Method, device and equipment for real-time updating of dynamic obstacle in perception and planning coordination

Abstract

The application relates to a dynamic obstacle real-time updating method, device and equipment based on perception and planning cooperation, and relates to the technical field of computer vision. The method is executed by a computer device, perception data of a dynamic obstacle is acquired, perception data meeting preset conditions is written into a temporary layer of a local cost map of a robot navigation system, and a clearance operation is performed on the cost of the dynamic obstacle in the temporary layer at a decay rate higher than that of a regular obstacle layer in the local cost map. Through a shared timestamp synchronization mechanism, updating of the local cost map and recalculation of a robot local planner are triggered.

Description

Technical Field

[0001] This invention relates to the field of computer vision technology, and in particular to a method, apparatus, and device for real-time updating of dynamic obstacles through perception and planning coordination. Background Technology

[0002] In mobile robot navigation systems, the local cost map and the local planner together form the core of the robot's real-time obstacle avoidance and motion control. The local cost map, centered on the robot, integrates sensor observation data to construct a dynamic environment model. The local planner, based on this map, outputs velocity commands that meet kinematic constraints through velocity sampling and trajectory optimization, enabling the robot to smoothly avoid obstacles and safely pass through unknown or dynamic environments.

[0003] In existing technologies, typical Nav2 navigation frameworks typically use lidar or pseudo-laser layers as the primary source of information on dynamic obstacles, and achieve obstacle avoidance through periodic cost map updates and planner recalculation.

[0004] However, the timing mismatch between the sensor release frequency and the cost map update, expansion layer calculation, and planner control cycle in this architecture results in a significant link delay from the perception of dynamic obstacles to their avoidance, and the efficiency of real-time dynamic obstacle updates in the coordination of perception and planning is low. Summary of the Invention

[0005] This invention provides a method, apparatus, and device for real-time updating of dynamic obstacles through perception and planning coordination, which can improve the efficiency of real-time updating of dynamic obstacles through perception and planning coordination. The technical solution is as follows: On the one hand, a method for real-time updating of dynamic obstacles through perception and planning coordination is provided, the method including: Acquire perception data of dynamic obstacles; Perception data that meets preset conditions is written into a temporary layer of the local cost map of the robot navigation system; For dynamic obstacle costs in the temporary layer, perform a clearing operation at a higher decay rate than that of the regular obstacle layer in the local cost map; The shared timestamp synchronization mechanism triggers the update of the local cost map and the recalculation of the robot's local planner.

[0006] On the other hand, a dynamic obstacle real-time update device that coordinates perception and planning is provided, the device comprising: The perception data acquisition module is used to acquire perception data of dynamic obstacles; The perception data writing module is used to write perception data that meets preset conditions into a temporary layer of the local cost map of the robot navigation system. The clearing operation execution module is used to perform a clearing operation on the dynamic obstacle cost in the temporary layer at a higher decay rate than that of the regular obstacle layer in the local cost map; The update calculation trigger module is used to trigger the update of the local cost map and the recalculation of the robot's local planner through a shared timestamp synchronization mechanism.

[0007] In another aspect, a computer device is provided, comprising a processor and a memory, wherein the memory stores at least one instruction, at least one program, code set or instruction set, wherein the at least one instruction, at least one program, code set or instruction set is loaded and executed by the processor to realize the above-mentioned dynamic obstacle real-time update method for perception and planning coordination.

[0008] On another front, a computer-readable storage medium is provided, which stores at least one instruction, at least one program, code set, or instruction set, wherein the at least one instruction, at least one program, code set, or instruction set is loaded and executed by a processor to achieve the above-mentioned method for real-time dynamic obstacle updating in coordination of perception and planning.

[0009] In another aspect, a computer program product is provided, comprising a computer program stored in a computer-readable storage medium. A processor of a computer device reads the computer program from the computer-readable storage medium and executes the computer program, causing the computer device to perform the dynamic obstacle real-time update method for perception and planning coordination provided in the various optional implementations described above.

[0010] The technical solution provided by this invention may include the following beneficial effects: The computer equipment is equipped with a shared timestamp queue, which records the timestamp of perception fusion completion. The temporary layer plugin checks if this timestamp is greater than the previous processing timestamp. Upon detecting new data, it immediately triggers the generation of a local cost map update (instead of waiting for a fixed period). Simultaneously, it notifies the planner via an update flag, allowing the planner to read the latest cost map and recalculate the trajectory within the next control cycle. This compresses the perception-avoidance delay to within one control cycle. The temporary layer is configured with a decay rate several times higher than the regular layer, ensuring that the area occupied by dynamic obstacles is quickly cleared after they disappear. This prevents outdated obstacles from continuously occupying the cost map and causing the planner to repeatedly calculate invalid obstacle avoidance paths, thus reducing the number of invalid recalculations by the planner. In summary, this solution significantly shortens the response latency of the perception-planning link through an event-driven synchronous triggering mechanism and a differentiated rapid decay strategy, while reducing the waste of computing resources and effectively improving the real-time update efficiency of dynamic obstacles.

[0011] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and are not intended to limit the invention. Attached Figure Description

[0012] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with the invention and, together with the description, serve to explain the principles of the invention.

[0013] Figure 1 This is a system configuration diagram of a dynamic obstacle real-time update method involving perception and planning coordination according to an embodiment of the present invention; Figure 2 This is a flowchart of a dynamic obstacle real-time update method for perception and planning coordination provided in one embodiment of the present invention; Figure 3 This is a block diagram of a dynamic obstacle real-time update device for perception and planning coordination provided in an exemplary embodiment of the present invention. Detailed Implementation

[0014] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numerals in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present invention. Rather, they are merely examples of apparatuses and methods consistent with some aspects of the invention as detailed in the appended claims.

[0015] Figure 1 This is a system configuration diagram of a dynamic obstacle real-time update method involving perception and planning coordination according to an embodiment of the present invention. Figure 1 As shown, the system can be executed by computer device 100, which can be deployed inside robot 10 or a computer device that remotely controls robot 10. In this embodiment, the robot includes, but is not limited to, service robots, warehouse logistics robots, inspection robots, or autonomous vehicles.

[0016] Computer device 100 acquires perception data of dynamic obstacles; writes perception data that meets preset conditions into a temporary layer of the local cost map of the robot navigation system; performs a clearing operation on the dynamic obstacle cost in the temporary layer at a rate higher than that of the regular obstacle layer in the local cost map; and triggers the update of the local cost map and the recalculation of the robot's local planner through a shared timestamp synchronization mechanism.

[0017] Please refer to Figure 2 , Figure 2This is a flowchart of a dynamic obstacle real-time update method for perception and planning coordination provided in an embodiment of the present invention. The dynamic obstacle real-time update method for perception and planning coordination can be executed by a computer device, for example, the computer device described above... Figure 1 The service provider equipment 130 shown is shown.

[0018] The aforementioned method for real-time dynamic obstacle updates that combines perception and planning may include the following steps: Step 210: Acquire perception data of dynamic obstacles.

[0019] The aforementioned dynamic obstacles are those that are in motion relative to the robot.

[0020] In some embodiments, a computer device may use a target detection algorithm to determine whether the current obstacle is a dynamic obstacle.

[0021] The aforementioned perception data refers to data related to dynamic obstacles, including but not limited to: bounding box data, category data, fusion distance data, temporal continuity data, and depth validity.

[0022] The detection box data is the smallest axis-aligned bounding area represented by a set of parameters in the robot's corresponding sensor coordinate system or image coordinate system. The detection box data is used to locate the spatial range of dynamic obstacles.

[0023] Category data consists of discrete labels and their confidence levels that identify the predefined semantic categories to which dynamic obstacles belong.

[0024] The fused distance data is robot coordinate system distance data obtained by fusing the target detection results with the depth / pseudo-laser results and can be used by the Nav2 local cost map.

[0025] Temporal continuity data is a set of parameters that describe the consistency of identity and motion evolution of the same dynamic obstacle across consecutive time frames.

[0026] Depth validity is a binary or continuous metric that measures whether obstacle distance values ​​obtained from sensors or fusion algorithms are reliable and usable.

[0027] In some embodiments, the computer device can directly acquire perception data of dynamic obstacles through sensors set up on the robot, or it can acquire perception data of dynamic obstacles through a built-in target detection model.

[0028] Step 220: Write the perception data that meets the preset conditions into a temporary layer of the local cost map of the robot navigation system.

[0029] The above preset conditions are that the target category belongs to the dynamic obstacle category set, the detection confidence is greater than the threshold, the detection box is valid, the fusion distance is valid, the detection frame and the depth frame are time-synchronized, and the target can be transformed into the robot coordinate system.

[0030] Specifically, regarding the above target category conditions: the target category must belong to the preset dynamic obstacle category set.

[0031] Examples include person, pet, car, pallet, bicycle, and forklift. Categories with fixed backgrounds, walls, or those that do not affect the robot's movement are not written to the dynamic temporary layer.

[0032] Specifically, regarding the above detection confidence condition: the target detection confidence must be greater than the preset threshold.

[0033] For example, the confidence threshold for dynamic obstacle categories can be set to 0.6, meaning the target detection confidence is >0.6.

[0034] Specifically, regarding the above detection frame conditions: the area, aspect ratio, and position of the detection frame should meet the preset range. A detection frame that is too small may result in false detections, while a detection frame that is too large may indicate close-range occlusion or detection anomalies.

[0035] For example, the ratio of the detection box area to the image area can be set to be within [0.001, 0.8]; the aspect ratio of the detection box can be set to be within a reasonable range, for example, [0.2, 5.0].

[0036] Specifically, regarding the above fusion distance condition: the fusion distance corresponding to the detection box should be within the effective range of the robot's local obstacle avoidance.

[0037] For example, the range of the virtual LaserScan can be set to 0.15 m ≤ d fused ≤ 6.0 m, d fused This represents the blending distance value between a person and dynamic obstacles during the holiday.

[0038] Specifically, regarding the aforementioned depth quality conditions: there must be a sufficient number of effective depth points within the detection frame area, or there should be stable distance observations within the corresponding pseudo-laser sector.

[0039] For example, the effective depth pixel ratio within the detection box can be set to ≥ 30%, or the number of effective points within the corresponding sector can be ≥ K, where K can be any integer from 5 to 20.

[0040] Specifically, regarding the above time synchronization conditions: the difference between the timestamps of the detection result and the depth result is less than the synchronization threshold.

[0041] For example, the detection result matches the depth frame synchronization within ±20 ms, which can be set to |tdet -t depth |≤20 ms, t det The timestamp representing the test result, t depth The timestamp representing the depth result.

[0042] Specifically, regarding the coordinate transformation conditions mentioned above: the computer device can transform the target from the image coordinate system or camera coordinate system to the robot coordinate system based on the camera's intrinsic and extrinsic parameters and the robot coordinate system transformation relationship, thereby obtaining the target's position (x, y) or occupied area. If the coordinate transformation fails, it will not be written to the temporary layer.

[0043] The aforementioned robot navigation system is a complete hardware and software system that enables a robot to move autonomously and safely from a starting point to a target point. It includes core functions such as perception, localization, planning, and control, and is used to control the robot.

[0044] In this embodiment, the temporary layer has the characteristics of short lifespan, high refresh rate, independent decay strategy and read / write separation.

[0045] In some embodiments, the temporary layer of the local cost map of the robot navigation system is provided with a corresponding circular buffer, hash table or queue structure for storing relevant data of dynamic obstacles, and the computer device can write the perception data that meets the preset conditions into it.

[0046] Based on one or more of the above embodiments, in some embodiments, step 220 can be implemented as follows: converting the perception data that meets the preset conditions into dynamic obstacle grid data; calculating the occupied area of ​​the dynamic obstacle in the cost map based on the dynamic obstacle grid data; and writing the dynamic obstacle grid data into a temporary layer of the local cost map of the robot navigation system based on the occupied area.

[0047] The aforementioned dynamic obstacle raster data refers to a set of raster indexes obtained by computer equipment after converting dynamic obstacles from a continuous world coordinate system representation to a discrete raster map representation.

[0048] The area occupied by the dynamic obstacle in the cost map refers to the overall spatial extent of all the grids covered by the dynamic obstacle grid data. The occupied area can be represented by a set of grids or described by a minimum bounding rectangle.

[0049] In some embodiments, the computer device can acquire dynamic target data (including horizontal and vertical coordinates in the world coordinate system, occupied radius, and timestamp) that meets preset conditions, and load parameters of the local cost map (resolution, map origin coordinates, and size) and a temporary layer array. The world coordinates of the dynamic obstacle are subtracted from the map origin coordinates and then divided by the resolution to obtain the raster row and column number corresponding to the center of the dynamic obstacle. The occupied radius is then divided by the resolution and rounded up to obtain the raster radius. All candidate rasteres are traversed with the center raster as the center and the raster radius as the boundary. The actual distance from the center of each candidate raster to the center of the obstacle is calculated and compared with the occupied radius. The row and column numbers of rasteres with a distance less than or equal to the radius are recorded as dynamic obstacle raster data (i.e., the set of occupied raster indices).

[0050] The computer device scans the row and column numbers of all grid cells in the raster data, finds the minimum row number, maximum row number, minimum column number, and maximum column number, and determines a minimum bounding rectangle as the area occupied by the dynamic obstacle in the cost map based on the values ​​corresponding to the minimum row number, maximum row number, minimum column number, and maximum column number. Based on the rectangular boundary of the occupied area, the corresponding area of ​​the temporary layer within the rectangle is first reset to the free space algebraic value. Then, each grid cell in the dynamic obstacle raster data is traversed to check whether its row and column numbers are within the valid range of the map. If they are valid, the value of the corresponding position in the temporary layer is set to the fatal obstacle algebraic value, and the current timestamp is recorded.

[0051] Step 230: Perform a clearing operation on the dynamic obstacle cost in the temporary layer at a higher decay rate than that of the regular obstacle layer in the local cost map.

[0052] The aforementioned dynamic obstacle cost refers to the passage cost that a robot must bear when a certain grid location is occupied by a dynamic obstacle (e.g., pedestrians, other vehicles, or temporarily appearing objects) in the local cost map of a robot navigation system.

[0053] The aforementioned decay rate, which is higher than that of the regular obstacle layer in the local cost map, means that the cost of dynamic obstacles in the temporary layer decreases over time (e.g., how many generations of value are reduced per second) faster than the cost in the regular obstacle layer. In other words, the cost of the temporary layer reaches zero in a shorter time.

[0054] The aforementioned clearing operation refers to the process by which a computer device periodically reduces the value of the dynamic obstacle cost in a temporary layer according to a preset decay rate until it becomes zero (representing free space), thereby removing the information occupied by the invalid dynamic obstacles.

[0055] Based on one or more of the above embodiments, in some embodiments, step 230 can be implemented as follows: obtaining the unobserved duration and target lifespan of the dynamic obstacle, wherein the unobserved duration is the duration between the current update time of the cost map and the last effective observation time of the dynamic obstacle, and the target lifespan is the maximum time that the dynamic obstacle is allowed to exist in the temporary layer; and according to the unobserved duration and target lifespan of the dynamic obstacle, performing a clearing operation on the cost of the dynamic obstacle in the temporary layer at a rate higher than the decay rate of the regular obstacle layer in the local cost map.

[0056] Furthermore, in some embodiments, the computer device reduces the cost value according to a decay function if the unobserved duration of the dynamic obstacle does not exceed the target lifespan; if the unobserved duration of the dynamic obstacle exceeds the target lifespan, the dynamic obstacle raster data is deleted, and the raster area occupied by the dynamic obstacle raster data is marked as an area to be cleared.

[0057] The aforementioned last valid observation time refers to the timestamp of the last time the computer equipment effectively observed the dynamic target and completed detection, distance fusion, and coordinate transformation. The timestamp can be the image frame acquisition time, i.e., the header.stamp added to the image frame by the camera driver; it can also be the detection result output time, i.e., the time when the target detection network completes inference and outputs the detection box; or it can be the fusion result generation time, i.e., the time when the detection box and depth data are matched and the robot coordinate system position is generated.

[0058] In this scheme, the fusion result generation time is preferably used as the last observation time. The fusion result generation time indicates that the target has completed the entire process from image detection to spatial localization and can be directly used for updating the cost map.

[0059] In some embodiments, after the computer device obtains the unobserved duration and target lifecycle of a dynamic obstacle, it compares the unobserved duration and the target lifecycle. If the unobserved duration is less than or equal to the target lifecycle, the dynamic obstacle is retained in the temporary layer, and the cost of the dynamic obstacle is reduced according to a preset decay function (linear decay or exponential decay). If the unobserved duration exceeds the lifecycle, the computer device deletes the dynamic obstacle from the temporary layer and marks the grid area it occupied as an area to be cleared.

[0060] During the cost map update phase, the temporary layer performs a clearing or restoration operation on the grids in the area to be cleared. If the area is not occupied by other layers (e.g., static obstacle layer or pseudo laser layer), it is restored to the free cost value; if other layers are marked as obstacles, the cost value of the other layers is retained.

[0061] For example, the computer device can be set to a target lifetime of 0.5 seconds, with a linear decay function and a decay factor of 0.2 per second (i.e., the cost decreases by 20% of the initial value every second). A short hold window is also set to 0.15 seconds: when the unobserved duration is less than or equal to 0.15 seconds, the dynamic obstacle cost in the temporary layer remains at the initial lethal value (e.g., 254); when the unobserved duration is between 0.15 seconds and 0.5 seconds, the current cost equals the initial cost using the linear formula. [1 - attenuation coefficient × (duration of no observation - 0.15 seconds)], when the duration of no observation is greater than 0.5 seconds, the computer device will delete the data of the dynamic obstacle from the cache and mark the corresponding grid area as pending clearing.

[0062] During the cost map update phase, the temporary layer performs a clearing or restoration operation on the grid within the area to be cleared: if the cost map area is not occupied by other layers (e.g., a normal obstacle layer or a pseudo laser layer), it is restored to an empty cost value; otherwise, the markers of other layers are retained.

[0063] To avoid obstacles flashing frequently in the field of view due to momentary missed detections, a short holding window (e.g., 0.15 seconds) can be set. Within the window, the cost of the dynamic obstacle remains unchanged. After the window expires but is still within the lifecycle (e.g., 0.5 seconds), linear or exponential decay is performed. After the lifecycle expires, the target is deleted.

[0064] The lifecycle can be configured differently for different types of dynamic targets: a shorter lifecycle (e.g., 0.2 seconds) is set for targets that are falsely detected in a single frame, a standard lifecycle (e.g., 0.5 seconds) is set for targets that are confirmed in multiple consecutive frames, and the lifecycle of the occupied area can be appropriately extended according to the speed of the target approaching the robot, while still implementing a rapid decay strategy, thereby achieving a balance between safety and passage efficiency.

[0065] In this embodiment, through the above mechanism, the temporary layer can quickly write high-cost data when a dynamic target appears and quickly release the passage space when the dynamic target disappears.

[0066] Step 240: Trigger the update of the local cost map and the recalculation of the robot's local planner through the shared timestamp synchronization mechanism.

[0067] In some embodiments, step 240 above can be implemented as follows: obtaining a shared timestamp queue, the shared timestamp queue containing the fusion completion timestamp of dynamic obstacles, the cost map update timestamp, and an update flag, the update flag being used to indicate whether it is necessary to trigger a recalculation of the robot local planner; if the fusion completion timestamp is later than the fusion completion timestamp recorded when the cost map update was last triggered, triggering an update of the local cost map and a recalculation of the robot local planner.

[0068] The aforementioned shared timestamp queue refers to a time-series synchronization data structure maintained by the computer device among the perception fusion module, the local cost map update module, and the local planner. It records the perception data generation time, dynamic obstacle writing time, cost map update time, and planner trigger time. The computer device can directly read the shared timestamp queue from the perception fusion module.

[0069] The computer device will retrieve the fusion completion timestamp from the shared timestamp queue and compare it with the historical fusion completion timestamp recorded when the cost map update was last triggered. If the current fusion completion timestamp is later than the historical timestamp, it means that new perceptual fusion data has been written to the shared timestamp queue. The value of the update flag determines whether to perform subsequent operations: when the update flag is true, the update operation of the local cost map is triggered (that is, the dynamic obstacle costs in the temporary layer are fused into the main cost map), and the robot's local planner is recalculated (for example, the local path planning algorithm is rerun to generate a new obstacle avoidance trajectory).

[0070] The aforementioned updates to the local cost map include, but are not limited to, prioritizing new data, local refreshing, discarding expired data, triggering changes, and recalculating the plan.

[0071] New data priority means that new observations of the same target overwrite old observations, avoiding the use of outdated locations by the local planner.

[0072] Partial refresh refers to updating only the area affected by dynamic targets, reducing the computational overhead of updating the entire map.

[0073] Expiration and discarding refers to the process where, when the timestamp of perceived data becomes too old, the computer device discards the data and no longer writes it to the temporary layer.

[0074] Change-triggered updates mean that temporary layer updates are triggered only when a target is added, its location changes, its value changes, or the target expires, thus avoiding meaningless repeated refreshes.

[0075] Planning recalculation refers to the process where, after the temporary layer of the cost map has been updated, the local planner reads the latest map and recalculates the velocity command or local trajectory.

[0076] Based on one or more of the above embodiments, in some embodiments, the robot local planner is a DWB planner or a TEB planner.

[0077] The DWB (Dynamic Window Based Planner) planner is an extension of the traditional dynamic window method. It selects the optimal motion command by sampling multiple sets of control commands in the velocity space and evaluating the cost of each trajectory using a local cost map.

[0078] The TEB (Time Elastic Band Planner) planner expresses the robot path as a series of timestamped pose sequences. By constructing a hypergraph that includes obstacle avoidance, dynamic constraints, velocity limits, and time optimization, it uses nonlinear optimization methods to solve for a smooth and time-optimal trajectory.

[0079] The above solution fundamentally solves the link delay problem caused by the timing mismatch between sensor release frequency, cost map update, and planner control cycle in the background technology by introducing a shared timestamp synchronization mechanism and a temporary layer fast decay strategy.

[0080] In this embodiment of the invention, the computer device is equipped with a shared timestamp queue. This queue records the timestamp of completed perception fusion and checks if the timestamp is greater than the previous processing timestamp in the temporary layer plugin. Once new data is detected, a local cost map update is immediately triggered (instead of waiting for a fixed period). Simultaneously, an update flag is used to notify the planner, enabling the planner to read the latest cost map and recalculate the trajectory within the next control cycle. This compresses the delay in perception and avoidance to within one control cycle. The temporary layer is configured with a decay rate several times higher than that of a regular layer, allowing dynamic obstacles to quickly clear their occupied areas after disappearing. This prevents outdated obstacles from continuously occupying the cost map and causing the planner to repeatedly calculate invalid obstacle avoidance paths, thus reducing the number of invalid recalculations by the planner. In summary, this solution significantly shortens the response latency of the perception-planning link through an event-driven synchronous triggering mechanism and a differentiated rapid decay strategy, while reducing the waste of computing resources and effectively improving the real-time update efficiency of dynamic obstacles.

[0081] In some application embodiments, the computer device performs dynamic obstacle avoidance according to the following process: First, the perception fusion module generates dynamic obstacle messages and writes them to a shared timestamp queue. Then, the temporary layer plugin reads the latest fusion completion timestamp from the queue. If this timestamp is greater than the previous processing timestamp, it determines that a new perception result exists and reads the corresponding target data from the dynamic obstacle cache. It then performs target input judgment, coordinate transformation, and raster mapping, and calls the local update interface of the cost map to update the boundary of the target's influence area to the main cost map. Simultaneously, it updates the temporary layer's write timestamp and sets the update flag to true in the shared timestamp queue. Finally, the local planner (DWB planner or TEB planner) reads this flag or the cost map update timestamp in the next control cycle. If it finds that the cost map has been updated, it re-executes velocity sampling, trajectory scoring, or trajectory optimization, outputting new velocity commands to enable the robot to promptly perform deceleration, detour, or stop actions.

[0082] Please refer to Figure 3 The diagram illustrates a block diagram of a real-time dynamic obstacle update device for perception and planning coordination provided by an exemplary embodiment of the present invention. This device can be implemented as all or part of a robot system through hardware or a combination of hardware and software to achieve the above-described... Figure 2 All or part of the steps in the illustrated embodiments. For example... Figure 3 As shown, the dynamic obstacle real-time update device that coordinates perception and planning includes: The perception data acquisition module 301 is used to acquire perception data of dynamic obstacles; The perception data writing module 302 is used to write perception data that meets preset conditions into a temporary layer of the local cost map of the robot navigation system. The clearing operation execution module 303 is used to perform a clearing operation on the dynamic obstacle cost in the temporary layer at a higher decay rate than that of the regular obstacle layer in the local cost map. The update calculation trigger module 304 is used to trigger the update of the local cost map and the recalculation of the robot's local planner through a shared timestamp synchronization mechanism.

[0083] In an exemplary embodiment, a computer device is also provided, the computer device including a processor and a memory, the memory storing instructions which are executed by the processor.

[0084] In an exemplary embodiment, a computer program product is also provided, comprising computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions to implement all or part of the steps of the methods shown in the above embodiments of the present invention.

[0085] In an exemplary embodiment, a computer-readable storage medium is also provided, which stores a computer program that is loaded and executed by a processor to implement all or part of the steps of the methods shown in the above embodiments of the present invention.

[0086] Those skilled in the art will understand that all or part of the steps of the above embodiments can be implemented by hardware, or by a program instructing related hardware. The program can be stored in a computer-readable storage medium, such as a read-only memory, a disk, or an optical disk.

[0087] Although the operations are described in a specific order in the accompanying drawings, this should not be construed as requiring these operations to be performed in the specific order or serial order shown, or requiring all of the operations shown to obtain the desired result. In certain environments, multitasking and parallel processing may be advantageous.

[0088] Any step, operation, or procedure described herein may be performed or implemented using one or more hardware or software modules, either alone or in combination with other devices. In one embodiment, the software module is implemented using a computer program product comprising a computer-readable medium containing computer program code, which is executable by a computer processor to perform any or all of the described steps, operations, or procedures.

[0089] The foregoing description of embodiments of the invention has been provided for purposes of illustration and description. The foregoing description is not exhaustive and is not intended to limit the invention to the exact forms disclosed; various modifications and variations may be made in accordance with the foregoing teachings, or may be derived from the practice of the invention. These embodiments were chosen and described to illustrate the principles of the invention and its practical application, enabling those skilled in the art to utilize the invention in various embodiments and with various modifications to suit the particular purpose of the concept.

[0090] It can be further understood that, unless otherwise specified, "connection" includes both direct connections where no other components exist between the two parties and indirect connections where other components exist between them.

[0091] It is further understood that although the operations are described in a specific order in the accompanying drawings in the embodiments of the present invention, this should not be construed as requiring these operations to be performed in the specific order or serial order shown, or requiring all the operations shown to obtain the desired result. In certain environments, multitasking and parallel processing may be advantageous.

[0092] Other embodiments of the invention will readily occur to those skilled in the art upon consideration of the specification and practice of the invention disclosed herein. This invention is intended to cover any variations, uses, or adaptations of the invention that follow the general principles of the invention and include common knowledge or customary techniques in the field of the invention not disclosed herein. The specification and examples are to be considered exemplary only, and the true scope and spirit of the invention are indicated by the following claims.

[0093] It should be understood that the present invention is not limited to the precise structure described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of the invention is limited only by the appended claims.

[0094] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for real-time updating of dynamic obstacles through coordinated perception and planning, characterized in that, The methods include: Acquire perception data of dynamic obstacles; Perception data that meets preset conditions is written into a temporary layer of the local cost map of the robot navigation system; For dynamic obstacle costs in the temporary layer, perform a clearing operation at a higher decay rate than that of the regular obstacle layer in the local cost map; The shared timestamp synchronization mechanism triggers the update of the local cost map and the recalculation of the robot's local planner.

2. The method according to claim 1, characterized in that, Perception data that meets preset conditions is written into a temporary layer of the local cost map of the robot navigation system, including: The perceived data that meets the preset conditions is converted into dynamic obstacle grid data; Calculate the area occupied by dynamic obstacles in the cost map based on dynamic obstacle grid data; Based on the occupied area, dynamic obstacle grid data is written into a temporary layer of the local cost map of the robot navigation system.

3. The method according to claim 1, characterized in that, For dynamic obstacle costs in the temporary layer, a clearing operation is performed at a higher decay rate than that of the regular obstacle layer in the local cost map, including: Obtain the unobserved duration and target lifetime of dynamic obstacles. The unobserved duration is the time between the current update time of the cost map and the last valid observation time of the dynamic obstacle. The target lifetime is the maximum time that the dynamic obstacle is allowed to exist in the temporary layer. Based on the unobserved duration of dynamic obstacles and the target lifecycle, the cost of dynamic obstacles in the temporary layer is cleared at a higher decay rate than that of regular obstacle layers in the local cost map.

4. The method according to claim 3, characterized in that, Based on the unobserved duration of dynamic obstacles and the target's lifecycle, a clearing operation is performed on the cost of dynamic obstacles in the temporary layer at a higher decay rate than that of the regular obstacle layer in the local cost map, including: If the unobserved duration of a dynamic obstacle does not exceed the target's lifetime, the cost is reduced according to the decay function. If the unobserved duration of a dynamic obstacle exceeds the target's lifecycle, the dynamic obstacle raster data will be deleted, and the raster areas occupied by the dynamic obstacle raster data will be marked as areas to be cleared.

5. The method according to claim 1, characterized in that, The shared timestamp synchronization mechanism triggers updates to the local cost map and recalculations of the robot's local planner, including: Obtain the shared timestamp queue, which contains the timestamp of dynamic obstacle fusion completion, the timestamp of cost map update, and the update flag. The update flag is used to indicate whether the robot's local planner needs to be recalculated. If the fusion completion timestamp is later than the fusion completion timestamp recorded when the cost map update was last triggered, an update to the local cost map and a recalculation of the robot's local planner will be triggered.

6. The method according to any one of claims 1 to 5, characterized in that, The robot's local planner is either a DWB planner or a TEB planner.

7. A dynamic obstacle real-time update device that coordinates perception and planning, characterized in that, The device includes: The perception data acquisition module is used to acquire perception data of dynamic obstacles; The perception data writing module is used to write perception data that meets preset conditions into a temporary layer of the local cost map of the robot navigation system. The clearing operation execution module is used to perform a clearing operation on the dynamic obstacle cost in the temporary layer at a higher decay rate than that of the regular obstacle layer in the local cost map; The update calculation trigger module is used to trigger the update of the local cost map and the recalculation of the robot's local planner through a shared timestamp synchronization mechanism.

8. A computer device, characterized in that, The computer device includes a processor and a memory, the memory storing instructions which are executed by the processor to implement the dynamic obstacle real-time update method for perception and planning coordination as described in any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that, The storage medium stores instructions that are executed by the processor of a computer device to implement the dynamic obstacle real-time update method for perception and planning coordination as described in any one of claims 1 to 6.

10. A computer program product, characterized in that, The computer program product includes computer instructions stored in a computer-readable storage medium; the computer instructions are read and executed by the processor of a computer device to implement the dynamic obstacle real-time update method for perception and planning coordination as described in any one of claims 1 to 6.

Images