adjusted to the weight value when 1-SOT, that is, K T′ ; K E will be adjusted to the weight value when 1-SOE, that is, K E′ , and the processed data is as follows:

[0021]

[0022] In the formula, is the weight value corresponding to [1-SOT i 1-SOE i ], i=1, 2, …, n.

[0023] The weight mathematical representation under different task demands is fitted by using RBFNN, as follows:

[0024]

[0025] RBFNN includes input layer, hidden layer, and output layer. In the formula, w j is the connection weight of RBFNN hidden layer neuron and network output; h j ([SOT SOE]) is the hidden layer activation function.

[0026] In the training process, the input data is The output data is

[0027] Through the above weight mathematical representation, the appropriate weight value under different task requirements is obtained.

[0028] Further, the specific method of Step3 is:

[0029] Based on the weight mathematical representation constructed in Step2, the weighted A* algorithm with dynamic weight adjustment is designed. Considering the comprehensive optimization of task time and motion energy consumption, a comprehensive cost function containing task time cost and motion energy consumption cost is designed. The algorithm searches the passable path node, dynamically adjusts the weight value of each cost according to the real-time changing task requirement, calculates the path node with the minimum comprehensive cost, and finally realizes the reasonable switching of different motion modes, and plans a short-time energy-saving task path.

[0030] The comprehensive cost function is designed as follows:

[0031]

[0032] In the formula, cost c is the comprehensive cost of the current search path node; T c is the task time cost, i.e. the time used for the current task; is the current weight value of the task time cost; E t is the motion energy consumption cost, i.e. the current vehicle energy consumption; is the current weight value of the motion energy consumption cost; H c is the heuristic term, i.e. the Euclidean distance from the current search path node to the key point; based on the weight mathematical representation constructed in Step2, according to the real-time changing task requirement, and are in a dynamic adjustment state.

[0033] Specific planning process:

[0034] In the early stage of planning, i.e. at the path point A, the vehicle has sufficient energy storage. At the path point A, the current weight value of the task time cost The current weight value of the motion energy consumption cost is set to be large and small With more attention to the task time cost, save task time, at this time, through the cost calculation and comparison of different path nodes in Step3 by the comprehensive cost function, the vehicle takes the air flight mode at A point;

[0035] As the planning proceeds, especially at the end of the planning, i.e. at path point B, the vehicle is insufficient in energy storage, and the SOE decreases significantly faster than the SOT. At path point B, the current weight value of the task time cost The current weight value of the motion energy consumption cost Based on the weight mathematical representation constructed in Step2, the weight values of each cost are dynamically adjusted, at this time, Through the cost calculation and comparison of different path nodes in Step3 by the comprehensive cost function, the vehicle takes the ground driving mode at B point to ensure energy-saving driving and avoid early depletion of vehicle energy.

[0036] The present application has the technical effects:

[0037] 1. A weight dynamically adjusted flight vehicle path planning method is proposed, a comprehensive cost function is designed, and the comprehensive optimization of the task time and motion energy consumption costs is mainly considered. The present method searches for the path node with the minimum comprehensive cost (including the mode switching path point), realizes the reasonable switching of different motion modes, and plans a short-time energy-saving task path.

[0038] 2. The weight mathematical representation under different task requirements is constructed by RBFNN, and the weight values of different costs are dynamically adjusted during the path node search and cost calculation process to meet the real-time changes of task requirements. BRIEF DESCRIPTION OF DRAWINGS

[0039] Figure 1 An algorithm flowchart is provided for the present application;

[0040] Figure 2 A planning process schematic diagram is provided for the present application. DETAILED DESCRIPTION

[0041] The specific technical solutions of the present application are described in combination with the drawings. Firstly, the planning data under different fixed weight values are collected, including the fixed weight values of each cost, the task time after planning, and the vehicle energy consumption after planning. The data collection process does not consider the task demand limit of the flying vehicle, i.e. it is assumed that the vehicle has sufficient energy storage and the task has sufficient time limit. Secondly, according to the actual vehicle energy storage and the task time limit, the above data are processed, and the weight mathematical representation under different task demands is further fitted by using a radial basis function neural network (RBFNN). Finally, a weighted A* algorithm with dynamic weight adjustment is designed, based on the above weight mathematical representation, the weight values of different costs are dynamically adjusted according to the real-time changing task demand, the path node (including the mode switching path point) with the minimum comprehensive cost is searched, the different motion modes are reasonably switched, and a short-time energy-saving task path is planned.

[0042] The specific process is shown in Figure 1

[0043] Step 1: Planning data collection. The grid map is rasterized, the starting point and the ending point are input, and different fixed weight values are set for the task time cost and the motion energy consumption cost. The path is planned and the planning data is collected. The planning data includes the fixed weight values of each cost, the task time after planning, and the vehicle energy consumption after planning. It should be noted that the task demand limit of the flying vehicle is not considered in this stage, i.e. it is assumed that the vehicle has sufficient energy storage and the task has sufficient time limit; in other words, the vehicle has sufficient energy storage, so the weight value of the energy consumption cost can be as small as possible, and the task has sufficient time limit, so the weight value of the task time cost can be as small as possible.

[0044] Step 2: Construction of weight mathematical representation. For the real-time changing task demand, the task demand representation quantities SOT and SOE are designed, which are specifically defined as follows:

[0045]

[0046] In the formula, SOT (State of Time) is the current task time condition, T c is the current task time, T t is the actual task time limit. SOE (State of Energy) is the current vehicle energy storage condition, E t is the current vehicle energy consumption, E t is the actual total vehicle energy storage.

[0047] ​Based on the above formula, the planning data collected in Step 1 is processed. It should be noted that in Step 1, the task demand constraints of the flying vehicle are not considered (assuming that the vehicle has sufficient energy storage and the task has sufficient time limit), and the collected data may exceed the actual task demand constraints (actual vehicle energy shortage, actual task time limit). The data exceeding the limit should be removed, that is, to ensure that T c ≤T t , E c ≤E t . The SOT and SOE under different weight values are solved There are n groups of data.

[0048] The weight value of the task time cost is K T , and the weight value of the motion energy cost is K E . Each group of [SOT i SOE i ] has a corresponding i = 1, 2, …, n, as follows:

[0049]

[0050] In practical applications, taking SOE as an example, when SOE is small, the vehicle energy storage is insufficient, and the current K E should be increased to pay more attention to the motion energy cost. In this method, the current K E will be adjusted to the weight value when SOE is 1-SOE, that is, K E′ . The adjustment of K T is the same. Based on the above analysis, the further processed data is as follows:

[0051]

[0052] In the formula, is the weight value corresponding to [1-SOT i 1-SOE i ], i = 1, 2, …, n.

[0053] Based on the above processed data, the weight mathematical representation under different task demands is further fitted by RBFNN, as follows:

[0054]

[0055] RBFNN includes input layer, hidden layer, and output layer. In the formula, w j is the connection weight between the hidden layer neurons of RBFNN and the network output; h j ([SOT SOE]) is the hidden layer activation function. It should be noted that in the training process, the input data is The output data is

[0056] By the above weight mathematical representation, appropriate weight values under different task requirements can be obtained. For example, at the beginning of planning, the vehicle has sufficient energy storage, and a larger K T and a smaller K E can be set in advance to save time, and the vehicle will frequently fly over obstacles; as the planning proceeds, the current vehicle energy storage decreases, i.e., SOE decreases, and the SOE decreases at a speed significantly faster than SOT, at which time the task requirement changes, and the vehicle needs to travel economically, based on the above-obtained weight mathematical representation under different task requirements, K T and K E are adaptively adjusted, i.e., K T decreases and K E increases. In summary, during the planning process, the task requirement changes in real time (SOT and SOE change in real time), based on the above weight mathematical representation, K T and K E will also be dynamically adjusted accordingly.

[0057] Step 3: Task path planning. Based on the weight mathematical representation constructed in Step 2, a weighted A* algorithm with dynamic weight adjustment is designed. This algorithm mainly considers the comprehensive optimization of the task time and motion energy consumption, and a comprehensive cost function including the task time cost and the motion energy consumption cost is designed. In the actual planning process, the algorithm searches for passable path nodes, dynamically adjusts the weight values of each cost according to the real-time changing task requirement, calculates the path node with the minimum comprehensive cost (including the modal switching path point), and finally realizes the reasonable switching of different motion modes, and plans a short-time energy-saving task path. The comprehensive cost function is designed as follows:

[0058]

[0059] In the formula, cost c is the comprehensive cost of the current search path node; T c is the task time cost, i.e., the current task time; is the current weight value of the task time cost; E c is the motion energy consumption cost, i.e., the current vehicle energy consumption; is the current weight value of the motion energy consumption cost; H c is the heuristic term, i.e., the Euclidean distance from the current search path node to the key point; based on the weight mathematical representation constructed in Step 2, the task requirement changes in real time, and are in a dynamic adjustment state.

[0060] The specific planning process is shown in Figure 2 .

[0061] At the beginning of planning, i.e. at path point A, the vehicle has sufficient energy storage. At path point A, the current weight value of the task time cost The current weight value of the motion energy consumption cost A large And a small In order to pay more attention to the task time cost and save task time, through the cost calculation and comparison of different path nodes in Step 3, the vehicle adopts the air flight mode at point A;

[0062] As the planning proceeds, especially at the end of the planning, i.e. at path point B, the vehicle has insufficient energy storage, and the SOE decreases significantly faster than the SOT. At path point B, the current weight value of the task time cost The current weight value of the motion energy consumption cost Based on the weight mathematical representation constructed in Step 2, the weight values of each cost are dynamically adjusted, at this time, And Through the cost calculation and comparison of different path nodes in Step 3, the vehicle adopts the ground driving mode at point B to ensure energy-saving driving and avoid early depletion of vehicle energy.

[0063] The method mainly considers the comprehensive optimization of the task time and motion energy consumption costs, but in subsequent applications, it is not limited to the above two costs, and the addition of multiple costs in the comprehensive cost function in the future can be considered as the protection scope of the patent.

[0064] The application proposes a weight dynamically adjusted flight vehicle path planning method, designs a comprehensive cost function, which includes the task time cost and the motion energy consumption cost, based on the above function, searches for the path node (including the mode switching path point) with the minimum comprehensive cost, realizes the reasonable switching of different motion modes, and plans a short-time energy-saving task path.

[0065] The application collects planning data under different fixed weight values, designs a task demand representation quantity according to the real-time changing task demand, processes the above collected data by using the designed representation quantity, based on the processed data, fits the weight mathematical representation under different task demands by using the RBFNN, and further, in the path node searching and cost calculation process, dynamically adjusts the weight values of different costs based on the constructed weight mathematical representation to meet the real-time changing task demand.

Claims

1. A flying vehicle path planning method with dynamic weight adjustment, characterized in that: The following steps are involved: Step 1: Planning data collection: collect planning data under different fixed weight values. The data collection process does not consider the mission requirements of the flight vehicle, that is, it assumes that the vehicle has sufficient energy storage and the mission time limit is sufficient. Step 2: Constructing a weighted mathematical representation: Based on the actual vehicle energy storage and mission time constraints, the above planning data is processed and a radial basis function neural network (RBFNN) is used to further fit the weighted mathematical representations under different mission requirements. Step 3: Task path planning: Design a weighted A* algorithm with dynamic weight adjustment. Based on the mathematical representation of the weights, dynamically adjust the weight values ​​of different costs according to the real-time changing task requirements, search for the path node with the lowest comprehensive cost, including the mode switching path point, reasonably switch between different motion modes, and plan a short-term energy-saving task path.

2. The method for dynamically adjusting weights for a flying vehicle path planning according to claim 1, characterized in that: Step 1: Grid the map, input the starting and ending points, set multiple sets of different fixed weight values ​​for the task time cost and movement energy cost, plan the path and collect planning data; the planning data includes the fixed weight values ​​of each cost, the task time after the planning is completed, and the vehicle energy consumption after the planning is completed.

3. The method for dynamically adjusting weights for a flying vehicle path planning according to claim 1, characterized in that: Step 2: Design task requirement representations SOT and SOE based on real-time changing task requirements. The specific definitions are as follows: Where SOT is the current task time, T c is the time taken for the current task, T t It is the actual task limit time; SOE is the current vehicle energy storage situation, E c is the current vehicle energy consumption, E t is the actual total vehicle energy storage; Based on the above formula, the planning data collected in Step 1 is processed. In Step 1, the mission requirements of the flight vehicle are not considered. The collected data may exceed the actual mission requirements. The data exceeding the limit should be eliminated, that is, to ensure T c ≤T t , E c ≤E t ; The SOT and SOE under multiple groups of different weight values ​​are There are n sets of data in total; Set the weight of the task time cost to K T , the weight of the energy consumption cost is K E ; Each of the above groups [SOT i SOE i ] have corresponding As shown below: In practical applications, the current K T The weight value will be adjusted to 1-SOT, that is, K T' ;K E The weight value will be adjusted to 1-SOE, that is, K E' , the processed data is as follows: Where, Yes [1-SOT i 1-SOE i ] corresponding weight value, i=1,2,…,n; RBFNN is used to fit the weight mathematical representation under different task requirements, as shown below: RBFNN includes input layer, hidden layer, and output layer; where w j is the connection weight between the hidden layer neurons of RBFNN and the network output; h j ([SOT SOE]) is the hidden layer activation function; During training, the input data is The output data is Through the above weight mathematical representation, appropriate weight values ​​under different task requirements are obtained.

4. The method for dynamically adjusting weights for a flying vehicle path planning according to claim 1, wherein: Step 3: Based on the mathematical representation of weights constructed in Step 2, a weighted A* algorithm with dynamic weight adjustment is designed. Considering the comprehensive optimization of the two costs of task time and motion energy consumption, a comprehensive cost function that includes task time cost and motion energy consumption cost is designed. The algorithm searches for traversable path nodes, dynamically adjusts the weight values ​​of each cost according to the real-time changing task requirements, calculates the path node with the minimum comprehensive cost, and ultimately achieves reasonable switching between different motion modes and plans a short-term energy-saving task path. The comprehensive cost function is designed as follows: Where cost c is the comprehensive cost of the current search path node; T c is the task time cost, that is, the time taken for the current task; is the current weight value of the task time cost; E c is the energy cost of motion, i.e. the current energy consumption of the vehicle; is the current weight value of the energy cost of exercise; H c is the heuristic term, i.e., the Euclidean distance from the current search path node to the focus; based on the weight mathematical representation constructed in Step 2, for real-time changing task requirements, and In a dynamic adjustment state.

5. A flying vehicle path planning method with dynamic weight adjustment according to any one of claims 1 to 4, characterized in that: The specific planning process is: At the initial stage of planning, that is, at path point A, the vehicle has sufficient energy storage; at path point A, the current weight value of the mission time cost The current weight of the energy cost of exercise Set a large and small In order to pay more attention to the mission time cost and save mission time, at this time, the vehicle adopts the air flight mode at point A through the comprehensive cost function in Step 3 to calculate and compare the costs of different path nodes; As the planning progresses, especially at the end of the planning, that is, at path point B, the vehicle energy storage is insufficient, and the SOE decreases significantly faster than the SOT; at path point B, the current weight value of the mission time cost The current weight of the energy cost of exercise Based on the weight mathematical representation constructed in Step 2, the cost weight values ​​are dynamically adjusted. At this time, and By calculating and comparing the costs of different path nodes through the comprehensive cost function in Step 3, the vehicle adopts a ground driving mode at point B to ensure energy-saving driving and avoid premature exhaustion of vehicle energy.

Images