Obstacle avoidance method and system for unmanned logistics vehicles

Abstract

The present invention provides an obstacle avoidance method and system for unmanned logistics vehicles, belonging to the technical field of unmanned logistics vehicle path planning. The method of the present invention determines the timing of obstacle avoidance based on the drivable area, multi-frame obstacle data and collision detection, and uses static reference lines to complete obstacle avoidance trajectory planning. The present invention judges the timing of obstacle avoidance based on the safe and efficient delivery goals of unmanned logistics vehicles and the operating characteristics of fixed route driving, avoids incorrect obstacle avoidance and frequent obstacle avoidance that makes the vehicle unsafe, and generates an optimal trajectory to achieve obstacle avoidance, thereby ensuring the operational safety of unmanned logistics vehicles while improving transportation efficiency.

Description

Technical Field

[0001] The present invention belongs to the technical field of unmanned logistics vehicle path planning, and specifically relates to an unmanned logistics vehicle obstacle avoidance method and system. Background Art

[0002] With the advancement of autonomous driving technology, the unmanned logistics industry in factories has rapidly developed due to its closed, fixed, and specialized working environments. As a key element in the unmanned logistics chain, unmanned logistics vehicles can quickly achieve large-scale operations in factory material distribution and product transportation. This improves factory cargo delivery efficiency, significantly reduces labor costs, and mitigates safety risks, effectively promoting the transformation of traditional factories into intelligent ones.

[0003] Typical system architecture of unmanned logistics vehicles is as follows: Figure 1 As shown in the figure, the system mainly includes an environmental perception module, a decision-making and planning module, and a motion control module. The decision-making and planning module further includes a behavior planning submodule and a path planning submodule. The path planning submodule is responsible for generating a local trajectory that complies with driving regulations, meets kinematic constraints, and is safe and collision-free, based on the vehicle's positioning information, local environment information, and static reference line information.

[0004] Factory environments often face scenarios where obstacles intrude into the set safety distance, forcing unmanned logistics vehicles to stop for long periods of time to avoid them. However, neither waiting on the spot nor manual takeover can achieve the goal of "reducing labor costs and improving transportation efficiency in unmanned logistics delivery." Therefore, when obstacles require extended avoidance time, a circumvention method needs to be designed to ensure the normal operation of unmanned logistics transportation.

[0005] For enclosed areas like factory areas, the Lanelet framework is typically used to extract information such as lanes and regional speed limits from high-precision maps before performing local planning. The current common approach to obstacle avoidance is to select a path only if the offset between the obstacle avoidance path data and the prior global path data is less than a preset offset threshold. However, this threshold determination method can lead to the vehicle swaying from side to side at certain moments, causing hesitation and wasting time.

[0006] Therefore, there is an urgent need for an obstacle avoidance method for unmanned logistics vehicles in factory areas that is suitable for static obstacles, which can ensure safety and improve efficiency according to the actual usage scenarios of unmanned logistics vehicles in factory areas. Summary of the Invention

[0007] The purpose of the present invention is to solve the problems existing in the above-mentioned prior art and provide an unmanned logistics vehicle obstacle avoidance method and system to avoid the unmanned logistics vehicle from having to avoid obstacles for a long time during operation and affecting the operating efficiency.

[0008] The present invention is achieved through the following technical solutions:

[0009] The first aspect of the present invention provides an obstacle avoidance method for an unmanned logistics vehicle. The method determines the timing of obstacle avoidance based on the drivable area, multi-frame obstacle data and collision detection, and uses a static reference line to complete obstacle avoidance trajectory planning.

[0010] A further improvement of the present invention is:

[0011] The method comprises:

[0012] Step 1: Obtain lane boundary information and lane speed limit information from the high-precision map;

[0013] Step 2: Generate a static reference line based on the prior global path and use the static reference line as the main static reference line;

[0014] Step 3: When the lateral and longitudinal distances between the obstacle in front and the vehicle are both less than the set safety distance, calculate the distance D that the main static reference line needs to move to avoid the obstacle. m , and then translate the main static reference line as a whole by D m Generate new static reference lines;

[0015] Step 4: Determine whether the obstacle can be circumvented based on the obstacle avoidance conditions. If so, proceed to step 5; if not, proceed to step 7.

[0016] Step 5: Complete path planning based on the new static reference line, generate a local trajectory, and control the vehicle to drive along the local trajectory to avoid obstacles;

[0017] Step 6: As the vehicle moves along the local trajectory to avoid obstacles, it performs a collision check on the main static reference line. If the collision check passes, it immediately returns to the main static reference line and proceeds to step 8.

[0018] Step 7: The vehicle stops at the current position to avoid obstacles and returns to step 4;

[0019] Step 8: Complete local trajectory planning based on the main static reference line.

[0020] A further improvement of the present invention is:

[0021] The operation of generating a static reference line based on the prior global path in the second step includes:

[0022] The data 30 meters before and 15 meters after the vehicle's current position are intercepted on the prior global path as the static reference line.

[0023] A further improvement of the present invention is:

[0024] The operation of calculating the distance Dm that the main static reference line needs to move to avoid the obstacle in the third step includes:

[0025] (31) Obtain the coordinates of each vertex of the obstacle. In the vehicle body coordinate system, take the horizontal coordinate of the leftmost vertex of the obstacle as the offset;

[0026] (32) The translation distance D of the main static reference line is calculated using the following formula m :

[0027] D m =E w / 2+d w -d

[0028] Among them, E w is the width of the vehicle; d w is the preset anti-collision distance; d is the horizontal coordinate of the leftmost vertex of the obstacle.

[0029] A further improvement of the present invention is:

[0030] In the third step, the main static reference line is translated as D m The operations for generating new static reference lines include:

[0031] Shift the main static reference line to the left as a whole D m Creates a new static guide.

[0032] A further improvement of the present invention is:

[0033] The obstacle avoidance conditions in the fourth step are as follows:

[0034] Condition 1: The new static reference line does not exceed the current lane;

[0035] Condition 2: The distance of the obstacle that intrudes into the safety distance relative to the vehicle remains unchanged in multiple frames of data;

[0036] Condition 3: The collision detection based on the new static reference line passes;

[0037] Obstacles can be circumvented only when the above three conditions are met at the same time.

[0038] A further improvement of the present invention is:

[0039] The fourth step includes:

[0040] (41) Based on the lane information provided by the high-precision map, the Lanelet framework is used to determine whether the new static reference line is still in the current lane. If so, condition 1 is satisfied.

[0041] (42) If the obstacle that has intruded into the safety distance remains a static obstacle within 10 seconds and the changes in its lateral and longitudinal distances relative to the vehicle are both less than the set error values, then condition 2 is determined to be satisfied;

[0042] (43) If the new static reference line passes the collision detection, it is determined that condition three is satisfied.

[0043] A further improvement of the present invention is:

[0044] The collision detection operation includes:

[0045] (S1) First, use the AABB method to construct the convex polygon envelope of the vehicle;

[0046] (S2) traversing each trajectory point in the predicted trajectory of each obstacle and the new static reference line of the ego vehicle, checking whether there is overlap between the convex polygonal envelope of the ego vehicle and the convex polygonal envelope of the obstacle, and removing obstacles that do not overlap;

[0047] (S3) For overlapping obstacles, the OBB method is used to construct the convex polygonal envelope of the ego vehicle, that is, the OBB envelope of the ego vehicle, and accurate collision detection is performed based on the separating axis theorem.

[0048] A further improvement of the present invention is:

[0049] The operations of accurate collision detection based on the separating axis theorem include:

[0050] (S31) Projecting the OBB envelope of the ego vehicle and the OBB envelope of the obstacle in four directions, namely, the longitudinal axis of the ego vehicle, the transverse axis of the ego vehicle, the longitudinal axis of the obstacle, and the transverse axis of the obstacle, respectively, to obtain: the projection length ap of the vector formed by the center line connecting the OBB envelope of the ego vehicle and the OBB envelope of the obstacle on the projection line in the transverse axis direction of the ego vehicle, half the projection length bp of the OBB envelope of the ego vehicle on the projection line in the transverse axis direction of the ego vehicle, and half the projection length cp of the OBB envelope of the obstacle on the projection line in the transverse axis direction of the ego vehicle;

[0051] (S32) determining in turn whether ap in each direction is less than or equal to bp+cp, if yes, determining that the direction will collide, if no, determining that the direction will not collide;

[0052] (S33) If the determination results in all four directions are that there is no collision, then it is determined that the new static reference line passes the collision detection; otherwise, it is determined that the new static reference line does not pass the collision detection.

[0053] A second aspect of the present invention provides an obstacle avoidance system for an unmanned logistics vehicle, the system comprising:

[0054] Map processing module, used to obtain lane boundary information and lane speed limit information from high-precision maps;

[0055] The environmental perception module is used to obtain information about the vehicle's surroundings, identify obstacles, and output the dynamic and static classification of the obstacle, the coordinates of the obstacle's center point and each vertex in the vehicle's coordinate system, the obstacle's convex polygon envelope, and the predicted trajectory of the dynamic obstacle;

[0056] The collision detection module is used to construct the convex polygon envelope of the ego vehicle and determine whether the ego vehicle has a collision risk;

[0057] The moving distance calculation module is used to calculate the distance required to move to bypass the static reference line of the obstacle based on the coordinates of the obstacle;

[0058] A reference line generation module, used to generate static reference lines;

[0059] The obstacle avoidance decision module is used to determine whether obstacles can be avoided based on the drivable area, multi-frame obstacle data, and collision detection;

[0060] Path planning module, used to generate vehicle motion trajectory.

[0061] Compared with the existing technology, the beneficial effects of the present invention are: the present invention judges the timing of obstacle avoidance based on the safe and efficient delivery goals of unmanned logistics vehicles and the operating characteristics of fixed route driving, avoids incorrect obstacle avoidance and frequent obstacle avoidance that lead to vehicle safety, and generates the optimal trajectory to achieve obstacle avoidance, thereby ensuring the safety of unmanned logistics vehicles while improving transportation efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] Figure 1 Existing unmanned logistics vehicle system architecture;

[0063] Figure 2 Schematic diagram of OBB envelope and AABB envelope;

[0064] Figure 3 Schematic diagram of eliminating non-collision obstacles;

[0065] Figure 4 Projection along the vehicle's transverse axis;

[0066] Figure 5 A structural diagram of the system of the present invention;

[0067] Figure 6 A block diagram of the steps of the method of the present invention. DETAILED DESCRIPTION

[0068] The present invention is further described in detail below with reference to the accompanying drawings:

[0069] Unmanned logistics vehicles in factory environments always aim to complete material distribution in the factory safely and efficiently. Therefore, the unmanned logistics vehicle system uses the prior global path as a reference to plan the vehicle to travel on a fixed route, without actively overtaking or changing lanes. However, in order to solve scenarios such as truck unloading that occupy the unmanned logistics route for a long time, it is necessary to design an obstacle avoidance strategy to ensure the normal operation of unmanned logistics transportation.

[0070] Existing obstacle avoidance methods based on threshold judgment can cause logistics vehicles to hesitate and are not very reliable. This clearly does not meet the requirements for safe and efficient operation of unmanned logistics vehicles. Unmanned logistics vehicles typically travel on fixed routes, so the difficulty of obstacle avoidance methods lies in determining the timing of obstacle avoidance and planning the obstacle avoidance trajectory.

[0071] In order to solve the above problems, the present invention provides a new obstacle avoidance method for unmanned logistics vehicles in factory areas. The method mainly determines the timing of obstacle avoidance based on three restrictions: drivable area, multi-frame obstacle data and collision detection, and completes obstacle avoidance trajectory planning through static reference lines to achieve a safe and efficient obstacle avoidance strategy.

[0072] The system of the present invention comprises:

[0073] 1. Map processing module: The map processing module uses an existing algorithm based on the Lanelet framework to obtain lane boundary information and lane speed limit information from the high-precision map.

[0074] 2. Environmental perception module. The environmental perception module includes but is not limited to various existing devices for obtaining environmental information around the vehicle body, such as lidar and cameras. It can obtain and identify environmental obstacles, including but not limited to pedestrians, vehicles and other targets, and output information such as the dynamic and static classification of the obstacle, the coordinates of the center point and each vertex of the obstacle in the vehicle body coordinate system, the convex polygon envelope of the obstacle, and the predicted trajectory of dynamic obstacles.

[0075] 3. Collision detection module, which is used to construct a convex polygon envelope of the ego vehicle, traverse each trajectory point of the predicted trajectory of each obstacle, and traverse each trajectory point of the ego vehicle's planned trajectory to check whether there is any overlap between the convex polygon envelope of the ego vehicle and the convex polygon envelope of the obstacle, to determine whether there is a collision risk between the ego vehicle and the obstacle.

[0076] 4. A moving distance calculation module, which calculates the distance required to move to bypass the static reference line of the obstacle based on the coordinates of the obstacle.

[0077] 5. A reference line generation module, which generates a static reference line for path planning by intercepting data 30 meters before and 15 meters after the vehicle's current position based on the prior global path.

[0078] 6. Obstacle Avoidance Decision Module: The obstacle avoidance decision module decides whether to avoid obstacles and whether to return to the main road based on three constraints: the drivable area, multi-frame obstacle data, and collision detection.

[0079] The present invention adds the above-mentioned collision detection module, moving distance calculation module, reference line generation module and obstacle avoidance decision module to the "behavior planning submodule" in the existing system. Figure 5 shown.

[0080] 7. Path planning module, which generates a vehicle motion trajectory that complies with driving regulations, meets kinematic constraints, and is safe and collision-free based on static reference lines, vehicle positioning information, and environmental perception information.

[0081] In order to achieve the above purpose, Figure 6 As shown, the present invention includes the following steps:

[0082] Step 1: Obtain lane boundary information and lane speed limit information from the high-precision map based on the Lanelet framework. This step is completed by the existing map processing module and implemented using existing algorithms, so I will not elaborate on it here.

[0083] Step 2: Based on the prior global path, data is intercepted 30 meters before and 15 meters after the vehicle's current position to generate a static reference line, which is used as the main static reference line. This step is completed by the reference line generation module.

[0084] The a priori global path is a pre-recorded path for an unmanned logistics vehicle, consisting of multiple discrete a priori trajectory points spaced 1 meter apart. The present invention uses the vehicle's positioning information to obtain the a priori trajectory point closest to the current position. The data for the first 30 points and the last 15 points of this a priori trajectory point are then taken and used to generate a primary static reference line using spline interpolation. The primary static reference line is also composed of discrete trajectory points.

[0085] Step 3: After the environment perception module detects that the lateral and longitudinal distances between the obstacle in front and the vehicle are both less than the set safety distance, the moving distance calculation module obtains the obstacle coordinates from the information output by the environment perception module and calculates the distance D that the main static reference line needs to move to avoid the obstacle based on the obstacle coordinates. m , then move the main static reference line to the left as a whole D m Generate a new static reference line. The third step is completed by the moving distance calculation module.

[0086] Specifically, calculate the distance D that the main static reference line needs to move to avoid obstacles m The operations include:

[0087] (31) The moving distance calculation module obtains the coordinates of each vertex of the obstacle from the information output by the environment perception module. In the vehicle body coordinate system, the horizontal coordinate of the leftmost vertex of the obstacle is taken as the offset;

[0088] (32) Main static reference line translation distance D m The calculation process is:

[0089] D m =E w / 2+d w -d

[0090] Among them, D m E is the translation distance of the main static reference line; w is the width of the vehicle; d w is the preset anti-collision distance; d is the horizontal coordinate of the leftmost vertex of the obstacle.

[0091] In addition, considering the actual traffic rules and the principles of safety and efficiency, the present invention stipulates that unmanned logistics vehicles can only bypass obstacles from the left. Therefore, in the third step, the main static reference line is shifted to the left as a whole by D m Creates a new static guide.

[0092] The fourth step is to determine whether the obstacle can be avoided based on the obstacle avoidance conditions. If it can be avoided, the process proceeds to the fifth step; if it cannot be avoided, the process proceeds to the seventh step. This step is completed by the obstacle avoidance decision module.

[0093] In the fourth step, there are three obstacle avoidance conditions:

[0094] Condition 1: The new static reference line does not exceed the current lane;

[0095] Condition 2: The distance of the obstacle that intrudes into the safety distance relative to the vehicle remains unchanged in multiple frames of data;

[0096] Condition 3: The collision detection based on the new static reference line passes.

[0097] Obstacle avoidance can only be achieved when all three of the above conditions are met simultaneously. This means that the new static reference line is used for path planning in step 5. If any of the conditions is not met, the process proceeds to step 7, where the vehicle remains at its current position and stops to avoid the obstacle.

[0098] The specific process of judging whether the above conditions are met is as follows:

[0099] (41) Based on the lane information provided by the high-precision map, the Lanelet framework is used to determine whether the new static reference line is still in the current lane (the existing interface function of the Lanelet framework can be used to determine whether a line is in a certain lane. This is a mature technology and will not be described here). If it is still in the current lane, it is determined that the first condition for circumventing the obstacle is met.

[0100] (42) In order to eliminate the possibility of misidentification of single-frame data of the environmental perception module and misclassification of dynamic and static obstacles, it is determined whether the obstacles that invade the safe distance in the obstacle information data output by the environmental perception module remain static obstacles within 10 seconds and whether the changes in their lateral and longitudinal distances relative to the vehicle are less than the set error values. If so, it is determined that the second condition for obstacle avoidance is met.

[0101] (43) If the new static reference line passes the collision detection, it is determined that the obstacle avoidance condition 3 is met. The collision detection is completed by the collision detection module.

[0102] Furthermore, the collision detection operation includes:

[0103] (S1) In order to accurately describe the obstacle outline, the environment perception module uses the OBB method to construct the obstacle convex polygon envelope. However, judging whether two OBB envelopes collide will take a long time and consume a lot of computing power. Therefore, the present invention first uses the AABB method to construct the ego-vehicle convex polygon envelope for rough collision detection, quickly eliminating obstacles that are obviously non-intersecting, which is time-saving and saves computing power. OBB and AABB are two common envelope boxes, such as Figure 2 shown.

[0104] (S2) Based on the predicted obstacle trajectory output by the environment perception module, traverse each trajectory point in the predicted trajectory of each obstacle and the new static reference line of the ego vehicle, check whether there is overlap between convex polygons, that is, check whether there is overlap between the AABB envelope of the ego vehicle and the OBB envelope of the obstacle, and quickly eliminate non-collision obstacles (that is, place the AABB envelope of the ego vehicle and the OBB envelope of the obstacle at each trajectory point, check whether the two envelopes overlap, and eliminate those that do not overlap), such as Figure 3 As shown in the figure, the black dots are trajectory points on the new static reference line. The black dots are the AABB envelope of the ego vehicle, and 1 to 6 are the OBB envelopes of the obstacles. Only the envelope of obstacle 5 overlaps with the envelope of the ego vehicle, so obstacles 1, 2, 3, 4, and 6 can be directly removed.

[0105] (S3) For overlapping envelopes (i.e. Figure 3 The OBB method is used to construct the convex polygon envelope of the ego vehicle and perform accurate collision detection based on the separating axis theorem. If all four sides pass the detection, it is considered to have passed the collision detection.

[0106] Furthermore, the specific process of accurate collision detection based on the separating axis theorem is as follows:

[0107] (S31) Project the OBB envelope of the vehicle and the OBB envelope of the obstacle in four directions: the longitudinal axis of the vehicle, the transverse axis of the vehicle (i.e., the long side and short side of the OBB envelope of the vehicle), the longitudinal axis of the obstacle, and the transverse axis of the obstacle (i.e., the long side and short side of the OBB envelope of the obstacle). Projection examples are as follows: Figure 4 As shown in the figure, ap is the projection length of the vector formed by the center line connecting the OBB envelope frame of the ego vehicle and the center line of the OBB envelope frame of the obstacle on the projection line of the ego vehicle's transverse axis, bp is half of the projection length of the OBB envelope frame of the ego vehicle on the projection line of the ego vehicle's transverse axis, and cp is half of the projection length of the OBB envelope frame of the obstacle on the projection line of the ego vehicle's transverse axis.

[0108] (S32) If all four directions satisfy the following conditions: the projection length of the vector formed by the center line connecting the OBB envelopes of the ego vehicle and the obstacle on the projection line is not greater than the sum of half the projection lengths of the OBB envelopes of the ego vehicle and the obstacle on the projection line, then it is considered that there will be a collision; otherwise, there will be no collision. That is, whether ap in each direction is less than or equal to bp+cp is determined in turn. If so, it is determined that there will be a collision in that direction; if not, it is determined that there will be no collision in that direction.

[0109] (S33) If the determination results in all four directions are that there is no collision, then it is determined that the new static reference line passes the collision detection; otherwise, it is determined that the new static reference line does not pass the collision detection.

[0110] Step 5: Complete path planning based on the new static reference line, generate a local trajectory, and control the vehicle to drive along the local trajectory to avoid obstacles. This step is completed by the path planning module and implemented using existing algorithms. The operations of step 5 are briefly described as follows:

[0111] (51) The current posture information of the vehicle is converted into the Frenet coordinate system based on the transverse and longitudinal directions of the new static reference line (the Frenet coordinate system is a coordinate system established by the tangent vector and normal vector of the new static reference line, longitudinal S: along the direction of the reference line; transverse l: normal to the reference line) as the planning starting point, and the initial state of the vehicle in the Frenet coordinate system is obtained, which mainly includes the transverse distance l, longitudinal distance S, displacement, velocity and acceleration.

[0112] (52) The longitudinal trajectory is actually the relationship between speed and time, that is, speed planning. Based on the set sampling duration and sampling interval, multiple sampling moments can be determined. At each sampling moment, the maximum and minimum vehicle speeds at that moment can be obtained based on the maximum acceleration and deceleration value and the maximum cruising speed of the vehicle. The maximum and minimum speeds are used as the planning end points. Based on the planning starting point and planning end point states, polynomial curve fitting is performed to obtain multiple longitudinal trajectories. Each longitudinal trajectory is a continuous curve.

[0113] (53) The transverse trajectory is the relationship between the transverse distance and the longitudinal distance in the Frenet coordinate system. Conventional point sampling is used. Based on the set transverse point distance and longitudinal sampling distance, multiple planning end points can be determined. Polynomial curve fitting is performed based on the starting and end points to obtain multiple transverse trajectories. Each transverse trajectory is a continuous curve.

[0114] (54) Feasibility screening is performed on a series of longitudinal trajectories, and each longitudinal trajectory is judged in turn to determine whether the speed meets the set vehicle speed range and whether the acceleration meets the set acceleration range. If yes, the longitudinal trajectory is considered to be a feasible longitudinal trajectory; if not, the longitudinal trajectory is deleted.

[0115] (55) Cost evaluation is performed on the lateral trajectory and the feasible longitudinal trajectory respectively, and the lateral trajectory and the feasible longitudinal trajectory are sorted from small to large according to the cost evaluation value. The cost evaluation value is the sum of each cost evaluation item multiplied by the corresponding weight. The lateral trajectory cost evaluation items include lateral offset and lateral comfort; the longitudinal trajectory cost evaluation items include centripetal acceleration, longitudinal comfort and collision possibility. These cost evaluations are all implemented using existing algorithms and will not be described in detail here.

[0116] (56) The lateral trajectory and longitudinal trajectory with the lowest cost evaluation are selected for synthesis (synthesis is performed using the existing method, which will not be repeated here), and a local trajectory with a timestamp in the Frenet coordinate system is obtained. The local trajectory is subjected to collision detection (the collision detection adopts the same method as the above steps (S1) to (S3), and the local trajectory is used instead of the new static reference line for collision detection). If the collision detection is passed, the local trajectory in the Frenet coordinate system is converted to the world coordinate system, and the vehicle travels along the local trajectory. If the collision detection of the local trajectory fails, the planning fails, and the seventh step is entered, that is, the vehicle remains at the current position and stops to avoid obstacles.

[0117] Step 6: As the vehicle travels along the local trajectory to avoid obstacles, it performs a collision check on the main static reference line (collision check uses the same method as steps (S1) to (S3) above, replacing the new static reference line with the main static reference line for collision check). If the collision check passes, it indicates that the target obstacle has been bypassed and no other obstacles have intruded into the safe distance, and the vehicle can return to the main road. That is, while traveling along the local trajectory, it continuously performs collision checks on the main static reference line. Once the collision check passes, it immediately returns to the main static reference line and proceeds to step 8. Steps 5 and 6 are completed by the path planning module.

[0118] Step 7: The vehicle stops at the current position to avoid obstacles and returns to step 4;

[0119] Step 8: Complete local trajectory planning based on the main static reference line. This is achieved using existing algorithms and will not be described in detail here.

[0120] The present invention judges the opportunity to avoid obstacles based on three limiting aspects: the drivable area, multi-frame dynamic and static classification of obstacles, and the judgment of the center point coordinates and each vertex coordinate data in the vehicle body coordinate system and collision detection, thereby reducing safety hazards during obstacle avoidance and improving obstacle avoidance safety.

[0121] The present invention uses a moving static reference line for path planning, and the output optimal trajectory can not only correctly bypass obstacles but also return to the main road, meeting the operational requirements of fixed-route driving of unmanned logistics vehicles in the factory area.

[0122] After actual vehicle verification, the method of the present invention can complete the obstacle avoidance timing decision-making that meets the actual operation needs of unmanned logistics vehicles in the factory area; and the method of the present invention can generate the optimal trajectory for unmanned logistics vehicles in the factory area to safely bypass static obstacles and return to the default fixed driving route.

[0123] The above technical solution is only one embodiment of the present invention. For those skilled in the art, it is easy to make various types of improvements or modifications based on the principles disclosed in the present invention, and it is not limited to the technical solution described in the above specific embodiments of the present invention. Therefore, the above description is only preferred and does not have a restrictive meaning.

Claims

1. A method for unmanned logistics vehicles to avoid obstacles, characterized by: The method determines the timing of obstacle avoidance based on the drivable area, multi-frame obstacle data, and collision detection, and uses static reference lines to complete obstacle avoidance trajectory planning. The collision detection operation includes: (S1) First, use the AABB method to construct the convex polygon envelope of the vehicle; (S2) traversing each trajectory point in the predicted trajectory of each obstacle and the new static reference line of the ego vehicle, checking whether there is overlap between the convex polygonal envelope of the ego vehicle and the convex polygonal envelope of the obstacle, and removing obstacles that do not overlap; (S3) For overlapping obstacles, the OBB method is used to construct the convex polygonal envelope of the ego vehicle, that is, the OBB envelope of the ego vehicle, and accurate collision detection is performed based on the separating axis theorem; The method comprises: Step 1: Obtain lane boundary information and lane speed limit information from the high-precision map; Step 2: Generate a static reference line based on the prior global path and use the static reference line as the main static reference line; Step 3: When the lateral and longitudinal distances between the obstacle in front and the vehicle are both less than the set safety distance, calculate the distance D that the main static reference line needs to move to avoid the obstacle. m , and then translate the main static reference line as a whole by D m Generate new static reference lines; Step 4: Determine whether the obstacle can be circumvented based on the obstacle avoidance conditions. If so, proceed to step 5; if not, proceed to step 7. The obstacle avoidance conditions are as follows: Condition 1: The new static reference line does not exceed the current lane; Condition 2: The distance of the obstacle that intrudes into the safety distance relative to the vehicle remains unchanged in multiple frames of data; Condition 3: The collision detection based on the new static reference line passes; Obstacles can be bypassed only when the above three conditions are met at the same time; Step 5: Complete path planning based on the new static reference line, generate a local trajectory, and control the vehicle to drive along the local trajectory to avoid obstacles; Step 6: As the vehicle moves along the local trajectory to avoid obstacles, it performs a collision check on the main static reference line. If the collision check passes, it immediately returns to the main static reference line and proceeds to step 8. Step 7: The vehicle stops at the current position to avoid obstacles and returns to step 4; Step 8: Complete local trajectory planning based on the main static reference line.

2. The obstacle avoidance method for an unmanned logistics vehicle according to claim 1, characterized in that: The operation of generating a static reference line based on the prior global path in the second step includes: intercepting data 30 meters before and 15 meters after the current position of the vehicle on the prior global path as the static reference line.

3. The obstacle avoidance method for an unmanned logistics vehicle according to claim 1, characterized in that: The calculation in the third step is the distance D that the main static reference line needs to move in order to avoid obstacles. m The operations include: (31) Obtain the coordinates of each vertex of the obstacle. In the vehicle body coordinate system, take the horizontal coordinate of the leftmost vertex of the obstacle as the offset; (32) The translation distance D of the main static reference line is calculated using the following formula m : D m =E w / 2+d w -d Among them, E w is the width of the vehicle; d w is the preset anti-collision distance; d is the horizontal coordinate of the leftmost vertex of the obstacle.

4. The obstacle avoidance method for an unmanned logistics vehicle according to claim 3, characterized in that: In the third step, the main static reference line is translated as D m The operations for generating a new static reference line include: Shift the main static reference line to the left as a whole D m Creates a new static guide.

5. The obstacle avoidance method for an unmanned logistics vehicle according to claim 1, characterized in that: The fourth step includes: (41) Based on the lane information provided by the high-precision map, the Lanelet framework is used to determine whether the new static reference line is still in the current lane. If so, condition 1 is determined to be satisfied; (42) If the obstacle that has intruded into the safety distance remains a static obstacle within 10 seconds and the changes in its lateral and longitudinal distances relative to the vehicle are both less than the set error values, then condition 2 is determined to be satisfied; (43) If the new static reference line passes the collision detection, it is determined that condition three is satisfied.

6. The obstacle avoidance method for an unmanned logistics vehicle according to claim 5, characterized in that: The operations of accurate collision detection based on the separating axis theorem include: (S31) Projecting the OBB envelope of the ego vehicle and the OBB envelope of the obstacle in four directions, namely, the longitudinal axis of the ego vehicle, the transverse axis of the ego vehicle, the longitudinal axis of the obstacle, and the transverse axis of the obstacle, respectively, to obtain: the projection length ap of the vector formed by the center line connecting the OBB envelope of the ego vehicle and the OBB envelope of the obstacle on the projection line in the transverse axis direction of the ego vehicle, half the projection length bp of the OBB envelope of the ego vehicle on the projection line in the transverse axis direction of the ego vehicle, and half the projection length cp of the OBB envelope of the obstacle on the projection line in the transverse axis direction of the ego vehicle; (S32) determining in turn whether ap in each direction is less than or equal to bp+cp, if yes, determining that the direction will collide, if no, determining that the direction will not collide; (S33) If the determination results in all four directions are that there is no collision, then it is determined that the new static reference line passes the collision detection; otherwise, it is determined that the new static reference line does not pass the collision detection.

7. An unmanned logistics vehicle obstacle avoidance system, characterized by: The system comprises: Map processing module, used to obtain lane boundary information and lane speed limit information from high-precision maps; The environmental perception module is used to obtain information about the vehicle's surroundings, identify obstacles, and output the dynamic and static classification of the obstacle, the coordinates of the obstacle's center point and each vertex in the vehicle's coordinate system, the obstacle's convex polygon envelope, and the predicted trajectory of the dynamic obstacle; The collision detection module is used to construct the convex polygon envelope of the ego vehicle and determine whether the ego vehicle has a collision risk; Collision detection operations include: (S1) First, use the AABB method to construct the convex polygon envelope of the vehicle; (S2) traversing each trajectory point in the predicted trajectory of each obstacle and the new static reference line of the ego vehicle, checking whether there is overlap between the convex polygonal envelope of the ego vehicle and the convex polygonal envelope of the obstacle, and removing obstacles that do not overlap; (S3) For overlapping obstacles, the OBB method is used to construct the convex polygonal envelope of the ego vehicle, that is, the OBB envelope of the ego vehicle, and accurate collision detection is performed based on the separating axis theorem; The moving distance calculation module is used to calculate the distance required to move to bypass the static reference line of the obstacle based on the coordinates of the obstacle; A reference line generation module, used to generate static reference lines; The obstacle avoidance decision module is used to determine whether obstacles can be avoided based on the drivable area, multi-frame obstacle data, and collision detection; Path planning module, used to generate vehicle motion trajectory.

Images