Logistics robot control method and system based on artificial intelligence

Abstract

The invention discloses a logistics robot control method and system based on artificial intelligence, and relates to the technical field of robots, and the method comprises the following steps: 1, each logistics robot is connected with a data collection API when starting to execute a task, and automatically records task data, all logistics robots upload task data to a cloud learning platform through a wireless network, and a preliminary experience database is constructed; the logistics robot is connected with the data acquisition AP I to record the task data in real time and upload the task data to the cloud learning platform to construct the preliminary experience database, the task data of all robots are integrated, the system is dynamically trained by using online learning, the multi-dimensional task experience database is established, and the task experience data of the system is dynamically trained. And according to the current logistics state, an experience area is intelligently selected to perform task allocation and control strategy optimization, upper layer decision and lower layer feedback are dynamically optimized based on real-time feedback, common accumulation and sharing of experience are realized, and operation strategies of all robots are continuously optimized.

Description

Technical Field

[0001] The present invention relates to the technical field of robots, and particularly to a control method and system for a logistics robot based on artificial intelligence. Background Art

[0002] With the rapid growth of e-commerce and global logistics demand, the logistics industry's demand for automation and intelligence has gradually increased. Traditional logistics management methods often rely on manual operations, which are not only inefficient but also easily affected by human factors. Therefore, many enterprises have begun to apply robot technology to the logistics field to achieve the automation and intelligence of the logistics process.

[0003] The existing technologies have the following deficiencies: Traditional logistics management methods can no longer meet the market's requirements for high efficiency and low cost, resulting in frequent manual intervention and low operation efficiency. Although existing control methods for logistics robots based on artificial intelligence can automate some tasks, they show insufficient environmental adaptability, poor task execution flexibility, insufficient real-time data processing capabilities, and an imperfect experience sharing mechanism. Due to the lack of an effective feedback mechanism and dynamic optimization strategy, it is difficult to perform optimally in complex and changing environments, limiting the success rate and efficiency of robots in logistics tasks. There is also a lack of an effective experience sharing and collective learning mechanism among robots, so that successful experiences and lessons learned cannot be effectively transmitted among all robots.

[0004] The above information disclosed in the background art section is only used to enhance the understanding of the background of the present disclosure, and thus it may include information that does not constitute the prior art known to those of ordinary skill in the art. Summary of the Invention

[0005] The purpose of the present invention is to provide a control method and system for a logistics robot based on artificial intelligence to solve the problems in the above background art.

[0006] To achieve the above purpose, the present invention provides the following technical solution: A control method for a logistics robot based on artificial intelligence, comprising the following steps:

[0007] Step 1: When each logistics robot starts to execute a task, it connects to a data collection API and automatically records task data, including task environment status, task execution status, and task feedback information. All logistics robots upload the task data to a cloud learning platform through a wireless network to build a preliminary experience database;

[0008] Step 2: Integrate the task data of all logistics robots through the cloud learning platform and perform dynamic training according to the real-time task data using online learning;

[0009] Step 3: Establish a multi-dimensional task experience database. Divide the upper-layer decision-making to automatically select an experience area for logistics task allocation according to the current logistics status, optimize the execution control strategy, and divide the lower-layer feedback to timely feedback the experience of the corresponding task data to the cloud according to the logistics task;

[0010] Step 4: Dynamically optimize the upper-layer decision-making and the lower-layer feedback based on the feedback, continuously update to form a closed loop, enable all logistics robots to jointly accumulate experience when performing tasks, and share the experience through the cloud learning platform to optimize the operation strategies of all logistics robots.

[0011] Preferably, each logistics robot connects to the data collection API when starting to execute a task, automatically records the task data, including the task environment status, task execution status, and task feedback information. The task environment status includes the current position of the logistics robot, the position of obstacles, and the task target. The task execution status includes the task start time, task completion time, and whether the task is successful. The task feedback information includes task execution problems, obstacles encountered during the task, and task success rate. During the process of automatically recording the task data, the precise coordinates of the current position of the logistics robot are accurately recorded through the global positioning system and the internal positioning algorithm. The internal sensor group is used to detect the surrounding environment where the logistics robot is located, and the position of obstacles is dynamically recorded. According to the task target position and the predetermined task information, the task execution status is compared with the preset task completion status, and the task feedback information is recorded by combining the task environment status and the task execution status through the internal information transmission group. All logistics robots upload the task data to the cloud learning platform through the wireless network to construct a preliminary experience database, and divide the preliminary experience database into the first experience area, the second experience area, the third experience area, and the fourth experience area. The first experience area includes all the execution records of successfully completed tasks, the second experience area includes all the execution records of failed tasks, the third experience area includes the task execution status under different environmental variables, and the fourth experience area includes the influence of manual intervention on the task execution status.

[0012] Preferably, the task data of all logistics robots is integrated through a cloud learning platform. Based on timestamps, the task data of different logistics robots within the same time period is correlated to obtain diverse task execution states under the same environmental variables. Corresponding tags are assigned to each task objective. Based on the integrated task dataset, a basic deep reinforcement learning model is constructed through the cloud learning platform for rapid adaptation, continuously optimizing the task rules and control strategies of the logistics robots. The task data uploaded in real time from the logistics robots is received through online learning and processed using incremental learning methods. The task execution state is used as an immediate reward signal to input into the basic deep reinforcement learning model for online update. An incremental feedback loop mechanism is established to immediately update the weights of the deep reinforcement learning model using new task data after each task execution. The task data is weighted based on environmental variables, and its specific formula is:

[0013]

[0014] where, w final represents the weight of each task execution state after update, β represents the feedback influence factor of task feedback on task data, E t represents the specific index of the task execution state at the current timestamp, α t represents the time influence factor of the timestamp on the current task execution state, E i,t represents the i-th task execution state at timestamp t, E env represents the specific index of the task execution state under the current environmental variable, E j,env represents the j-th task execution state under the environmental variable env, ω represents the time influence factor of the environmental variable on the current task execution state, ω pre represents the weight of each task execution state after the previous update, N represents the total number of task data participating in weight calculation, M represents the total number of task data participating in environmental variable calculation. Tasks executed in complex environments are preferably selected to help all logistics robots update learning objectives.

[0015] Preferably, the preliminary experience database is reclassified based on the success rate dimension, including high success rate area, medium success rate area, and low success rate area. The preliminary experience database is reclassified based on the environmental condition dimension, including humidity area and temperature area. The preliminary experience database is reclassified based on the task feature dimension, including task type area and task complexity area. The logistics API is connected to monitor the current logistics status in real time and divide upper-layer decisions to train high-level experience area decisions. Based on the current logistics status, the optimal experience area after reclassification is dynamically selected through a decision algorithm for logistics task allocation. After the task is allocated, resource optimization is automatically performed according to the task execution state, and lower-layer feedback is divided to feed back the experience of the corresponding task data to the cloud in real time according to the logistics task.

[0016] Preferably, the decision algorithm is retrained regularly based on the latest feedback. Once a new control strategy is generated, the logistics robot will immediately receive the updated control strategy for execution in the next round of tasks, continuously updating to form a closed loop, and a feedback gain model is constructed for quantitative improvement. The specific formula is as follows:

[0017]

[0018] Among them, F(t) represents the feedback value of the currently executed task, α represents the feedback gain factor, F(t - 1) represents the feedback value at the previous time stamp, γ represents the historical feedback gain factor, F k represents the total feedback value of the past k task executions, C t represents the complexity factor of the currently executed task, δ represents the immediate reward factor, R(t) represents the immediate reward signal for the task executed at time t, and the optimal experience area is completely converted into a collective knowledge base, enabling all logistics robots to jointly accumulate experience when executing tasks, using collective wisdom to identify common problems and successful control strategies through machine learning algorithms, generating the best experience for all logistics robots to reference, and sharing the experience through the cloud learning platform to optimize the operation strategies of all logistics robots.

[0019] In the above technical solution, the technical effects and advantages provided by the present invention are as follows:

[0020] 1. By connecting to the data acquisition API and uploading task data in real time, the logistics robot can dynamically record and analyze the task environment status and execution status, improving the accuracy and reliability of the data.

[0021] 2. Using the deep reinforcement learning model constructed by the cloud learning platform, the robot can automatically optimize the task strategy based on task data and real-time feedback, and through incremental learning of task data, update the decision rules in a timely manner to achieve intelligent task allocation and execution, enhancing the autonomous learning and adaptation ability of the logistics robot.

[0022] 3. Establish a multi-dimensional task experience database, and reclassify the preliminary experience database according to different dimensions based on the task data of all logistics robots, enabling the logistics robot to select the most suitable execution strategy under different environmental conditions.

[0023] 4. Completely convert the optimal experience area into a collective knowledge base, enabling different logistics robots to share successful and failed experiences, forming a stronger collaborative working ability for logistics robots.

[0024] 5. By constructing a feedback gain model, the effectiveness of the control strategy can be quantitatively improved, ensuring that the logistics robot can improve its task execution ability during continuous learning, forming an effective closed-loop feedback mechanism, and maintaining long-term efficient operation. Brief Description of the Drawings

[0025] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings required to be used in the embodiments. Obviously, the drawings described below are only some embodiments recorded in the present invention. For those of ordinary skill in the art, other drawings can also be obtained based on these drawings.

[0026] Figure 1 It is a method flow chart of a logistics robot control method based on artificial intelligence according to the present invention.

[0027] Figure 2 It is a module schematic diagram of a logistics robot control system based on artificial intelligence according to the present invention. Detailed Embodiments

[0028] Now, the exemplary embodiments will be described more fully with reference to the accompanying drawings. However, the exemplary embodiments can be implemented in various forms and should not be construed as limited to the examples set forth herein; rather, these exemplary embodiments are provided so that the present disclosure will be more thorough and complete, and will fully convey the concept of the exemplary embodiments to those skilled in the art.

[0029] Embodiment 1

[0030] The present invention provides a logistics robot control method and system based on artificial intelligence as Figure 1 shown, including the following steps:

[0031] Step 1: When each logistics robot starts to execute a task, it connects to the data acquisition API and automatically records task data, including the task environment status, task execution status, and task feedback information. All logistics robots upload the task data to the cloud learning platform through a wireless network to build a preliminary experience database;

[0032] When each logistics robot starts to execute a task, it connects to the data collection API and automatically records task data, including the task environment status, task execution status, and task feedback information. The task environment status includes the current position of the logistics robot, the position of obstacles, and the task target. The task execution status includes the task start time, task completion time, and whether the task is successful. The task feedback information includes task execution problems, obstacles encountered during the task, and the task success rate. During the process of automatically recording task data, the precise coordinates of the current position of the logistics robot are accurately recorded through the Global Positioning System and internal positioning algorithms. The internal sensor group is used to detect the surrounding environment of the logistics robot, and the position of obstacles is dynamically recorded. According to the task target position and predetermined task information, the task execution status is compared with the preset task completion status, and the task feedback information is recorded by combining the task environment status and task execution status through the internal information transfer group. All logistics robots upload the task data to the cloud learning platform through wireless networks to build a preliminary experience database. The preliminary experience database is divided into a first experience area, a second experience area, a third experience area, and a fourth experience area. The first experience area includes all execution records of successfully completed tasks. The second experience area includes all execution records of failed tasks. The third experience area includes the task execution status under different environmental variables. The fourth experience area includes the impact of manual intervention on the task execution status.

[0033] Step 2: Integrate the task data of all logistics robots through the cloud learning platform and perform dynamic training based on the real-time task data using online learning;

[0034] Integrate the task data of all logistics robots through the cloud learning platform, associate the task data of different logistics robots within the same time period based on the time stamp, obtain diverse task execution statuses under the same environmental variables, assign corresponding labels (such as handling, distribution, inventory counting, etc.) to each task target, and build a basic deep reinforcement learning model (such as Convolutional Neural Network (CNN) or Long Short-Term Memory Network (LSTM)) through the cloud learning platform based on the integrated task dataset for rapid adaptation, continuously optimizing the task rules and control strategies of the logistics robots. Receive the task data uploaded in real time from the logistics robots through online learning and process it using the incremental learning method. Use the task execution status as an immediate reward signal to input into the basic deep reinforcement learning model for online update, establish an incremental feedback loop mechanism to immediately update the weights of the deep reinforcement learning model using the new task data after each task execution, and weight the task data based on the time stamp, environmental variables, and task feedback. The specific formula is:

[0035]

[0036] where, w finalrepresents the weight of the execution status of each task after update, β represents the feedback impact factor of task feedback on task data, E t represents the specific index of the task execution status at the current timestamp, α t represents the time impact factor of the timestamp on the current task execution status, E i,t represents the execution status of the i-th task at timestamp t, E env represents the specific index of the task execution status in the current environmental variables, E j,env represents the execution status of the j-th task in the environmental variable env, ω represents the time impact factor of the environmental variable on the current task execution status, ω pre represents the weight of the execution status of each task after the previous update, N represents the total number of task data participating in the weight calculation, M represents the total number of task data participating in the environmental variable calculation. Prioritize the tasks executed in complex environments to help all logistics robots update their learning goals.

[0037] Step 3: Establish a multi-dimensional task experience database. Divide the upper-layer decision to automatically select the experience area for logistics task allocation according to the current logistics status, optimize the execution control strategy, and divide the lower-layer feedback to feedback the experience of the corresponding task data to the cloud in real time according to the logistics tasks;

[0038] Re-classify the preliminary experience database based on the success rate dimension, including the high success rate area, the medium success rate area, and the low success rate area. The high success rate area includes task data with a success rate higher than 80%, the medium success rate area includes task data with a success rate between 50% and 80%, and the low success rate area includes task data with a success rate lower than 50%. Re-classify the preliminary experience database based on the environmental condition dimension, including the humidity area and the temperature area. Re-classify the preliminary experience database based on the task characteristics dimension, including the task type area and the task complexity area. Connect to the logistics API to monitor the current logistics status in real time and divide the upper-layer decision to train the decision of the high-level experience area (such as A3C). Based on the current logistics status, dynamically select the optimal experience area after re-classification through a decision algorithm (such as C4.5 or random forest) for logistics task allocation. After the task is allocated, automatically optimize the resources according to the task execution status. Divide the lower-layer feedback to feedback the experience of the corresponding task data to the cloud in real time according to the logistics tasks.

[0039] Step 4: Dynamically optimize the upper-layer decision and the lower-layer feedback based on the feedback, continuously update to form a closed loop, enable all logistics robots to accumulate experience together when executing tasks, and share the experience through the cloud learning platform to optimize the operation strategies of all logistics robots;

[0040] Regularly retrain the decision-making algorithm based on the latest feedback. Once a new control strategy is generated, the logistics robot will immediately receive the updated control strategy for execution in the next round of tasks, continuously update to form a closed loop, and build a feedback gain model for quantitative improvement. The specific formula is as follows:

[0041]

[0042] Among them, F(t) represents the feedback value of the currently executed task, α represents the feedback gain factor, F(t - 1) represents the feedback value at the previous time stamp, γ represents the historical feedback gain factor, F k represents the total feedback value of the past k task executions, C t represents the complexity factor of the currently executed task, δ represents the immediate reward factor, R(t) represents the immediate reward signal for the task executed at time t, and completely convert the optimal experience area into a collective knowledge base, enabling all logistics robots to accumulate experience together when executing tasks, using collective wisdom to identify common problems and successful control strategies through machine learning algorithms (such as DQN, DDPG, etc.), generating the best experience for all logistics robots to reference, and sharing the experience through the cloud learning platform to optimize the operation strategies of all logistics robots.

[0043] The present invention provides a logistics robot control system based on artificial intelligence as Figure 2 shown, including a record uploading module, an integration training module, a decision feedback module, and a sharing optimization module;

[0044] Record uploading module: Each logistics robot connects to the data acquisition API when starting to execute a task, automatically records task data, including task environment status, task execution status, and task feedback information. All logistics robots upload the task data to the cloud learning platform through wireless networks to build a preliminary experience database;

[0045] In this embodiment, the specifically described record uploading module enables each logistics robot to automatically record and upload task data when starting to execute a task, relying on the connected data collection API. In the specific process, each robot first accurately obtains its current coordinates through the global positioning system and the internal positioning algorithm, and uses the internal sensor group to detect the surrounding environment of the logistics robot. The internal sensor group specifically includes an environmental temperature sensor, a photosensitive sensor, an environmental humidity sensor, and an infrared sensor. During the process of the logistics robot executing a task, the task execution status is compared with the preset task completion status according to the task target location and the predetermined task information, and the task feedback information is recorded by combining the task environment status and the task execution status through the internal information transmission group. All logistics robots upload the task data to the cloud learning platform through the wireless network to build a preliminary experience database, and divide the preliminary experience database into a first experience area, a second experience area, a third experience area, and a fourth experience area. Among them, the first experience area includes all the execution records of successfully completed tasks, the second experience area includes all the execution records of failed tasks, the third experience area includes the task execution status under different environmental variables, and the fourth experience area includes the impact of manual intervention on the task execution status. According to experiments, the third experience area can include that the task execution success rate is 90% in an environment with a humidity of 70% and a temperature of 22°C. According to experiments, the third experience area can also include that during the task execution process of the logistics robot, it is manually intervened once, resulting in the task completion time increasing to 15 minutes, but the task is successfully completed.

[0046] By connecting to the data collection API and uploading task data in real time, the logistics robot can dynamically record and analyze the task environment status and execution status, improving the accuracy and reliability of the data.

[0047] Integration training module: Integrate the task data of all logistics robots through the cloud learning platform, and use online learning to perform dynamic training based on real-time task data;

[0048] In this embodiment, the specifically described integration training module correlates the task data of different logistics robots within the same time period based on the time stamp to obtain diverse task execution statuses under the same environmental variables. According to experiments, logistics robot A successfully executed a task at time t = 2025-10-01 9:00:00 with environmental variables of 70% humidity and 22°C temperature.

[0049] Using the deep reinforcement learning model built by the cloud learning platform, the robot can automatically optimize the task strategy based on task data and real-time feedback, update the decision rules in a timely manner through incremental learning of the task data, realize intelligent task allocation and execution, and improve the autonomous learning and adaptation ability of the logistics robot.

[0050] Decision feedback module: Establish a multi-dimensional task experience database, divide the upper-layer decision-making to automatically select the experience area for logistics task allocation according to the current logistics status, optimize the execution control strategy, and divide the lower-layer feedback to timely feedback the experience of the corresponding task data to the cloud according to the logistics task;

[0051] Sharing and optimization module: Dynamically optimize the upper-layer decision-making and lower-layer feedback based on the feedback, continuously update to form a closed loop, enable all logistics robots to accumulate experience jointly when performing tasks, and share the experience through the cloud learning platform to optimize the operation strategies of all logistics robots.

[0052] Only some exemplary embodiments of the present invention are described above by way of illustration. Undoubtedly, for those of ordinary skill in the art, various different ways can be used to modify the described embodiments without departing from the spirit and scope of the present invention. Therefore, the above drawings and descriptions are illustrative in nature and should not be construed as limiting the protection scope of the claims of the present invention.

Claims

1. A logistics robot control method based on artificial intelligence, characterized in that: The following steps are involved: Step 1: Each logistics robot connects to the data collection API when starting to perform a task, and automatically records the task data, including the task environment status, task execution status, and task feedback information. All logistics robots upload the task data to the cloud learning platform via the wireless network to build a preliminary experience database; Step 2: Integrate the task data of all logistics robots through the cloud learning platform, and use online learning to perform dynamic training based on real-time task data; Step 3: Establish a multi-dimensional task experience database, divide the upper-level decision-making, automatically select the experience area according to the current logistics status to allocate logistics tasks, optimize the execution control strategy, divide the lower-level feedback, and feed back the experience of the corresponding task data to the cloud in real time according to the logistics tasks; Step 4: Dynamically optimize upper-level decisions and lower-level feedback based on feedback, continuously update to form a closed loop, enable all logistics robots to accumulate experience together when performing tasks, share experience through the cloud learning platform, and optimize the operating strategies of all logistics robots.

2. The artificial intelligence-based logistics robot control method according to claim 1, characterized in that: In the step one, in the process of automatically recording task data, the precise coordinates of the current position of the logistics robot are accurately recorded through the global positioning system and the internal positioning algorithm, the surrounding environment of the logistics robot is detected by the internal sensor group, and the obstacle position is dynamically recorded, and the task execution status is compared with the preset task completion status according to the task target position and the predetermined task information, and the task feedback information is recorded by the internal information transmission group in combination with the task environment status and the task execution status. All logistics robots upload the task data to the cloud learning platform through the wireless network; the preliminary experience database is divided into a first experience area, a second experience area, a third experience area and a fourth experience area, wherein the first experience area includes the execution records of all successfully completed tasks, the second experience area includes the execution records of all failed tasks, the third experience area includes the task execution status of different environmental variables, and the fourth experience area includes the impact of manual intervention on the task execution status.

3. The artificial intelligence-based logistics robot control method according to claim 1, characterized in that: In the step 2, the dynamic training specifically includes: based on the integrated task data set, building a basic deep reinforcement learning model through a cloud learning platform for rapid adaptation, receiving task data uploaded in real time from the logistics robot through online learning and processing it using an incremental learning method, inputting the task execution status as an immediate reward signal into the basic deep reinforcement learning model for online updating, establishing an incremental feedback loop mechanism to immediately update the deep reinforcement learning model weights with new task data after each task is executed, weighting the task data based on environmental variables, giving priority to tasks performed in complex environments, and helping all logistics robots update their learning goals.

4. The artificial intelligence-based logistics robot control method according to claim 3 is characterized in that: The specific formula for weighting task data based on environmental variables is: Among them, w final represents the weight of each task execution state after update, β represents the feedback influence factor of task feedback on task data, E t Indicates the specific indicator of the task execution status at the current timestamp, α t Indicates the time impact factor of the timestamp on the current task execution status, E i,t represents the execution status of the i-th task at timestamp t, E env Indicates the specific indicator of the task execution status in the current environment variable, E j,env represents the jth task execution status when the environment variable env is set, ω represents the time influence factor of the environment variable on the current task execution status, ω pre It represents the weight of each task execution status after the last update, N represents the total number of task data involved in the weight calculation, and M represents the total number of task data involved in the environment variable calculation.

5. The artificial intelligence-based logistics robot control method according to claim 1, characterized in that: In the step three, the establishment of a multi-dimensional task experience database specifically includes: reclassifying the preliminary experience database based on the success rate dimension, including a high success rate area, a medium success rate area and a low success rate area, reclassifying the preliminary experience database based on the environmental condition dimension, including a humidity area and a temperature area, and reclassifying the preliminary experience database based on the task characteristics dimension, including a task type area and a task complexity area.

6. The artificial intelligence-based logistics robot control method according to claim 1, characterized in that: In step 4, the decision algorithm is regularly retrained based on the latest feedback. Once a new control strategy is generated, the logistics robot will immediately receive the updated control strategy, continuously update it to form a closed loop, build a feedback gain model for quantitative improvement, and completely convert the task experience database into a collective knowledge base, so that all logistics robots can accumulate experience together when performing tasks, use collective wisdom to identify common problems and successful control strategies through machine learning algorithms, generate the best experience for reference by all logistics robots, and share experience through a cloud learning platform to optimize the operating strategies of all logistics robots.

7. The artificial intelligence-based logistics robot control method according to claim 6, characterized in that: The specific formula for quantitative improvement of the feedback gain model is: Among them, F(t) represents the feedback value of the current task, α represents the feedback gain factor, F(t-1) represents the feedback value of the previous timestamp, γ represents the historical feedback gain factor, and F k represents the total feedback value of the past k task executions, C t represents the complexity factor of the current task, δ represents the immediate reward factor, and R(t) represents the immediate reward signal for executing the task at time t.

8. The artificial intelligence-based logistics robot control system according to claim 1, characterized in that: Record upload module: Each logistics robot connects to the data collection API when starting to perform a task, automatically recording task data, including task environment status, task execution status, and task feedback information. All logistics robots upload task data to the cloud learning platform via wireless network to build a preliminary experience database; Integrated training module: Integrate the task data of all logistics robots through the cloud learning platform, and use online learning to conduct dynamic training based on real-time task data; Decision feedback module: Establish a multi-dimensional task experience database, divide the upper-level decision-making, automatically select the experience area according to the current logistics status to allocate logistics tasks, optimize the execution control strategy, divide the lower-level feedback according to the logistics tasks, and feed back the experience of the corresponding task data to the cloud in real time; Shared Optimization Module: Dynamically optimize upper-level decisions and lower-level feedback based on feedback, continuously update to form a closed loop, enable all logistics robots to accumulate experience together when performing tasks, and share experience through the cloud learning platform to optimize the operating strategies of all logistics robots.

Images