An Improved Path Planning Method Based on Adaptive Parameters

Abstract

The application discloses an improved path planning method based on adaptive parameters, comprising the following steps: S1, initial path planning is carried out through a DWA algorithm; S2, on the basis of the initial path, the weight of the evaluation function parameter is designed for different obstacle types and a steering strategy is prepared, S2-1, when the intelligent agent is passing through between dense obstacles, the influence of the course item on the trajectory of the initial path is reduced by designing an adaptive rate, S2-2, when facing narrow gap obstacles, the weight of the distance parameter in the evaluation function is reduced; S2-3, when facing trap obstacles, a steering strategy is formulated in the DWA algorithm; S3, the moving obstacles in the dynamic environment are predicted to avoid obstacles. Through the method, the intelligent agent can plan a real-time path in a complex environment containing a large number of static and dynamic obstacles, path constraints and performance constraints, and the intelligent agent can avoid the dynamic obstacles in advance, reducing the hidden danger of side collision.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of path planning of intelligent agents, and particularly to an improved path planning method based on adaptive parameters. BACKGROUND

[0002] Mobile intelligent agents have achieved fruitful results in recent years, and are widely applied in autonomous underwater vehicles (AUVs), unmanned ground vehicles (UGVs), unmanned aerial vehicles (UAVs), and the like. When multiple mobile intelligent agents cooperate to explore or move, a basic and important problem to be considered is path planning, which aims to design and deploy appropriate algorithms according to different environments, motion limitations of the intelligent agents, and other factors.

[0003] At present, common technical means can complete optimal path planning for static unknown environments, and have high accuracy, and can perfectly avoid obstacles. However, in reality, unknown environments are often a combination of static and dynamic environments. Therefore, conventional path planning methods cannot accurately avoid dynamic obstacles.

[0004] To solve the above technical problems, some people select a dynamic window algorithm (DWA), which plans a path for the intelligent agent in an unknown environment and has real-time obstacle avoidance capability, thereby achieving the obstacle avoidance function of dynamic obstacles and static obstacles in an unknown environment. However, in order to realize real-time calculation of the current environment, the calculation amount is very large. Therefore, in recent years, the DWA algorithm has been continuously improved, mainly including improvement of calculation amount, improvement of window accuracy, improvement of boundary definition of irregular obstacles, and adaptive improvement of evaluation function parameters, and good progress has been made.

[0005] However, the existing path planning method selecting DWA still has the following defects:

[0006] 1. When facing narrow gap obstacles, the original algorithm will cause sudden deceleration and jitter due to the too fast iteration judgment of invalid trajectories and the limited prediction trajectories not meeting the requirement of crossing the gap.

[0007] 2. In actual environments, various trap obstacles often appear, such as non-convex region traps, straight line traps, and the like. Among them, the straight line trap obstacle as shown in FIG. 1B refers to the intelligent agent, the obstacle, and the target point being on a straight line. In this case, due to symmetry, the intelligent agent cannot actively turn to bypass the obstacle. In addition, the non-convex region trap as shown in FIG. 1A can appear in warehouses, houses, and other structures. When local information is obtained by using limited visual sensors or other environmental sensors, it is difficult to issue a warning before falling into the non-convex region trap, and it is also difficult to decelerate outside the trap. Figure 2 Figure 2

[0008] ​​3、Because the dynamic obstacle is difficult to be mathematically defined in the unknown environment, only the relative position thereof can be obtained through the sensor of the intelligent agent, so that the original DWA algorithm cannot avoid the lateral collision.

[0009] 4、The adaptive improvement of the existing evaluation function parameters is only for the speed weight in one environment, and the heading angle evaluation function and the distance evaluation function in the evaluation function are not considered, so that the path planning still has certain defects. SUMMARY

[0010] According to the deficiencies of the prior art, the application provides an improved path planning method based on adaptive parameters, which realizes real-time path planning for an intelligent agent in a complex environment containing a large number of static and dynamic obstacles, path constraints and performance constraints.

[0011] In order to solve the above technical problems, the technical scheme of the application is as follows:

[0012] An improved path planning method based on adaptive parameters comprises the following steps:

[0013] S1, initial path planning is performed through a DWA algorithm;

[0014] S2, on the basis of the initial path, the weight of the evaluation function parameter is designed for different obstacle types, and a steering strategy is made,

[0015] S2-1, when the intelligent agent passes through the dense obstacles, the influence of the heading item on the trajectory of the initial path is reduced through the design of an adaptive rate, wherein the expression of the heading angle adaptive rate a(t) is as follows:

[0016]

[0017] Wherein, λ1>1 is the weight of the adaptive parameter of the heading angle, p(t) is the position of the intelligent agent at t, ob(t) indicates the position of the obstacle closest to the intelligent agent at t, ζ1 indicates the distance threshold, ρ1 is one of the adjustable parameters for adjusting the adaptive change rate, and is set to ρ1>1;

[0018] S2-2, when facing the narrow gap obstacles, the weight of the distance parameter in the evaluation function is reduced, wherein the design of the distance adaptive rate β(t) is as follows:

[0019]

[0020] Wherein, λ2>0 is the weight of the distance adaptive parameter;

[0021] S2-3, when facing the trap obstacles, a steering strategy is made in the DWA algorithm, wherein the steering strategy is defined as follows:

[0022]

[0023] where ω turn is the rotation angle of the agent in unit time t, the agent rotates counterclockwise or clockwise, which is determined by ω turn > 0 and ω turn < 0, respectively;

[0024] S3, predicting and avoiding obstacles for moving obstacles in a dynamic environment.

[0025] As a preferred, the step S1 comprises the following sub-steps:

[0026] S1-1, obtaining the state of the agent at the next time by integral solution of the velocity v and the angle ω in the preset time interval;

[0027] S1-2, calculating the feasible velocity range V d that the agent can move in the preset time interval;

[0028] S1-3, considering the feasible velocity range V a limited by the safe braking distance between the agent and the obstacle;

[0029] S1-4, obtaining the maximum velocity space V s limited by the physical performance of the agent itself;

[0030] S1-5, combining the feasible velocity range V d , the feasible velocity range V a and the maximum velocity space V s , obtaining the intersection of the three, that is, the velocity window V r available for the agent, which is expressed as: V r = V d ∪ V a ∪ V s ;

[0031] S1-6, traversing each velocity in the velocity window V r available for the agent through the evaluation function.

[0032] As a preferred, the state at the next time in the step S1-1 comprises absolute coordinates, velocity and moving angle.

[0033] As a preferred, in the step S1-1, the method for obtaining the state of the agent at the next time is:

[0034] defining the general motion equation of the agent:

[0035]

[0036] where t0 denotes the initial time, v represents the speed of the agent, and 0 represents the heading angle of the robot, with the horizontal line as the starting angle, x(t n ) represents the x-coordinate of the agent at time t n ,

[0037] At the same time, a completely discrete expression can also be obtained:

[0038]

[0039] where,

[0040]

[0041] When the angular velocity ω i = 0, the agent will move along a straight line, and when ω i ≠ 0, it will make a circular motion with the center at and the radius :

[0042]

[0043]

[0044]

[0045] As a preferred, in the step S1-2, the agent can move in the predetermined time interval V d range of feasible speeds.

[0046]

[0047] where (v a , ω a ) represents the actual speed and angular velocity of the agent.

[0048] As a preferred, in the step 1-3, the method for obtaining the feasible speed range V a considering the safety braking distance limit between the agent and the obstacle:

[0049]

[0050] where dist(v, ω) represents the distance between the agent and the nearest obstacle under the speed v and angular velocity ω.

[0051] As a preferred, in the step S1-6, the method for traversing each speed in the agent's selectable speed window V r :

[0052] G(v, ω) = σ (a * heading (v, ω) + b * dist (v, ω) + g * velocity (v, ω))

[0053] Wherein, a, b, g are weight coefficients respectively, Gvelocity(v,ω) represents the speed score of the agent, heading(v,ω) represents the heading angle score of the agent, dist(v,ω) represents the distance score between the agent and the nearest obstacle, and the three evaluation indexes are normalized by σ(·).

[0054] As a preferred, the step S3 comprises the following sub-steps:

[0055] S3-1, observation stage: capturing the absolute coordinates of the moving obstacle at t n time in the limited sensor detection range Continuously observing m-n (m>n) time slices to establish position information, i.e. coordinate matrix

[0056]

[0057] S3-2, prediction stage: predicting the stored position information , using a high-order fitting function based on the least square idea to predict the horizontal and vertical coordinates respectively, to obtain a fitting coefficient combination

[0058] Due to the irregular motion of the dynamic obstacle, the fixed fitting coefficient is not enough to accurately predict the motion trend of the dynamic obstacle, therefore, after t m time, the rolling window idea is used to update and accurately calculate the motion trend of the dynamic obstacle until the obstacle exceeds the observation range of the sensor;

[0059] S3-3, application stage: obtaining the position of the dynamic obstacle after t m +s*Δt time by prediction

[0060] As a preferred, in the step S3-2, the prediction process is based on time series, i.e. using the position information recorded in the current time period to predict the possible position of the obstacle in the future Wherein, LS is a fitting function based on the least square method.

[0061] The present application has the following characteristics and beneficial effects:

[0062] By optimizing the weight of the evaluation function in the DWA algorithm, the agent can quickly and safely pass through the dense obstacles and narrow gap obstacles. The designed steering strategy can also help the agent escape from the non-convex region trap obstacle and straight line obstacle with high probability.

[0063] In addition, based on the path tracking of the dynamic obstacle, the future movement trend of the dynamic obstacle is obtained by the fitting prediction method, which reduces the probability of collision of the agent in the dynamic environment and avoids most lateral collision cases. BRIEF DESCRIPTION OF DRAWINGS

[0064] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed to be used in the embodiments or the prior art description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0065] Figure 1 The specific steps for passing through the narrow gap obstacle in the present embodiment are shown in the schematic diagram.

[0066] Figure 2 The schematic diagram of the two types of trap obstacles, straight line trap obstacle and non-convex region trap obstacle, in the present embodiment is shown.

[0067] Figure 3 The schematic diagram of the trap obstacle escaped by the steering strategy in the present embodiment is shown.

[0068] Figure 4 The schematic diagram of the number of times that the improved DWA algorithm is better than the original algorithm under different obstacle densities in the present embodiment is shown.

[0069] Figure 5 The schematic diagram of the comparison of the failure number and the average steering number of the two algorithms under different obstacle densities in the present embodiment is shown.

[0070] Figure 6 The flow chart of the dynamic obstacle trajectory prediction step in the present embodiment is shown.

[0071] Figure 7 The dynamic obstacle avoidance route with prediction obstacle avoidance in the present embodiment is shown. DETAILED DESCRIPTION

[0072] It should be noted that the embodiments in the present application and the features in the embodiments can be combined with each other without conflict.

[0073] In the description of this invention, it should be understood that the terms "center," "longitudinal," "lateral," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicating orientations or positional relationships based on the orientations or positional relationships shown in the accompanying drawings, are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention. Furthermore, the terms "first," "second," etc., are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined with "first," "second," etc., may explicitly or implicitly include one or more of that feature. In the description of this invention, unless otherwise stated, "a plurality of" means two or more.

[0074] In the description of this invention, it should be noted that, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art will understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0075] This invention provides an improved path planning method based on adaptive parameters, such as... Figure 6 As shown, it includes the following steps:

[0076] S1. Plan the initial path using the DWA algorithm.

[0077] Specifically, S1-1 obtains the agent's state at the next moment by integrating the velocity v and angle ω within a preset time interval. The state at the next moment includes the absolute coordinates, velocity, and movement angle.

[0078] In this embodiment, taking the horizontal coordinate as an example, the general motion equation of the agent is first defined:

[0079]

[0080] Where t0 represents the initial time, v represents the agent's velocity, θ represents the robot's heading angle, with the horizontal line as the starting angle, and x(t) n ) represents t n The x-coordinate of the agent at any given time.

[0081] At the same time, a completely discrete form of expression can also be obtained:

[0082]

[0083] in,

[0084]

[0085] When angular velocity ω i When ω = 0, the agent will move along a straight line. i ≠0 o'clock With the center of the circle, Circular motion with radius:

[0086]

[0087]

[0088]

[0089] S1-2, Calculate the feasible speed range V that the agent can move within a preset time interval. d .

[0090] Specifically, given a finite linear and angular acceleration, the agent can move within a feasible velocity range V over a predetermined time interval. d How to obtain:

[0091]

[0092] Among them, (v a ,ω a ) represents the actual velocity and angular velocity of the intelligent agent.

[0093] In this embodiment, the predetermined time interval is set to 1-5 seconds.

[0094] S1-3, Feasible speed range V considering safe braking distance limits with respect to obstacles. a ;

[0095] Specifically, the feasible speed range V considering the safe braking distance limit between the vehicle and the obstacle. a How to obtain:

[0096]

[0097] Here, dist(v,ω) represents the distance between the agent and the nearest obstacle, given the velocity v and angular velocity ω.

[0098] S1-4. Obtain the maximum velocity space V, which is limited by the physical performance of the agent itself. s ;

[0099] S1-5, Combined with feasible speed range V dFeasible speed range V a and the maximum velocity space V s The intersection of these three factors is obtained, which is the agent's selectable velocity window V. r , represented as: V r =V d ∪V a ∪V s ;

[0100] S1-6. Traverse the agent's selectable velocity window V using the evaluation function. r Each speed in the middle.

[0101] Specifically, the agent can traverse selectable velocity windows V. r The method for each velocity in the data:

[0102] G(v,ω)=σ(α·heading(v,ω)+β·dist(v,ω)+γ·velocity(v,ω))

[0103] Where α, β, and γ are weighting coefficients, Gvelocity(v,ω) represents the agent's velocity score, heading(v,ω) represents the agent's heading angle score, and dist(v,ω) represents the distance score between the agent and the nearest obstacle. The three evaluation metrics will be normalized by σ(·).

[0104] S2. Based on the initial path, design weights for the evaluation function parameters for different obstacle types and formulate a turning strategy.

[0105] S2-1. When an agent moves through dense obstacles, in order to enable the agent to pass through the dense obstacles as quickly as possible, it is necessary to reduce the heading score. By designing an adaptive rate to reduce the influence of the heading term on the trajectory of the initial path, the agent moving through dense obstacles can maintain a stable speed while maintaining a safe distance. The expression of the heading angle adaptive rate α(t) is as follows:

[0106]

[0107] Where λ1>1 is the weight of the adaptive parameter of the heading angle, p(t) is the position of the agent at time t, ob(t) refers to the position of the nearest obstacle to the agent at time t, ζ1 refers to the distance threshold, and ρ1 is one of the adjustable parameters for adjusting the adaptive change rate, and ρ1>1 is set.

[0108] S2-2. When facing obstacles in narrow gaps, reduce the weight of the distance parameter in the evaluation function. The design of the distance adaptation rate β(t) is as follows:

[0109]

[0110] Where λ2>0 are the weights of the distance adaptive parameter;

[0111] Understandably, the original algorithm suffers from sudden deceleration and jitter due to its overly rapid iteration of invalid trajectories and the limited predicted trajectories failing to meet the requirements for traversing the gaps. Appropriately reducing the weight of the distance component in the evaluation function prevents the agent from excessively eliminating obstacles, thus allowing for a smoother passage through narrow gaps and obstacles.

[0112] like Figure 1 As shown, the black object represents a narrow gap obstacle, the circle represents the agent, and the extending arrow represents the finite predicted trajectory generated by the DWA algorithm. The three stages respectively reflect the process of the agent from facing the narrow gap obstacle to decelerating and passing through it.

[0113] Unlike the fixed weights in the original algorithm, in Figure 1 The (a) and (b) phase design allows for dynamic adjustment of the distance term weights, preventing the agent from excessively rejecting predicted trajectories that approach obstacles. As a result, the agent does not decelerate prematurely or oscillate excessively.

[0114] S2-3. When facing traps and obstacles, formulate a turning strategy in the DWA algorithm, where the turning strategy is defined as follows:

[0115]

[0116] Where, ω turn It is the angle of rotation of the agent within a unit time t. The agent rotates counterclockwise or clockwise, respectively, determined by ω. turn >0 and ω turn <0 decision;

[0117] Understandably, in real-world environments, various traps and obstacles often appear, such as non-convex area traps and straight-line traps.

[0118] in, Figure 2 The straight-line trap obstacle shown on the right refers to an situation where the agent, obstacle, and target point are on a straight line. In this case, due to symmetry, the agent cannot actively turn to bypass the obstacle. Furthermore, as... Figure 2 As shown on the left, non-convex area traps can occur in warehouses, houses, and other structures.

[0119] When using limited visual sensors or other environmental sensors to acquire local information, it is difficult to issue warnings before falling into traps in non-convex regions, and it is also difficult to decelerate outside the trap. Therefore, it is necessary to design strategies for escaping traps within local regions. The improved DWA algorithm introduces a critical collision determination: when the agent moves at extremely low speeds, due to the continuous motion requirement of the DWA algorithm, it is not possible to guarantee V... rAll predicted trajectories are safe, and this state is defined as a critical collision state. For this type of nearly stationary agent, the improved DWA algorithm drives the agent to turn and move in place, thereby attempting to navigate around the concave obstacle.

[0120] Furthermore, in order to verify the feasibility and superiority of the adaptive design of the evaluation function parameters and the steering strategy, this embodiment deployed 10, 15, 20, 25, 30, and 35 random circular obstacle maps under the conditions in Table 1, and repeated the simulation 100 times.

[0121] Table 1 Fixed Parameters and Adjustable Parameters

[0122]

[0123] The following results were obtained:

[0124] according to Figure 4 , Figure 5 The results show that the improved DWA algorithm outperforms the original DWA algorithm in terms of time, path length, number of failures, and average number of turns, especially when the obstacle density is 30.

[0125] S3. Predict and avoid moving obstacles in dynamic environments.

[0126] Understandably, dynamic obstacles are difficult to define mathematically in unknown environments, and their relative positions can only be obtained through the agent's sensors. Furthermore, the original DWA algorithm often cannot avoid lateral collisions. To address this, this study proposes a strategy whereby the agent predicts the movement trajectories (straight lines and curves) of dynamic obstacles within a certain range around it, thereby avoiding the obstacles. The specific implementation steps are as follows:

[0127] S3-1, Observation Phase: Capturing moving obstacles within the limited sensor detection range at t n Absolute coordinates of time Establish location information, i.e., coordinate matrix, by continuously observing time slices m and n (m > n).

[0128]

[0129] S3-2, Prediction Stage: Predicting the stored location information... To make predictions, a higher-order fitting function based on the least squares approach is used to predict the horizontal and vertical axes respectively, resulting in a combination of fitting coefficients. The prediction process is based on time series, that is, it uses the location information recorded in the current time period to predict the possible future location of obstacles. Where LS is the fitting function based on the least squares method;

[0130] Due to the irregular motion of dynamic obstacles, fixed fitting coefficients are insufficient to accurately predict the motion trend of dynamic obstacles. Therefore, t m After a certain time, the system uses a scrolling window to update in real time. and Accurately calculate the motion trend of dynamic obstacles until the obstacles are out of the sensor's observation range;

[0131] S3-3, Application Stage: Dynamic obstacles are predicted at time t... m Position after +s*Δt time It is added as a virtual static obstacle to the agent's obstacle map and updated in real time.

[0132] Through the above three-level obstacle avoidance strategy, such as Figure 7 As shown, the top three images show dynamic obstacle avoidance routes without predictive obstacle avoidance; the bottom three images show dynamic obstacle avoidance routes with predictive obstacle avoidance. It is clear that the agent can avoid dynamic obstacles in advance (the improved DWA algorithm determines deceleration or turning), reducing the risk of side collisions.

[0133] The embodiments of the present invention have been described in detail above with reference to the accompanying drawings, but the present invention is not limited to the described embodiments. For those skilled in the art, various changes, modifications, substitutions, and variations can be made to these embodiments, including components, without departing from the principles and spirit of the present invention, and these variations still fall within the protection scope of the present invention.

Claims

1. An improved path planning method based on adaptive parameters, characterized in that, The method comprises the following steps: S1, planning an initial path by a DWA algorithm; S2, on the basis of the initial path, designing weights of parameters of an evaluation function and making a steering strategy for different obstacle types, S2-1, when the agent is passing through dense obstacles, reducing the influence of a heading item on the trajectory of the initial path by designing an adaptive rate, wherein an expression of the heading angle adaptive rate a(t) is as follows: wherein λ1>1 is a weight of an adaptive parameter of the heading angle, p(t) is a position of the agent at t, ob(t) indicates a position of the closest obstacle to the agent at t, ζ1 indicates a distance threshold, ρ1 is one of adjustable parameters for adjusting the adaptive change rate, and ρ1>1 is set; S2-2, when facing narrow gap obstacles, reducing the weight of a distance parameter in the evaluation function, wherein a distance adaptive rate β(t) is designed as follows: wherein λ2>0 is a weight of the distance adaptive parameter; S2-3, when facing trap obstacles, making a steering strategy in the DWA algorithm, wherein the steering strategy is defined as follows: where ω turn is the rotation angle of the agent in unit time t, and the counterclockwise or clockwise rotation of the agent is determined by ω turn > 0 and ω turn < 0, respectively. S3, performing prediction obstacle avoidance for a moving obstacle in a dynamic environment.

2. The improved path planning method based on adaptive parameters according to claim 1, characterized in that, The step S1 comprises the following sub-steps: S1-1, obtaining a state of the agent at a next time by integral calculation of velocity v and angle ω within a preset time interval; S1-2, calculating a feasible speed range V that the intelligent agent can move within a preset time interval d ; S1-3, range of feasible speeds V taking into account the safety braking distance limit with obstacles a ; S1-4, obtaining the maximum speed space V of the physical performance limit of the agent itself s ; S1-5, the feasible velocity range V d , the feasible velocity range V a , and the maximum velocity space V s , the intersection of the three, i.e. the velocity window V r , which the agent can choose, is expressed as: V r = V d ∪ V a ∪ V s ; S1-6, traversing the speed window V of the agent by the evaluation function r each of the intermediate speeds.

3. The improved path planning method based on adaptive parameters according to claim 2, characterized in that, The state at the next time in the step S1-1 comprises absolute coordinates, velocity and moving angle.

4. The improved path planning method based on adaptive parameters according to claim 2, characterized in that, In the step S1-1, the state of the agent at the next time is obtained by the following method: a general motion equation of the agent is defined as follows: where t0 denotes the initial time, v represents the agent velocity, 0 represents the robot heading angle, with the horizontal line as the starting angle, x(t n ) represents the x coordinate of the agent at time t n . Meanwhile, a completely discrete expression is obtained as follows: wherein, When the angular velocity ω i = 0, the agent will move along a straight line, ω i ≠ 0, do the circular motion with the center at and the radius of :

5. The improved path planning method based on adaptive parameters according to claim 4, characterized in that, In the step S1-2, the agent can move in the feasible speed range V d Acquisition method: where (v a , ω a ) denotes the actual velocity and angular velocity of the agent.

6. The improved path planning method based on adaptive parameters according to claim 5, characterized in that, In the steps 1-3, the feasible speed range V is considered with the safety braking distance limit between the vehicle and the obstacle a Acquisition method: wherein dist(v, ω) represents a distance between the agent and the closest obstacle under the conditions of velocity v and angular velocity ω.

7. The improved path planning method based on adaptive parameters according to claim 6, characterized in that, In the step S1-6, the method of traversing each speed in the optional speed window V of the agent r may be: G(v, ω) = σ (a heading(v, ω) + β dist(v, ω) + γ velocity(v, ω)) wherein α, β and γ are weight coefficients, velocity(v, ω) represents a velocity score of the agent, heading(v, ω) represents a heading angle score of the agent, and dist(v, ω) represents a distance score between the agent and the closest obstacle, and the three evaluation indexes are normalized by σ(·).

8. The improved path planning method based on adaptive parameters according to claim 4, characterized in that, The step S3 comprises the following sub-steps: S3-1, observation phase: capture the absolute coordinates of the moving obstacle at t n Continuous observation m-n, where m>n, time slice builds location information, i.e. coordinate matrix ​ S3-2, prediction stage: on the stored position information Prediction is made, high-order fitting function based on least square idea is used to predict the horizontal and vertical coordinates respectively, and fitting coefficient combination is obtained Due to the irregular motion of the dynamic obstacle, the fixed fitting coefficient is not enough to accurately predict the motion trend of the dynamic obstacle, so that t m After a period of time, the idea of rolling window is adopted to update the fitting coefficient in real time and The motion trend of the dynamic obstacle is accurately calculated until the obstacle exceeds the observation range of the sensor; S3-3, Application stage: through prediction, get the position of dynamic obstacle at t m + s * Δt time and add it to the obstacle map of the agent as a virtual static obstacle, and update it in real time.

9. The improved path planning method based on adaptive parameters according to claim 8, characterized in that, In the step S3-2, the prediction process is based on time series, i.e. the position information recorded in the current time period is used to predict the possible future position of the obstacle where LS is a least squares based fitting function.

Images