Intelligent shuttle vehicle with body based on active perception and control method thereof
Through the embodied intelligent shuttle car with integrated active perception and independent decision-making capabilities, the existing four-way shuttle car is solved, and adaptive handling operations are realized, improving the intelligence and robustness of the system.
Patent Information
- Application Number
- CN202510735241.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-06-04
- Publication Date
- 2025-08-26
AI Technical Summary
The existing four-way shuttle vehicles are difficult to adapt to adaptively in dynamic and complex scenarios, resulting in inefficiency and high dependence on manual intervention, making it difficult to deal with changes in workplace layout and temporary obstacles.
The active perception module, the perceived data fusion module, the decision-making self-generating module and the perceived ability self-evolution module are adopted to integrate active perception, autonomous decision-making and efficient learning capabilities. Through self-adjustment of perception strategies, intelligent adjustment of walking movements and online evolution of perceived capabilities, adaptive handling operations are realized.
It significantly improves the operation efficiency and robustness in dynamic scenarios, reduces dependence on manual intervention, and improves the adaptability and intelligence level of the system.
Smart Images
Figure CN120540322A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of warehousing and transportation technology, and in particular to an embodied intelligent shuttle vehicle based on active perception and a control method thereof. Background Art
[0002] With the rapid development of logistics automation technology, four-way shuttles, as efficient material handling equipment, have been widely used in the field of warehousing and logistics. In the existing technology, the optimization of four-way shuttles mainly focuses on modular design (such as walking module, lifting and reversing module) and local improvements of scheduling algorithms (such as path planning, task allocation). However, these methods are usually based on preset rules or fixed environmental parameters, and it is difficult to cope with the challenges of dynamic and complex scenarios. For example, when the layout of the workplace changes, temporary obstacles appear, or task requirements are suddenly adjusted, the existing system needs to rely on manual reprogramming or parameter adjustment, resulting in low efficiency and insufficient adaptability. In response to the above problems, there is an urgent need for an embodied intelligent shuttle that integrates active perception, autonomous decision-making and efficient learning capabilities to achieve adaptive handling operations in complex scenarios, reduce dependence on manual intervention, and promote the leap of logistics automation to a higher level of intelligence. Summary of the Invention
[0003] The present invention provides an embodied intelligent shuttle vehicle based on active perception, comprising: an active perception module, a perception data self-integration module, a shuttle vehicle travel module, a decision self-generation module and a perception capability self-evolution module;
[0004] Active perception module, used to dynamically capture multimodal environmental data based on task-driven perception strategies;
[0005] Perception data fusion module, used to fuse the captured multimodal environmental data;
[0006] The shuttle vehicle travel module is used to drive the shuttle vehicle to move along the queried transport path and determine in real time whether to enter the decision-making self-generation module based on the environmental fusion data;
[0007] A decision-making self-generation module is used to adaptively generate and execute real-time adjustment decisions for walking movements;
[0008] The perception capability self-evolution module is used to evolve the shuttle vehicle's active perception capability online by combining data from multiple vehicles during operation.
[0009] The embodied intelligent shuttle vehicle based on active perception as described above, wherein the active perception module specifically includes: a transport task receiving submodule and a perception strategy self-adjustment submodule;
[0010] The transport task receiving submodule is used to receive the transport tasks issued by the host computer;
[0011] The perception strategy self-adjustment submodule is used to dynamically adjust the perception strategy driven by the transportation task and capture multimodal environmental data.
[0012] In the above-mentioned embodied intelligent shuttle based on active perception, the workflow of the perception strategy self-adjustment submodule is as follows:
[0013] Convert the shuttle's real-time mission state and environment state into a vector respectively;
[0014] Input the converted task vector and environment vector into the intelligent perception model to generate a set of perception strategies;
[0015] Execute the generated perception strategy and transmit the acquired multimodal environment data to the perception data integration module.
[0016] In the above-mentioned embodied intelligent shuttle vehicle based on active perception, the decision self-generation module specifically includes: an input set preparation submodule, a decision generation submodule, and a decision execution submodule;
[0017] The input set preparation submodule is used to generate an input set that can be recognized by the walking action intelligent adjustment model based on the current environment fusion data, the previous walking action, and the current shuttle state;
[0018] The decision generation submodule is used to input the prepared input set into the walking action intelligent adjustment model to obtain real-time adjustment decisions for walking actions;
[0019] The decision execution submodule is used to adjust the decision according to the generated walking action to control the shuttle vehicle to pass through the current section.
[0020] In the above-mentioned embodied intelligent shuttle vehicle based on active perception, the perception capability self-evolution module specifically includes: a multi-vehicle data division submodule, a fusion error calculation submodule, and a fusion weight update submodule;
[0021] The multi-vehicle data partitioning submodule is used to partition the perception data of multiple shuttle vehicles with overlapping transportation paths into the same set;
[0022] The fusion error calculation submodule is used to calculate the fusion error based on the fusion weights of the divided multi-vehicle data and the current multimodal environment data;
[0023] The fusion weight update submodule is used to update the fusion weight of the multimodal environment data online according to the calculated fusion error.
[0024] The present invention also provides a control method for an embodied intelligent shuttle vehicle based on active perception, comprising:
[0025] Step S210: Dynamically capture multimodal environmental data according to a task-driven perception strategy;
[0026] Step S220: fusing the captured multimodal environmental data;
[0027] Step S230: driving the shuttle vehicle to move along the queried transport route, and determining in real time whether the moving action needs to be adjusted based on the environmental fusion data;
[0028] Step S240: If the walking motion needs to be adjusted, the walking motion intelligent adjustment model is used to adaptively generate a real-time adjustment decision for the walking motion and execute it; if the walking motion does not need to be adjusted, the decision is ignored;
[0029] Step S250: Evolving the active perception capability of the shuttle vehicle online by combining data from multiple vehicles during operation.
[0030] The beneficial effects achieved by the present invention are as follows: it has three major characteristics: active perception, autonomous decision-making and efficient learning. Compared with the existing four-way shuttle system, it significantly improves the operating efficiency and robustness in dynamic scenarios. BRIEF DESCRIPTION OF THE DRAWINGS
[0031] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments described in the present invention. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.
[0032] Figure 1 This is a schematic diagram of an embodied intelligent shuttle vehicle based on active perception provided in Example 1 of the present application;
[0033] Figure 2 This is a flow chart of a control method for an embodied intelligent shuttle vehicle based on active perception provided in Example 2 of the present application. DETAILED DESCRIPTION
[0034] The following is a clear and complete description of the technical solutions in the embodiments of the present invention, in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.
[0035] Example 1
[0036] like Figure 1 As shown, the first embodiment of the present application provides an embodied intelligent shuttle vehicle based on active perception, including: an active perception module 11, a perception data self-integration module 12, a shuttle vehicle walking module 13, a decision self-generation module 14, and a perception capability self-evolution module 15;
[0037] Active perception module 11, used to dynamically capture multimodal environmental data according to task-driven perception strategy, specifically including: a transport task receiving submodule and a perception strategy self-adjustment submodule;
[0038] (1) The transport task receiving submodule is used to receive the transport task issued by the host computer;
[0039] The host computer is the central control system in warehousing logistics, which is used to issue delivery tasks to the shuttle vehicles. The delivery tasks must include three data: the cargo location, the quantity of goods to be picked up, and the delivery target location.
[0040] (2) The perception strategy self-adjustment submodule is used to dynamically adjust the perception strategy driven by the transportation task and capture multimodal environmental data;
[0041] This module integrates an intelligent perception model that can generate a set of perception strategies f(t) based on the current task state Task(t) and environment state Envi(t), where t is the current timestamp. The specific workflow of this module is as follows:
[0042] ① Convert the shuttle's real-time mission status and environment status into a vector respectively;
[0043] First, based on the key points from the current point to the cargo point, and then to the delivery target point, the transportation route is queried from the existing on-board map. Then, the status of the current task Task(t) is determined based on the position of the shuttle vehicle on the transportation route. When the vehicle is at the starting position, the task status is "start", when the vehicle is moving on the path, the task status is "walk", when the vehicle stops moving on the path, the task status is "stop", when the vehicle is at the cargo point, the task status is "pick up", when the vehicle is at the delivery target point, the task status is "unload", and when the vehicle leaves the delivery target point, the task status is "end". After determination, it is converted into a task vector τ; the environmental state Envi(t) is the currently perceived environmental information such as obstacle density, ambient light intensity, whether there is interference, etc., which is converted into a vector and represented as δ.
[0044] ② Input the converted task vector and environment vector into the intelligent perception model to generate a set of perception strategies;
[0045] The intelligent perception model is expressed as:
[0046] Where f(t) is the output perception strategy, α is the activation strategy set of the sensor, and α i is the i-th data item in α, when its value is 1, it means the i-th sensor is enabled, and when it is 0, it means the i-th sensor is disabled. β is the sensor configuration parameter policy set. iis the i-th vector in β, β i Each component in is the value of each configuration parameter of the i-th sensor, γ is the fusion vector of multimodal environmental data, γ i is the i-th component in γ, which represents the fusion weight of the i-th sensor data, τ and δ are the input task vector and environment vector, ε(β i ) is the power consumption generated by the configuration of the i-th sensor, is the acquisition delay caused by the configuration of the i-th sensor, E fuse (γ) is the weight vector γ for the fusion error of multimodal data, ω1, ω2, ω3, and ω4 are the reward factors for perception accuracy, energy consumption, end-to-end delay, and fusion error, respectively. i ranges from 1 to N, where N is the number of on-board sensors. It is used to return the (α, β, γ) when the result of the expression in the brackets is the smallest. where φ k is the kth configuration parameter of the i-th sensor The efficiency factor, k ranges from 1 to K, where K is the total number of configuration parameters for the i-th sensor, IoU(·) is the intersection-over-union function, and SensorCap i is the detectable range of the i-th sensor, Task Req τ is the detection requirement range of the task vector τ, is the current environment vector δ and the optimal working environment vector of the i-th sensor It should be noted that in the initial state of the fusion vector γ, that is, before self-evolution, each component is customized (by setting a different fusion vector for each vehicle to reduce the cold start time of the active perception module), and the corresponding E fuse (γ) is also assigned a value only after the self-evolution begins, and is initially set to 1. The detection requirement range of the task vector τ comes from a system preset table, which maintains the detection requirement ranges corresponding to various transportation tasks.
[0047] ③Execute the generated perception strategy and transmit the acquired multimodal environment data to the perception data integration module;
[0048] Based on the activation strategy set α in the perception strategy f(t), the sensor is started or shut down, and then each sensor is configured according to the configuration parameter strategy set β. Then, the configured sensors are used to capture real-time multimodal environmental data. Finally, the captured multimodal environmental data and the fusion weight vector γ are transmitted to the perception data integration module.
[0049] The perception data fusion module 12 is used to fuse the captured multimodal environmental data;
[0050] The captured multimodal environment data is weightedly fused using the received fusion weight vector, and the fused data is transmitted to the shuttle walking module.
[0051] The shuttle vehicle moving module 13 is used to drive the shuttle vehicle to move along the queried transport path and determine in real time whether to enter the decision-making self-generation module 14 based on the environmental fusion data;
[0052] The basis for judging whether to enter the decision-making self-production module is whether the total difference between the current fusion data and the previous fusion data is greater than the threshold (system preset). If it is greater, it enters the decision-making self-generation module; if it is less, it does not enter.
[0053] The decision self-generation module 14 is used to adaptively generate and execute real-time adjustment decisions for walking actions, specifically including: an input set preparation submodule, a decision generation submodule, and a decision execution submodule;
[0054] (1) Input set preparation submodule, which is used to generate an input set that can be recognized by the walking action intelligent adjustment model based on the current environment fusion data, the previous walking action, and the current shuttle state;
[0055] The current environment fusion data is recorded as F t (including the detected obstacle position), the previous walking action (such as acceleration, steering angle action coding) is recorded as A t-1 The current state of the shuttle (the vector composed of position, speed and heading angle) is recorded as V t , then the generated input set is represented as X=[F t ,A t-1 ,V t ], t is the current timestamp, and t-1 is the timestamp when the previous walking action was performed.
[0056] (2) a decision generation submodule, which is used to input the prepared input set into the walking action intelligent adjustment model to obtain real-time adjustment decisions for walking actions;
[0057] The walking action intelligent adjustment model is expressed as: Where Y is the output walking action real-time adjustment decision, X is the input set, Used to return the walking action A when the calculation result in the brackets is the largest t , A t ∈A, A is the discrete action space (including all walking action encodings), is an adjustable parameter, in is the adjustable weight, d jis the distance between the current position of the shuttle and the jth obstacle, j ranges from 1 to M, M is the number of obstacles detected, v is the current speed of the shuttle, θ is the angle deviation between the current heading and the target path, Indicates execution of action A t-1 and A t The acceleration and steering angular velocity of the shuttle change;
[0058] (3) Decision execution submodule, which is used to adjust the decision and control the shuttle to pass through the current road section according to the generated walking action;
[0059] The generated walking action adjustment decision is converted into a control command to drive the shuttle to move according to the decision to pass the current road section.
[0060] The perception capability self-evolution module 15 is used to online evolve the shuttle vehicle's active perception capability by combining multi-vehicle data during operation. Specifically, it includes: a multi-vehicle data division submodule, a fusion error calculation submodule, and a fusion weight update submodule;
[0061] (1) Multi-vehicle data partitioning submodule, used to partition the perception data of multiple shuttle vehicles with overlapping transport paths into the same set;
[0062] Perception data is the multimodal environmental data captured by the sensors on the shuttle. After being divided according to the transportation path, each set contains the perception data of multiple shuttles in the same area at different times. These sets are the basis of online evolutionary perception capabilities.
[0063] (2) Fusion error calculation submodule, which is used to calculate the fusion error based on the fusion weights of the divided multi-vehicle data and the current multimodal environment data;
[0064] Bring the divided set and the fusion weight of the current multimodal environment data into the calculation formula: In the fusion error E fuse (γ), where They represent the local environment map estimates of cars c and g in the u-th set, respectively. g∈Γ(c), Γ(c) is the set of other shuttles in the u-th set except car c, c ranges from 1 to uL, uL is the total number of shuttles in the u-th set, Ω cg represents the covariance matrix of the observation overlap region between cars c and g, λ is the regularization coefficient, Tr(·) is the trace function of the matrix, represents the fusion weight of the i-th sensor data on vehicle c, P i (c) It represents the covariance matrix estimated by the i-th sensor on vehicle c, i ranges from 1 to N, N is the total number of onboard sensors, and u ranges from 1 to U, U is the total number of divided sets.
[0065] (3) Fusion weight update submodule, which is used to update the fusion weight of multimodal environment data online according to the calculated fusion error;
[0066] This module updates the fusion weights of the sensor data on the ego vehicle based on the fusion error. The fusion weight update formula of the i-th sensor data on the ego vehicle is expressed as: in is the updated fusion weight, is the fusion weight before updating, E fuse (γ) is the fusion error, and ψ is the learning rate of the intelligent perception model. After updating the fusion weights of all sensor data, a new weight vector γ is obtained. The new weight vector γ is used to adjust the aforementioned intelligent perception model in real time to achieve the effect of online updating of perception capabilities.
[0067] Example 2
[0068] like Figure 2 As shown, the second embodiment of the present application provides a control method for an embodied intelligent shuttle vehicle based on active perception, including:
[0069] Step S210: Dynamically capture multimodal environmental data according to the task-driven perception strategy, which is specifically divided into the following sub-steps:
[0070] Step S211: receiving the transport task issued by the host computer;
[0071] The host computer is the central control system in warehousing logistics, which is used to issue delivery tasks to the shuttle vehicles. The delivery tasks must include three data: the cargo location, the quantity of goods to be picked up, and the delivery target location.
[0072] Step S212: Dynamically adjust the perception strategy driven by the transport task to capture multimodal environmental data;
[0073] The dynamic adjustment perception strategy is generated by the intelligent perception model. The specific implementation process is as follows:
[0074] ① Convert the shuttle's real-time mission status and environment status into a vector respectively;
[0075] First, based on the key points from the current point to the cargo point, and then to the delivery target point, the transportation route is queried from the existing on-board map. Then, the status of the current task Task(t) is determined based on the position of the shuttle vehicle on the transportation route. When the vehicle is at the starting position, the task status is "start", when the vehicle is moving on the path, the task status is "walk", when the vehicle stops moving on the path, the task status is "stop", when the vehicle is at the cargo point, the task status is "pick up", when the vehicle is at the delivery target point, the task status is "unload", and when the vehicle leaves the delivery target point, the task status is "end". After determination, it is converted into a task vector τ; the environmental state Envi(t) is the currently perceived environmental information such as obstacle density, ambient light intensity, whether there is interference, etc., which is converted into a vector and represented as δ.
[0076] ② Input the converted task vector and environment vector into the intelligent perception model to generate a set of perception strategies;
[0077] ③Execute the generated perception strategy and transmit the acquired multimodal environment data to the perception data integration module;
[0078] Based on the activation policy set in the perception policy, the sensor is started or shut down, and then each sensor is configured according to the configuration parameter policy set. Then, the configured sensors are used to capture real-time multimodal environmental data.
[0079] Step S220: fusing the captured multimodal environmental data;
[0080] The captured multimodal environmental data is weightedly fused using the fusion weight vector output by the intelligent perception model to obtain environmental fusion data.
[0081] Step S230: driving the shuttle vehicle to move along the queried transport route, and determining in real time whether the moving action needs to be adjusted based on the environmental fusion data;
[0082] The basis for judging whether the walking action needs to be adjusted is whether the total difference between the current fusion data and the previous fusion data is greater than the threshold (system preset). If it is greater, the walking action needs to be adjusted; if it is less, it is not necessary.
[0083] Step S240: If the walking motion needs to be adjusted, the walking motion intelligent adjustment model is used to adaptively generate a real-time adjustment decision for the walking motion and execute it; if the walking motion does not need to be adjusted, the decision is ignored;
[0084] The intelligent walking motion adjustment model is used to adaptively generate and execute real-time walking motion adjustment decisions. The specific steps are as follows:
[0085] Step S241: Generate an input set recognizable by the walking motion intelligent adjustment model based on the current environment fusion data, the previous walking motion, and the current shuttle vehicle state;
[0086] The current environment fusion data is recorded as F t (including the detected obstacle position), the previous walking action (such as acceleration, steering angle action coding) is recorded as A t-1 The current state of the shuttle (the vector composed of position, speed and heading angle) is recorded as V t , then the generated input set is represented as X=[F t ,A t-1 ,V t ], t is the current timestamp, and t-1 is the timestamp when the previous walking action was performed.
[0087] Step S242: inputting the prepared input set into the walking motion intelligent adjustment model to obtain a real-time adjustment decision for the walking motion;
[0088] Step S243: adjusting the decision to control the shuttle vehicle to pass through the current road section according to the generated walking action;
[0089] The generated walking action adjustment decision is converted into a control command to drive the shuttle to move according to the decision to pass the current road section.
[0090] Step S250: Evolving the shuttle's active perception capability online by combining data from multiple vehicles during operation. This is specifically divided into the following sub-steps:
[0091] Step S251: grouping the sensing data of multiple shuttle vehicles with overlapping transport paths into the same set;
[0092] Step S252: Calculating a fusion error based on the divided multi-vehicle data and the fusion weights of the current multimodal environment data;
[0093] Step S253: updating the fusion weight of the multimodal environment data online according to the calculated fusion error;
[0094] This module updates the fusion weights of the sensor data on the ego vehicle based on the fusion error. The fusion weight update formula of the i-th sensor data on the ego vehicle is expressed as: in is the updated fusion weight, is the fusion weight before updating, E fuse (γ) is the fusion error, and ψ is the learning rate of the intelligent perception model. After updating the fusion weights of all sensor data, a new weight vector γ is obtained. The new weight vector γ is used to adjust the aforementioned intelligent perception model in real time to achieve the effect of online updating of perception capabilities.
[0095] Corresponding to the above embodiment, an embodiment of the present invention provides a computer storage medium, comprising: at least one memory and at least one processor;
[0096] The memory is used to store one or more program instructions;
[0097] A processor is used to run one or more program instructions to execute a control method of an embodied intelligent shuttle vehicle based on active perception.
[0098] Corresponding to the above embodiment, an embodiment of the present invention provides a computer-readable storage medium, which contains one or more program instructions, and the one or more program instructions are used by a processor to execute a method for controlling an embodied intelligent shuttle vehicle based on active perception.
[0099] The embodiments disclosed in the present invention provide a computer-readable storage medium, in which computer program instructions are stored. When the computer program instructions are executed on a computer, the computer executes the above-mentioned active perception-based embodied intelligent shuttle control method.
[0100] In the embodiments of the present invention, the processor may be an integrated circuit chip having signal processing capabilities. The processor may be a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.
[0101] The methods, steps, and logic diagrams disclosed in the embodiments of the present invention can be implemented or executed. A general-purpose processor can be a microprocessor or any conventional processor. The steps of the methods disclosed in the embodiments of the present invention can be directly implemented and executed by a hardware decoding processor, or by a combination of hardware and software modules within the decoding processor. The software modules can be located in a storage medium well-established in the art, such as random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, or registers. The processor reads the information from the storage medium and, in conjunction with its hardware, completes the steps of the aforementioned methods.
[0102] The storage medium may be a memory and may be, for example, a volatile memory or a nonvolatile memory, or may include both volatile and nonvolatile memory.
[0103] Among them, the non-volatile memory can be a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), or a flash memory.
[0104] Volatile memory may be random access memory (RAM), which is used as an external cache memory. By way of example and not limitation, many forms of RAM are available, such as static random access memory (SRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double data rate synchronous dynamic random access memory (DDRSDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous link dynamic random access memory (SLDRAM), and direct RAM bus random access memory (DRRAM).
[0105] The storage media described in the embodiments of the present invention are intended to include, but are not limited to, these and any other suitable types of memory.
[0106] Those skilled in the art will appreciate that in one or more of the above examples, the functions described herein can be implemented using a combination of hardware and software. When software is used, the corresponding functions can be stored in a computer-readable medium or transmitted as one or more instructions or codes on a computer-readable medium. Computer-readable media include computer storage media and communication media, wherein communication media includes any medium that facilitates the transmission of computer programs from one place to another. The storage medium can be any available medium that can be accessed by a general-purpose or special-purpose computer.
[0107] The specific implementation methods described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific implementation method of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent replacements, improvements, etc. made on the basis of the technical solution of the present invention should be included in the scope of protection of the present invention.
Claims
1. An embodied intelligent shuttle vehicle based on active perception, characterized in that: include: Active perception module, perception data self-integration module, shuttle vehicle travel module, decision self-generation module and perception capability self-evolution module; Active perception module, used to dynamically capture multimodal environmental data based on task-driven perception strategies; Perception data fusion module, used to fuse the captured multimodal environmental data; The shuttle vehicle travel module is used to drive the shuttle vehicle to move along the queried transport path and determine in real time whether to enter the decision-making self-generation module based on the environmental fusion data; A decision-making self-generation module is used to adaptively generate and execute real-time adjustment decisions for walking movements; The perception capability self-evolution module is used to evolve the shuttle vehicle's active perception capability online by combining data from multiple vehicles during operation.
2. The embodied intelligent shuttle vehicle based on active perception according to claim 1, characterized in that: The active perception module specifically includes: a transport task receiving submodule and a perception strategy self-adjustment submodule; The transport task receiving submodule is used to receive the transport tasks issued by the host computer; The perception strategy self-adjustment submodule is used to dynamically adjust the perception strategy driven by the transportation task and capture multimodal environmental data.
3. The embodied intelligent shuttle vehicle based on active perception according to claim 2, characterized in that: The workflow of the perception strategy self-adjustment submodule is as follows: Convert the shuttle's real-time mission state and environment state into a vector respectively; Input the converted task vector and environment vector into the intelligent perception model to generate a set of perception strategies; Execute the generated perception strategy and transmit the acquired multimodal environment data to the perception data integration module.
4. The embodied intelligent shuttle vehicle based on active perception according to claim 1, characterized in that: The basis for judging whether to enter the decision self-production module is whether the total difference between the current fusion data and the previous fusion data is greater than the threshold. If it is greater, the decision self-generation module is entered; if it is less, it is not entered.
5. The embodied intelligent shuttle vehicle based on active perception according to claim 1, characterized in that: The decision self-generation module specifically includes: input set preparation submodule, decision generation submodule and decision execution submodule; The input set preparation submodule is used to generate an input set that can be recognized by the walking action intelligent adjustment model based on the current environment fusion data, the previous walking action, and the current shuttle state; The decision generation submodule is used to input the prepared input set into the walking action intelligent adjustment model to obtain real-time adjustment decisions for walking actions; The decision execution submodule is used to adjust the decision according to the generated walking action to control the shuttle vehicle to pass through the current section.
6. The embodied intelligent shuttle vehicle based on active perception according to claim 1, characterized in that: The perception capability self-evolution module specifically includes: multi-vehicle data division submodule, fusion error calculation submodule, and fusion weight update submodule; The multi-vehicle data partitioning submodule is used to partition the perception data of multiple shuttle vehicles with overlapping transportation paths into the same set; The fusion error calculation submodule is used to calculate the fusion error based on the fusion weights of the divided multi-vehicle data and the current multimodal environment data; The fusion weight update submodule is used to update the fusion weight of the multimodal environment data online according to the calculated fusion error.
7. A control method for an embodied intelligent shuttle vehicle based on active perception, characterized in that: include: Step S210: Dynamically capture multimodal environmental data according to a task-driven perception strategy; Step S220: fusing the captured multimodal environmental data; Step S230: driving the shuttle vehicle to move along the queried transport route, and determining in real time whether the moving action needs to be adjusted based on the environmental fusion data; Step S240: If the walking motion needs to be adjusted, the walking motion intelligent adjustment model is used to adaptively generate a real-time adjustment decision for the walking motion and execute it; if the walking motion does not need to be adjusted, the decision is ignored; Step S250: Evolving the active perception capability of the shuttle vehicle online by combining data from multiple vehicles during operation.
8. A computer storage medium, characterized in that include: at least one memory and at least one processor; a memory for storing one or more program instructions; A processor is configured to run one or more program instructions to execute the active perception-based embodied intelligent shuttle control method as claimed in claim 7.