Mobile robot navigation framework based on improved DWA in unknown environment and experimental method

Abstract

The invention discloses a mobile robot navigation framework based on improved DWA in an unknown environment and an experimental method. The mobile robot navigation framework comprises an environment sensing module used for acquiring obstacle information of the unknown environment in real time through a laser radar; the improved DWA planning module comprises a dynamic window generation unit for generating a speed window according to mechanical constraints of the robot, and a multi-factor evaluation unit; the multi-factor evaluation unit comprises a navigation function, an obstacle avoidance function, a speed function, an escape function and a target approaching function; the execution module is used for sending an optimal speed instruction output by the evaluation function to the motor controller; and the framework optimization module is used for cancelling dependence on a global map and a reference path and directly tracking a global target point. Through an improved navigation framework, the algorithm directly tracks a global target point, the dependence on a global map and a reference path is removed, and the speed of an optimal track is published to a motor controller for execution, so that navigation is realized and unknown obstacles are effectively processed under the condition of no reference path.

Description

Technical Field

[0001] The present invention relates to the technical field of robot local path planning, and in particular to a mobile robot navigation framework and experimental method based on improved DWA in an unknown environment. Background Art

[0002] Mobile robots have been widely adopted in various fields of production and life. However, most current navigation algorithms rely on prior information such as known maps, accurate positioning, and obstacle distribution for navigation and obstacle avoidance. This makes them difficult to adapt to the increasing number of application scenarios involving unknown factors. Furthermore, the development of navigation methods that can perceive and respond to environmental uncertainties in real time has been slow.

[0003] Traditional DWA (Dynamic Window Algorithm) algorithms are irrational, resulting in low efficiency and a tendency to fall into local optimality. Furthermore, traditional autonomous navigation frameworks rely on global maps and precise positioning information, making them difficult to operate in unknown environments. When unknown obstacles appear, traditional methods repeatedly run a global planner (such as A*) based on local perception, which reduces real-time performance and efficiency in environments with numerous obstacles. Summary of the Invention

[0004] The purpose of the present invention is to overcome the shortcomings of the existing technology. To achieve the above purpose, a mobile robot navigation framework and experimental method based on improved DWA in an unknown environment are adopted to solve the problems raised in the above background technology.

[0005] A mobile robot navigation framework based on improved DWA in unknown environments, including:

[0006] An environmental perception module, which is used to obtain obstacle information in an unknown environment in real time through a laser radar;

[0007] An improved DWA planning module, wherein the improved DWA planning module includes a dynamic window generation unit and a multi-factor evaluation unit;

[0008] The dynamic window generation unit is used to generate a speed window based on the robot's mechanical constraints and remove traditional obstacle constraints;

[0009] The multi-factor evaluation unit includes a navigation function, an obstacle avoidance function, a speed function, an escape function, and a target approach function;

[0010] The navigation function is used to calculate the target deviation angle based on the position after the distance is moved on the predicted trajectory;

[0011] The obstacle avoidance function is used to calculate the shortest distance from the starting point of the trajectory to the obstacle within a preset time interval and discard trajectories below the safety threshold;

[0012] The speed function is used to add an angular velocity smoothing scoring item to suppress trajectory oscillation;

[0013] The escape function is used to record the historical trajectory cost through a rasterized map and evaluate the similarity between the predicted trajectory and the historical trajectory;

[0014] The target approach function is used to calculate the shortest distance between the trajectory point and the target;

[0015] An execution module, configured to send the optimal speed instruction output by the evaluation function to a motor controller;

[0016] The framework optimization module is used to cancel the dependence on the global map and the reference path and directly track the global target point.

[0017] As a further solution of the present invention: the current state of the robot includes the current position, direction angle, linear velocity and angular velocity of the robot;

[0018] The motion parameters include the robot's maximum linear velocity, maximum angular velocity, acceleration and angular acceleration.

[0019] As a further solution of the present invention: there are mechanical constraints and obstacle constraints on the speed window; the robot stops before encountering an obstacle:

[0020]

[0021] Where obs(v,ω) is the shortest distance from the predicted trajectory to the obstacle; p is the safety factor.

[0022] As a further solution of the present invention: the navigation function is a heading(v,ω) function, which calculates the deviation angle by predicting the position of the end of the trajectory; the correction is to change the reference position for calculating the deviation angle to the position after several time intervals on the predicted trajectory, and to determine the distance d h The deviation angle is calculated by combining the robot linear velocity prediction position after the time interval. The time interval is the number n Δt,h for:

[0023] n Δt,h =fix(d h / v)

[0024] The final function form is as follows:

[0025] heading′(v,ω)=-N[|θ(r nΔt,h ,G)-θ(r nΔt,h )|];

[0026] Among them, θ(r nΔt,h ,G) is n Δt,hThe angle at which the robot points to the target after a time interval; θ(r nΔt,h ) is the robot orientation at that position; N represents normalization.

[0027] As a further solution of the present invention: the obstacle avoidance function is an obdist(v, ω) function, which adds a function of discarding dangerous trajectories. When calculating the shortest distance to an obstacle, only the position from the starting point of the trajectory to a certain time interval is considered, and an upper limit of the obstacle distance is set.

[0028] The modified obdist(v,ω) function not only has the function of evaluating the original trajectory, but also discards the trajectory from the starting point to n Δt,o The trajectories whose shortest distance from the obstacle after the time interval is less than the safety distance are in the form of:

[0029]

[0030] Among them, d nΔt,o is the starting point to n on the trajectory Δt,o The shortest distance between the robot's position and the obstacle in the time interval; d p is the farthest obstacle considered by the function, and R is the robot's safety radius.

[0031] As a further solution of the present invention: after modifying and adding each sub-evaluation function, the evaluation function of the improved DWA algorithm is obtained, and the formula is:

[0032] C(v,ω)=ω1·heading′(v,ω)+ω2·obdist′(v,ω)+ω3·velocity′(v,ω)+ω4·osc(v,ω)+ω5·goal(v,ω).

[0033] As a further solution of the present invention: the framework optimization module is integrated into the move_base node in the ROS system, directly receives the global target point and outputs the speed instruction to the motor control node.

[0034] The technical solution of the second aspect: an experimental method for a mobile robot navigation framework based on an improved DWA in an unknown environment as described in any of the above items, comprising the following steps:

[0035] Step S10: Experimental platform construction:

[0036] Configure a mobile robot platform equipped with lidar, IMU and odometer sensors, and install the ROS operating system;

[0037] Step S20: unknown environment construction:

[0038] Setting up static unknown obstacles and / or dynamic moving obstacles in the experimental field;

[0039] Step S30: Navigation framework deployment:

[0040] Load the improved navigation framework into the robot's main control computer and load the original map without any obstacles;

[0041] Step S40: Target-driven navigation:

[0042] Publish the global target point through Rviz software and start the autonomous operation of the navigation framework;

[0043] Step S50: Dynamic obstacle avoidance verification:

[0044] When a dynamic obstacle intrudes into the robot's path, the real-time obstacle avoidance trajectory is recorded;

[0045] Step S60: Performance evaluation:

[0046] Collect trajectory length, time consumption, and minimum obstacle distance indicators and compare them with traditional DWA results.

[0047] As a further solution of the present invention: in step S20, the static obstacles constitute a local optimal trap, and the dynamic obstacles are regularly moving pedestrian models.

[0048] As a further solution of the present invention: Step S60 verifies the global search capability in the following manner:

[0049] When the robot enters an encircled obstacle area, detect whether it escapes within the preset time;

[0050] Statistics on navigation success rate in unknown environments.

[0051] Compared with the prior art, the present invention has the following technical effects:

[0052] By adopting the above technical solution, the core mechanism of the dynamic window algorithm (DWA) is improved by relaxing the speed window constraint and constructing a multi-factor scoring system with five types of modified / new evaluation functions (covering target deviation angle optimization, dangerous trajectory elimination, angular velocity smoothing, local optimal escape and target approach). At the same time, the ROS navigation architecture is reconstructed to eliminate the dependence on the global map and reference path, so that the move_base node can directly drive the improved DWA to generate real-time speed instructions based on the target point coordinates.

[0053] An improved navigation framework was designed to enable the algorithm to directly track the global target point, removing its dependence on the global map and reference path. The speed of the optimal trajectory was published to the motor controller for execution, ensuring the real-time performance of the algorithm. This enabled it to navigate without the need for a reference path and effectively handle unknown obstacles. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] The specific embodiments of the present invention are described in detail below with reference to the accompanying drawings:

[0055] Figure 1 This is a schematic diagram of the improved autonomous navigation framework structure based on the move_base node according to the embodiment disclosed in this application;

[0056] Figure 2 A schematic diagram of a grid cost increment calculation method according to an embodiment disclosed in this application;

[0057] Figure 3 This is the planning effect diagram of the traditional DWA algorithm;

[0058] Figure 4 This is a planning effect diagram of the improved DWA algorithm according to the embodiment disclosed in this application;

[0059] Figure 5 The robot trajectory diagrams of the conventional and improved DWA of the embodiments disclosed in this application;

[0060] Figure 6 These are obstacle avoidance effect diagrams of the embodiments disclosed in this application without using the navigation framework of the present invention and after using the navigation framework of the present invention;

[0061] Figure 7 This is a flowchart of a method for using the navigation framework according to an embodiment disclosed in this application;

[0062] Figure 8 The original map of the static scene and the actual scene of the embodiment disclosed in this application;

[0063] Figure 9 This is the static scene navigation result of the embodiment disclosed in this application;

[0064] Figure 10 The original map and actual scene of the dynamic scene in the embodiment disclosed in this application;

[0065] Figure 11 This is the dynamic scene navigation result of the embodiment disclosed in this application. DETAILED DESCRIPTION

[0066] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0067] Please refer to Figure 1 In an embodiment of the present invention, a mobile robot navigation framework based on an improved DWA in an unknown environment includes:

[0068] An environmental perception module, which is used to obtain obstacle information in an unknown environment in real time through a laser radar;

[0069] An improved DWA planning module, wherein the improved DWA planning module includes a dynamic window generation unit and a multi-factor evaluation unit;

[0070] The dynamic window generation unit is used to generate a velocity window based on the robot's mechanical constraints, removing traditional obstacle constraints;

[0071] The multi-factor evaluation unit includes navigation function, obstacle avoidance function, speed function, escape function, and target approach function;

[0072] The navigation function is used to calculate the target deviation angle based on the position after the distance is moved on the predicted trajectory;

[0073] The obstacle avoidance function is used to calculate the shortest distance from the starting point of the trajectory to the obstacle within a preset time interval and discard trajectories that are below the safety threshold;

[0074] The velocity function is used to add an angular velocity smoothing score item to suppress trajectory oscillation;

[0075] The escape function is used to record the historical trajectory cost through the rasterized map and evaluate the similarity between the predicted trajectory and the historical trajectory;

[0076] The target approach function is used to calculate the shortest distance between the trajectory point and the target;

[0077] An execution module, the execution module is used to send the optimal speed instruction output by the evaluation function to the motor controller;

[0078] Framework optimization module,The framework optimization module is used to eliminate the dependence on the global map and reference path, and directly track the global target point.

[0079] In this embodiment, the current state of the robot includes the current position, direction angle, linear velocity and angular velocity of the robot;

[0080] The motion parameters include the robot's maximum linear velocity, maximum angular velocity, acceleration, and angular acceleration.

[0081] In this embodiment, there are mechanical and obstacle constraints on the speed window; the robot stops before hitting an obstacle:

[0082]

[0083] Where obs(v,ω) is the shortest distance between the predicted trajectory and the obstacle; p is the safety factor, which is generally 2.

[0084] In order to ensure the foresight of the predicted trajectory, the prediction time T is generally taken as 10 to 30 time intervals, during which the speed is considered constant. In reality, the robot's speed will be reselected at each time interval, and most positions on the predicted trajectory will not actually be reached, so the method of calculating the distance to the obstacle in formula (1) is too conservative. In addition, the acceleration terms a and α in formula (1) are selected from the acceleration values ​​of the robot during normal driving. This constraint is based on safety considerations and requires the robot to stop when encountering an obstacle, so it is more reasonable to use the acceleration when braking. The acceleration during braking is often much greater than the value during normal driving, so the value of the acceleration in formula (1) is also too conservative. If the terms a and α in (1) are replaced with larger braking accelerations, the constraint is almost useless.

[0085] In this embodiment, the navigation function is the heading(v,ω) function, which calculates the deviation angle by predicting the position of the end of the trajectory; the correction is to change the reference position for calculating the deviation angle to the position after several time intervals on the predicted trajectory, and to determine the distance d h The deviation angle is calculated by combining the robot linear velocity prediction position after the time interval. The time interval is the number n Δt,h for:

[0086] n Δt,h =fix(d h / v) (2)

[0087] The final function form is as follows:

[0088] heading′(v,ω)=-N[|θ(r nΔt,h ,G)-θ(r nΔt,h )|]; (3)

[0089] Among them, θ(r nΔt,h ,G) is n Δt,h The angle at which the robot points to the target after a time interval; θ(r nΔt,h ) is the robot orientation at that position; N represents normalization.

[0090] In this embodiment, the obstacle avoidance function is the obdist(v, ω) function, which adds a function to discard dangerous trajectories. When calculating the shortest distance to an obstacle, only the position from the starting point of the trajectory to a certain time interval is considered, and an upper limit on the obstacle distance is set.

[0091] In addition to the function of the original evaluation trajectory, the modified obdist(v,ω) function will also discard those trajectories whose shortest distance from the obstacle to the position after nΔt,o time intervals is less than the safety distance. Its form is:

[0092]

[0093] Among them, d nΔt,o is the starting point to n on the trajectory Δt,o The shortest distance between the robot's position and the obstacle in the time interval; d p is the farthest obstacle considered by the function, and R is the robot's safety radius.

[0094] The velocity function is velocity(v,ω). This function makes the robot move quickly. In practice, the original function score is modified because it does not take angular velocity into account, which can cause trajectory oscillation or distortion.

[0095] Generally speaking, angular velocity fluctuations can easily lead to oscillations, and excessive angular velocity can easily lead to distortion. Considering that a large angular velocity does not affect the safety of the robot when it is traveling at low speed, a scoring mechanism for angular velocity is added as follows (5):

[0096] S ω =p1|2*ω t-1 -ω t -ω t-2 |-N(p2ω t v). (5)

[0097] Where: ω Angular velocity score.ω t-2 ,ω t-1 and ω t They are the robot's angular velocity at the previous two moments, the previous moment, and the current moment respectively. p1 and p2 are coefficients, and the linear velocity term is added to comprehensively consider the relationship between the two velocities.

[0098] A new function, osc(v, ω), has been added. Traditional DWA is a purely local path planning algorithm, prone to falling into local optima when the global environment is unknown. To address this, this section adds the osc(v, ω) function, which helps the robot escape local optima (such as those surrounding obstacles), enhancing DWA's global search capabilities in unknown environments. This function evaluates the similarity between the predicted trajectory and the historical trajectory, with a weight of w4. The global map is rasterized, with each grid cell assigned a cost value that records the proximity of the historical trajectory to that grid cell. The higher the cost value of a grid cell, the more likely the robot is to wander around it.

[0099] Establish the coordinate system Ω with P as the origin P ,d ij is the distance between the grid with coordinates (i, j) and the origin P; w ij is the increment of the cost value of grid (i, j) after time t, calculated by formula (6). Figure 2 As shown in the figure, it is a schematic diagram of the grid cost increment calculation method.

[0100]

[0101] Among them, p v It is a balancing term for the robot's speed, used to avoid a surge in the cost value of the surrounding grid when the robot is traveling at low speed.

[0102] The evaluation criterion of the osc(v,ω) function is the total cost value of all grids swept by the predicted trajectory. The lower the total cost value, the higher the score.

[0103] When the robot walks into an encircling obstacle due to the lack of a global map, this function can help the robot escape from the local optimum instead of circling inside the obstacle. The newly added osc(v,ω) function is as follows:

[0104]

[0105] Among them,∑ tra occupied is the sum of all grid cost values ​​covered by the predicted trajectory;

[0106] Traditional DWAs have low efficiency when approaching a target, especially when surrounded by obstacles. A new goal(v,ω) function has been added to enhance the robot's ability to navigate to the target. This function evaluates how well a predicted trajectory approaches the target, calculating the closest distance to the target for each point in the trajectory. The smaller the shortest distance, the higher the score, with a weight of ω5. This helps trajectory 1 achieve a higher score, enabling the robot to steer toward the target in a timely manner.

[0107] While the heading(v,ω) function also performs navigation, it's important to retain both functions because they have different purposes and application scenarios. Although the reference position in the heading(v,ω) function is corrected, it's still possible for it to exceed the target, especially when the robot is very close to the target, which can cause the function to misjudge. In this case, goal(v,ω) can help the robot steer toward the target in a timely manner. On the other hand, when the robot is still far from the target, the shortest distances from all predicted trajectories to the target are large and similar. In this case, the goal(v,ω) function has little discriminative power (hence, it is designed to only take effect when the robot is less than 2 meters from the target), and the robot relies primarily on the heading(v,ω) function for navigation. Overall, these two functions help the robot reach the target from both the distance and heading perspectives, avoiding the limitations of a single function and enhancing the robot's navigation capabilities.

[0108] The newly added goal(v,ω) function is in the following form (8):

[0109]

[0110] Among them, dist min(tra,G) is the shortest distance from the predicted trajectory to the target; dist(r,G) is the distance from the robot to the target; G P The effective range of this function is 2m.

[0111] In this embodiment, after modifying and adding each sub-evaluation function, the evaluation function of the improved DWA algorithm is obtained, and the formula is:

[0112] C(v,ω)=ω1·heading′(v,ω)+ω2·obdist′(v,ω)+ω3·velocity′(v,ω)+ω4·osc(v,ω)+ω5·goal(v,ω).

[0113] In this embodiment, the framework optimization module is integrated into the move_base node in the ROS system, directly receives the global target point and outputs the speed command to the motor control node.

[0114] The traditional method and the method of the present invention were compared. Figure 3 and Figure 4 By comparison, it can be seen that the trajectory of the traditional method is very oscillating and takes a long time, while the trajectory of the method of the present invention is obviously smoother and maintains high-speed motion throughout the process, which significantly improves the trajectory quality. Figure 5 Comparing (a) and (b), it can be seen that the trajectory of the traditional algorithm falls into the local optimal area and cannot escape, while the trajectory of the method of the present invention can eventually leave the obstacle area and reach the end point, thereby improving the navigation success rate.

[0115] A comparative experiment was conducted between the existing method and the method of the present invention. Figure 6 (a), (c), (e) and (g) are the results of the existing method, where the number is marked as γ d The trajectory is the optimal trajectory. Figure 6 (b), (d), (f), and (h) are the results of using the proposed method. The corresponding indicators are shown in Table 1. Comparison shows that the proposed method achieves similar safety distance and trajectory length as the existing method, but significantly reduces trajectory time, indicating that the proposed method can more efficiently traverse dense obstacles.

[0116] The proposed improved DWA was verified in a real environment. The experiment used an XQ-4Pro robot, which was configured with an i7-4500U 1.8GHz CPU and ran on the ROS (Robot Operation System) system. The robot was equipped with sensors such as a gyroscope, an odometer, and a lidar. In the experiment, a workstation was used to remotely log in to the robot, and instructions were issued and navigation information was monitored through the Rviz software in the workstation. In ROS, autonomous navigation tasks involve multiple nodes, which can be divided into four categories according to their functions: sensor nodes, positioning nodes, path planning nodes (named move_base), and motor control nodes. The algorithm of the present invention is implemented in move_base based on C++.

[0117] Figure 7 This is a flowchart of the algorithm of the present invention. The global reference path is obtained by the A* global planner and then tracked by the DWA local planner. When the original reference path is not feasible, it is necessary to call the global planner again to find a new path based on the detected obstacles, which has low online efficiency in an environment with many unknown obstacles. The present invention takes into account the unknown environment and improves the navigation framework so that the algorithm directly follows the global target point, eliminating the dependence on the global map and reference path, and publishing the speed of the optimal trajectory to the motor controller for execution, thereby ensuring the real-time performance of the algorithm, and being able to achieve navigation and effectively handle unknown obstacles without the need for a reference path.

[0118] The technical solution of the second aspect: an experimental method for a mobile robot navigation framework based on an improved DWA in an unknown environment, including any of the above, comprising the following steps:

[0119] Step S10: Experimental platform construction:

[0120] Configure a mobile robot platform equipped with lidar, IMU and odometer sensors, and install the ROS operating system;

[0121] Step S20: unknown environment construction:

[0122] Setting up static unknown obstacles and / or dynamic moving obstacles in the experimental field;

[0123] In this embodiment, the static obstacles in step S20 constitute a local optimal trap, and the dynamic obstacles are regularly moving pedestrian models.

[0124] Step S30: Navigation framework deployment:

[0125] Load the improved navigation framework into the robot's main control computer and load the original map without any obstacles;

[0126] Step S40: Target-driven navigation:

[0127] Publish the global target point through Rviz software and start the autonomous operation of the navigation framework;

[0128] Step S50: Dynamic obstacle avoidance verification:

[0129] When a dynamic obstacle intrudes into the robot's path, the real-time obstacle avoidance trajectory is recorded;

[0130] Step S60: Performance evaluation:

[0131] Collect trajectory length, time consumption, and minimum obstacle distance indicators and compare them with traditional DWA results.

[0132] In this embodiment, step S60 verifies the global search capability in the following manner:

[0133] When the robot enters an encircled obstacle area, detect whether it escapes within the preset time;

[0134] Statistics on navigation success rate in unknown environments.

[0135] Two sets of physical experiments were conducted. Experiment 1 considered an unknown static environment. To reflect the unknown nature of obstacles, the original map loaded in the navigation framework was barrier-free, e.g. Figure 8 (a), while the actual experimental scenario includes multiple unknown obstacles, such as Figure 8 (b) shows the distribution of these obstacles. These obstacles significantly affect the robot's direct path to the goal and form some small local optimal regions. The robot starts from the starting point and moves based on the global goal position and the detectable surrounding obstacles.

[0136] Figure 9 It is the Rviz software interface and the actual position of the robot during the experiment. Figure 9 The reference path is obtained by A*, which only considers the wall obstacles in the original map. Figure 9 is the final navigation trajectory. It can be seen that the robot can flexibly move among obstacles and successfully reach the target. The resulting trajectory is significantly different from the reference trajectory, verifying the feasibility of the proposed navigation algorithm in an unknown static environment.

[0137] The scenario in Experiment 2 contains a pedestrian with unknown motion trend as a dynamic obstacle. The present invention effectively improves the DWA mechanism and can also handle scenarios with simple dynamic obstacles. Figure 10 The original map and the actual test scene are obstacle-free. In the experiment, a group of people moved back and forth in the scene, and their trajectory interfered with the robot's direct path to the target twice.

[0138] Figure 11 This is the Rviz software interface and the actual positions of the robot and pedestrian during the experiment. The sub-graphs are arranged in chronological order. Figure 11In (a), the reference path points directly to the target. It can be seen that when the pedestrian obstructs the robot's movement, the robot can turn to avoid it, such as Figure 11 (b). When the pedestrian moves away from the robot, the robot will return to the target direction, e.g. Figure 11 (c). Figure 11 (d) is the final navigation trajectory. It can be seen that the robot can autonomously avoid the pedestrian and successfully reach the target. The resulting trajectory is significantly different from the reference trajectory, verifying the feasibility of the proposed navigation algorithm in a simple unknown dynamic environment.

[0139] Although the embodiments of the present invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and variations may be made to these embodiments without departing from the principles and spirit of the present invention. The scope of the present invention is defined by the appended claims and their equivalents, and all should be included within the scope of protection of the present invention.

Claims

1. A mobile robot navigation framework based on improved DWA in unknown environments, characterized by: include: An environmental perception module, which is used to obtain obstacle information in an unknown environment in real time through a laser radar; An improved DWA planning module, wherein the improved DWA planning module includes a dynamic window generation unit and a multi-factor evaluation unit; The dynamic window generation unit is used to generate a speed window based on the robot's mechanical constraints and remove traditional obstacle constraints; The multi-factor evaluation unit includes a navigation function, an obstacle avoidance function, a speed function, an escape function, and a target approach function; The navigation function is used to calculate the target deviation angle based on the position after the distance is moved on the predicted trajectory; The obstacle avoidance function is used to calculate the shortest distance from the starting point of the trajectory to the obstacle within a preset time interval and discard trajectories below the safety threshold; The speed function is used to add an angular velocity smoothing scoring item to suppress trajectory oscillation; The escape function is used to record the historical trajectory cost through a rasterized map and evaluate the similarity between the predicted trajectory and the historical trajectory; The target approach function is used to calculate the shortest distance between the trajectory point and the target; An execution module, configured to send the optimal speed instruction output by the evaluation function to a motor controller; The framework optimization module is used to cancel the dependence on the global map and the reference path and directly track the global target point.

2. The mobile robot navigation framework based on improved DWA in an unknown environment according to claim 1, characterized in that: The current state of the robot includes the current position, direction angle, linear velocity and angular velocity of the robot; The motion parameters include the robot's maximum linear velocity, maximum angular velocity, acceleration and angular acceleration.

3. The mobile robot navigation framework based on improved DWA in an unknown environment according to claim 1, characterized in that: There are mechanical and obstacle constraints on the velocity window; the robot stops before hitting the obstacle: Where obs(v,ω) is the shortest distance from the predicted trajectory to the obstacle; p is the safety factor.

4. The mobile robot navigation framework based on improved DWA in an unknown environment according to claim 1, characterized in that: The navigation function is the heading(v,ω) function, which calculates the deviation angle by predicting the position of the end of the trajectory; the correction is to change the reference position for calculating the deviation angle to the position after several time intervals on the predicted trajectory, and to determine the distance d h The deviation angle is calculated by combining the robot linear velocity prediction position after the time interval. The time interval is the number n Δt,h for: n Δt,h =fix(d h / v) The final function form is as follows: heading′(v,ω)=-N[|θ(r nΔt,h ,G)-θ(r nΔt,h )|]; Among them, θ(r nΔt,h ,G) is n Δt,h The angle at which the robot points to the target after a time interval; θ(r nΔt,h ) is the robot orientation at that position; N represents normalization.

5. The mobile robot navigation framework based on improved DWA in an unknown environment according to claim 1, characterized in that: The obstacle avoidance function is the obdist(v, ω) function, which adds a function to discard dangerous trajectories. When calculating the shortest distance to an obstacle, only the position from the starting point of the trajectory to a certain time interval is considered, and an upper limit on the obstacle distance is set. The modified obdist(v,ω) function not only has the function of evaluating the original trajectory, but also discards the trajectory from the starting point to n Δt,o The trajectories whose shortest distance from the obstacle after the time interval is less than the safety distance are in the form of: Among them, d nΔt,o is the starting point to n on the trajectory Δt,o The shortest distance between the robot's position and the obstacle in the time interval; d p is the farthest obstacle considered by the function; R is the robot's safety radius.

6. The mobile robot navigation framework based on improved DWA in an unknown environment according to claim 1, characterized in that: After modifying and adding each sub-evaluation function, the evaluation function of the improved DWA algorithm is obtained, and the formula is: C(v,ω)=ω1·heading′(v,ω)+ω2·ovdist′(v,ω)+ω3·velocity′(v,ω)+ω4·osc(v,ω)+ω5·goal(v,ω).

7. The mobile robot navigation framework based on improved DWA in an unknown environment according to claim 1, characterized in that: The framework optimization module is integrated into the move_base node in the ROS system, directly receives the global target point and outputs the speed command to the motor control node.

8. An experimental method comprising the mobile robot navigation framework based on the improved DWA in an unknown environment as claimed in any one of claims 1 to 7, characterized in that: The following steps are involved: Step S10: Experimental platform construction: Configure a mobile robot platform equipped with lidar, IMU and odometer sensors, and install the ROS operating system; Step S20: unknown environment construction: Setting up static unknown obstacles and / or dynamic moving obstacles in the experimental field; Step S30: Navigation framework deployment: Load the improved navigation framework into the robot's main control computer and load the original map without any obstacles; Step S40: Target-driven navigation: Publish the global target point through Rviz software and start the autonomous operation of the navigation framework; Step S50: Dynamic obstacle avoidance verification: When a dynamic obstacle intrudes into the robot's path, the real-time obstacle avoidance trajectory is recorded; Step S60: Performance evaluation: Collect trajectory length, time consumption, and minimum obstacle distance indicators and compare them with traditional DWA results.

9. The experimental method of the mobile robot navigation framework based on the improved DWA in an unknown environment according to claim 8 is characterized in that: In step S20 , the static obstacles constitute a local optimal trap, and the dynamic obstacles are regularly moving pedestrian models.

10. The experimental method of the mobile robot navigation framework based on the improved DWA in an unknown environment according to claim 8, characterized in that: Step S60 verifies the global search capability by: When the robot enters an encircled obstacle area, it detects whether it escapes within the preset time; Statistics on the navigation success rate in unknown environments.

Images