Elastic resource scheduling method and device for robot navigation, electronic equipment and medium

Abstract

The invention provides an elastic resource scheduling method and device for robot navigation, electronic equipment and a medium. The method comprises the following steps: acquiring environment sensing data and task state data of a robot; according to the environment perception data, generating a dynamic environment risk field representing the level of the space risk; determining the navigation situation category of the robot according to the risk value distribution of the dynamic environment risk field; according to the navigation situation category and the task state data, determining real-time criticality levels of the plurality of navigation subtasks of the robot; dynamically allocating corresponding computing resources according to the real-time criticality level; according to the computing resources, an algorithm instance matched with the level of the computing resources is called and executed to execute the navigation task. According to the method, danger can be sensed in real time through the dynamic environment risk field, a large amount of computing power is injected for planning and controlling tasks at the critical moment, the rapid response capacity and obstacle avoidance reliability of the robot in the complex dynamic environment are ensured, and the risk of collision accidents is greatly reduced.

Description

Technical Field

[0001] This application relates to the field of robot autonomous navigation technology, and in particular to a method, apparatus, electronic device and medium for flexible resource scheduling of robot navigation. Background Technology

[0002] With the widespread application of robotics in warehousing and logistics, intelligent inspection, and service reception, the reliability of robot navigation in complex and dynamic environments faces severe challenges. Traditional robot navigation systems typically employ a fixed allocation strategy for computing resources, meaning that regardless of whether the robot is in an open area or a dense obstacle zone, its localization, path planning, and control algorithms operate with the same precision and frequency. This homogenized resource allocation method has significant drawbacks: in safe areas, the continuous operation of high-performance algorithms leads to a huge waste of computing resources, increasing system power consumption and heat dissipation burden, and shortening the battery life of mobile devices; while in dangerous areas with dense obstacles, fixed computing resources may not be able to meet sudden high computing demands, resulting in path planning delays and slow response to control commands, thereby increasing the risk of collisions. Especially in human-robot collaborative environments, the appearance of dynamic obstacles (such as pedestrians and other robots) is unpredictable, requiring the system to have millisecond-level real-time response capabilities.

[0003] Current technologies lack effective quantitative assessments of environmental risks and corresponding resource scheduling mechanisms, making it impossible to achieve efficient utilization of computing resources while ensuring navigation safety. Therefore, there is an urgent need for a navigation method capable of intelligently sensing environmental risks and dynamically adjusting computing resource allocation accordingly to resolve the contradiction between resource waste and real-time performance assurance. Summary of the Invention

[0004] In view of this, embodiments of this application provide a flexible resource scheduling method, apparatus, electronic device and medium for robot navigation, in order to solve the technical problem in the prior art that there is a lack of effective quantitative assessment of environmental risks and a resource scheduling mechanism linked to them, which makes it impossible to achieve efficient utilization of computing resources while ensuring navigation safety.

[0005] A first aspect of this application provides a flexible resource scheduling method for robot navigation, the method comprising: Acquire the robot's environmental perception data and task status data; Based on the environmental perception data, a dynamic environmental risk field characterizing the level of spatial risk is calculated and generated; Based on the risk value distribution of the dynamic environmental risk field, the navigation situation category of the robot is determined, and the navigation situation category includes at least critical situation and relaxed situation; Based on the navigation situation category and the task status data, determine the real-time criticality level of multiple navigation sub-tasks of the robot; Based on the real-time criticality level, the corresponding computing resources are dynamically allocated; Based on the computing resources, invoke and execute algorithm instances that match the computing resource level to perform the navigation task; Based on the execution results of the algorithm instance, motion control commands for the robot are output; simultaneously, based on the execution effect of the motion control commands and resource usage, the calculation parameters of the dynamic environment risk field and the allocation strategy of the computing resources are optimized.

[0006] A second aspect of this application provides a flexible resource scheduling device for robot navigation, comprising: The data acquisition module is used to acquire the robot's environmental perception data and task status data; The dynamic environmental risk field generation module is used to calculate and generate a dynamic environmental risk field that characterizes the level of spatial risk based on the environmental perception data. The navigation situation category determination module is used to determine the navigation situation category of the robot based on the risk value distribution of the dynamic environment risk field. The navigation situation category includes at least critical situations and relaxed situations. The real-time criticality level determination module is used to determine the real-time criticality level of multiple navigation sub-tasks of the robot based on the navigation situation category and the task status data. The computing resource allocation module is used to dynamically allocate corresponding computing resources according to the real-time criticality level; An execution module is used to invoke and execute an algorithm instance that matches the computing resource level, based on the computing resources, to perform a navigation task; The optimization module is used to output motion control commands for the robot based on the execution results of the algorithm instance; at the same time, it performs feedback optimization on the calculation parameters of the dynamic environment risk field and the allocation strategy of the computing resources based on the execution effect of the motion control commands and the resource usage.

[0007] A third aspect of this application provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the above-described method.

[0008] A fourth aspect of this application provides a readable storage medium storing a computer program that, when executed by a processor, implements the steps of the above-described method.

[0009] The beneficial effects of this application's embodiments compared to existing technologies are as follows: This application's embodiments acquire the robot's environmental perception data and task status data; based on the environmental perception data, calculate and generate a dynamic environmental risk field characterizing the level of spatial risk; determine the robot's navigation situation category based on the risk value distribution of the dynamic environmental risk field, the navigation situation category including at least critical and relaxed situations; determine the real-time criticality level of multiple navigation sub-tasks of the robot based on the navigation situation category and task status data; dynamically allocate corresponding computing resources based on the real-time criticality level; call and execute algorithm instances matching the computing resource level to perform navigation tasks based on the computing resources; output the robot's motion control commands based on the execution results of the algorithm instances; and simultaneously, optimize the calculation parameters of the dynamic environmental risk field and the allocation strategy of computing resources based on the execution effect of the motion control commands and resource usage. This application perceives dangers in real time through the dynamic environmental risk field and injects a large amount of computing power into planning and control tasks at critical moments, ensuring the robot's rapid response capability and obstacle avoidance reliability in complex dynamic environments, greatly reducing the risk of collision accidents. Breaking away from the rigid model of fixed resource allocation, the system automatically switches to a low-power operating state in safe areas, saving significant computing resources and effectively reducing overall system power consumption, making it particularly suitable for battery-powered mobile robots. The introduced feedback optimization mechanism allows the system to learn from practical operating experience and continuously adjust its parameters, thereby better adapting to different application scenarios and environmental characteristics, demonstrating good robustness and generalization ability. Attached Figure Description

[0010] To more clearly illustrate the technical solutions in the embodiments of this application, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0011] Figure 1 This is a schematic diagram illustrating an application scenario of this application embodiment; Figure 2 This is a flowchart illustrating a flexible resource scheduling method for robot navigation provided in an embodiment of this application; Figure 3 This is a schematic diagram of a flexible resource scheduling device for robot navigation provided in an embodiment of this application; Figure 4 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0012] In the following description, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of this application. However, those skilled in the art will understand that this application may also be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, circuits, and methods have been omitted so as not to obscure the description of this application with unnecessary detail.

[0013] With the widespread application of robotics in warehousing and logistics, intelligent inspection, and service reception, the reliability of robot navigation in complex and dynamic environments faces severe challenges. Traditional robot navigation systems typically employ a fixed allocation strategy for computing resources, meaning that regardless of whether the robot is in an open area or a dense obstacle zone, its localization, path planning, and control algorithms operate with the same precision and frequency. This homogenized resource allocation method has significant drawbacks: in safe areas, the continuous operation of high-performance algorithms leads to a huge waste of computing resources, increasing system power consumption and heat dissipation burden, and shortening the battery life of mobile devices; while in dangerous areas with dense obstacles, fixed computing resources may not be able to meet sudden high computing demands, resulting in path planning delays and slow response to control commands, thereby increasing the risk of collisions. Especially in human-robot collaborative environments, the appearance of dynamic obstacles (such as pedestrians and other robots) is unpredictable, requiring the system to have millisecond-level real-time response capabilities.

[0014] Current technologies lack effective quantitative assessments of environmental risks and corresponding resource scheduling mechanisms, making it impossible to achieve efficient utilization of computing resources while ensuring navigation safety. Therefore, there is an urgent need for a navigation method capable of intelligently sensing environmental risks and dynamically adjusting computing resource allocation accordingly to resolve the contradiction between resource waste and real-time performance assurance.

[0015] In view of the problems in the prior art, this application provides a novel flexible resource scheduling method for robot navigation. This method involves acquiring the robot's environmental perception data and task status data; calculating and generating a dynamic environmental risk field characterizing the level of spatial risk based on the environmental perception data; determining the robot's navigation situation category based on the risk value distribution of the dynamic environmental risk field, where the navigation situation category includes at least critical and relaxed situations; determining the real-time criticality level of multiple navigation sub-tasks based on the navigation situation category and task status data; dynamically allocating corresponding computing resources based on the real-time criticality level; calling and executing algorithm instances that match the computing resource level to perform navigation tasks based on the computing resources; outputting motion control commands for the robot based on the execution results of the algorithm instances; and simultaneously optimizing the calculation parameters of the dynamic environmental risk field and the allocation strategy of computing resources based on the execution effect of the motion control commands and resource usage. This application achieves real-time hazard perception through the dynamic environmental risk field and injects significant computing power into planning and control tasks at critical moments, ensuring the robot's rapid response capability and obstacle avoidance reliability in complex dynamic environments, and greatly reducing the risk of collision accidents. Breaking away from the rigid model of fixed resource allocation, the system automatically switches to a low-power operating state in safe areas, saving significant computing resources and effectively reducing overall system power consumption, making it particularly suitable for battery-powered mobile robots. The introduced feedback optimization mechanism allows the system to learn from practical operating experience and continuously adjust its parameters, thereby better adapting to different application scenarios and environmental characteristics, demonstrating good robustness and generalization ability.

[0016] The following will describe in detail, with reference to the accompanying drawings, a method and apparatus for flexible resource scheduling of robot navigation according to embodiments of this application.

[0017] Figure 1 This is a schematic diagram illustrating an application scenario according to an embodiment of this application. The application scenario may include terminal devices 101, 102, and 103, server 104, and network 105.

[0018] Terminal devices 101, 102, and 103 can be hardware or software. When terminal devices 101, 102, and 103 are hardware, they can be various electronic devices with displays that support communication with server 104, including but not limited to smartphones, tablets, laptops, and desktop computers. When terminal devices 101, 102, and 103 are software, they can be installed in the aforementioned electronic devices. Terminal devices 101, 102, and 103 can be implemented as multiple software programs or software modules, or as a single software program or software module; this application embodiment does not impose any limitations on this. Furthermore, various applications can be installed on terminal devices 101, 102, and 103, such as data processing applications, instant messaging tools, social platform software, search applications, shopping applications, etc.

[0019] Server 104 can be a server that provides various services, such as a backend server that receives requests sent by terminal devices with which it has established communication connections. This backend server can receive and analyze the requests sent by the terminal devices and generate processing results. Server 104 can be a single server, a server cluster consisting of several servers, or a cloud computing service center. This application embodiment does not limit this.

[0020] It should be noted that server 104 can be either hardware or software. When server 104 is hardware, it can be various electronic devices that provide various services to terminal devices 101, 102, and 103. When server 104 is software, it can be multiple software programs or software modules that provide various services to terminal devices 101, 102, and 103, or it can be a single software program or software module that provides various services to terminal devices 101, 102, and 103. This application embodiment does not impose any limitations on this.

[0021] Network 105 can be a wired network using coaxial cable, twisted pair, and fiber optic connection, or it can be a wireless network that enables interconnection of various communication devices without wiring, such as Bluetooth, Near Field Communication (NFC), and Infrared. This application embodiment does not limit this.

[0022] Users can establish a communication connection with server 104 via network 105 through terminal devices 101, 102, and 103 to receive or send information. Specifically, server 104 acquires the robot's environmental perception data and task status data; based on the environmental perception data, it calculates and generates a dynamic environmental risk field characterizing the level of spatial risk; based on the risk value distribution of the dynamic environmental risk field, server 104 determines the robot's navigation situation category, which includes at least critical and relaxed situations; based on the navigation situation category and task status data, it determines the real-time criticality level of multiple navigation sub-tasks of the robot; server 104 dynamically allocates corresponding computing resources based on the real-time criticality level; based on the computing resources, it calls and executes algorithm instances that match the computing resource level to perform navigation tasks; based on the execution results of the algorithm instances, server 104 outputs motion control commands for the robot; simultaneously, based on the execution effect of the motion control commands and resource usage, it performs feedback optimization on the calculation parameters of the dynamic environmental risk field and the allocation strategy of computing resources.

[0023] It should be noted that the specific types, quantities, and combinations of terminal devices 101, 102, and 103, server 104, and network 105 can be adjusted according to the actual needs of the application scenario, and this application embodiment does not impose any restrictions on this.

[0024] Figure 2 This is a flowchart illustrating a flexible resource scheduling method for robot navigation provided in an embodiment of this application. Figure 2 The elastic resource scheduling method for robot navigation can be derived from... Figure 1 The terminal device or server executes the command. For example... Figure 2 As shown, the flexible resource scheduling method for robot navigation includes: S201, acquire the robot's environmental perception data and task status data; S202, Based on environmental perception data, calculate and generate a dynamic environmental risk field that characterizes the level of spatial risk; S203. Based on the risk value distribution of the dynamic environmental risk field, determine the navigation situation category of the robot. The navigation situation category includes at least critical situation and relaxed situation. S204, based on the navigation situation category and task status data, determine the real-time criticality level of multiple navigation sub-tasks of the robot; S205 dynamically allocates corresponding computing resources based on the real-time criticality level; S206, Based on computing resources, invoke and execute an algorithm instance that matches the computing resource level to perform the navigation task; S207: Based on the execution results of the algorithm instance, output the robot's motion control commands; at the same time, based on the execution effect of the motion control commands and the resource usage, perform feedback optimization on the calculation parameters of the dynamic environment risk field and the allocation strategy of computing resources.

[0025] Specifically, in this embodiment, environmental perception data refers to comprehensive information about the surrounding environment collected by the robot's multimodal sensor system. This information mainly includes static obstacle information, dynamic object information, and semantic information. Specifically, static obstacle information describes the geometric features and spatial distribution of fixed objects in the environment; dynamic object information includes real-time parameters such as the motion vector, velocity, and orientation angle of moving targets; and semantic information uses a deep learning model to understand the scene and identify specific functional areas and traffic signs. Task status data reflects the robot's internal operating status, including core parameters such as the current navigation target, remaining battery power, system load, and task urgency. This step is crucial because it establishes the data foundation for all subsequent decisions. In practical applications, when a warehouse robot performs a handling task, it needs to simultaneously perceive the position of the shelves (static obstacles), the movement trajectory of the forklift (dynamic objects), and identify the loading and unloading area (semantic information). The organic combination of these data constitutes a complete understanding of the environmental situation.

[0026] Furthermore, the dynamic environmental risk field is a virtual potential field model constructed based on physical space. It quantifies various risk factors in the environment into continuously distributed risk values ​​using mathematical methods. The construction of this risk field includes three levels: the static risk layer calculates the Euclidean distance from the robot to each static obstacle based on the nearest neighbor distance algorithm, with the risk value increasing exponentially as the distance increases; the dynamic risk layer predicts the future trajectory of moving objects using Kalman filtering or particle filtering algorithms and calculates the collision probability based on a spatiotemporal intersection model; the semantic risk layer assigns basic risk weights to different functional areas based on a pre-established environmental knowledge base, with areas such as elevator entrances and stairwells receiving higher weights. These three levels of risk are fused through an adaptive weighting algorithm to ultimately generate a dynamic field that comprehensively reflects the distribution of environmental risks. The innovation of step S202 lies in transforming the discrete risk judgment in traditional navigation into a continuous and computable risk field, enabling the robot to perceive changes in risk gradients in the environment more precisely. For example, in a human-machine hybrid workshop, robots can not only perceive the static risks of the equipment itself, but also assess the dynamic risks brought about by personnel movement, and combine the semantic risks of regional functions (such as assembly areas and passageways) to form a comprehensive risk assessment of the environment.

[0027] Furthermore, the navigation situation category is a qualitative assessment of the robot's current operating environment based on a dynamic environmental risk field. This embodiment uses a dual-threshold mechanism to classify situations: a high-risk threshold identifies dangerous situations requiring immediate action; when the risk value of the robot's location and nearby path points consistently exceeds this threshold, the system determines it has entered a critical situation. A low-risk threshold identifies safe situations where operation can be relaxed; when the risk value of the area surrounding the robot consistently falls below this threshold, the system determines it is in a relaxed situation. This dual-threshold design avoids frequent state switching that might occur with a single threshold while ensuring the accuracy and stability of situation assessment. In actual operation, when the robot is traveling in a warehouse aisle and a moving forklift appears ahead, the risk value on its path quickly exceeds the high-risk threshold, triggering a critical situation. Conversely, when the robot is traveling in an open storage area with no obstacles, the risk value remains below the low-risk threshold, and the system maintains a relaxed situation. This step transforms continuous risk values ​​into discrete decision states, providing a clear basis for subsequent differentiated resource scheduling strategies and effectively solving the problem of unreasonable resource allocation caused by unclear environmental assessments in traditional methods.

[0028] Furthermore, the real-time criticality assessment of navigation subtasks is a multi-factor decision-making process that requires comprehensive consideration of external environmental risks and internal task requirements. This embodiment decomposes the navigation system into three core subtasks: the localization subtask is responsible for maintaining the robot's pose estimation, and its criticality depends on the localization accuracy requirements and the richness of environmental features; the path planning subtask is responsible for generating safe and achievable motion trajectories, and its criticality is affected by environmental complexity and path constraints; the motion control subtask is responsible for executing trajectory tracking control, and its criticality is determined by control accuracy requirements and system dynamic characteristics. Under critical situations, the path planning and motion control subtasks are assigned the highest criticality levels because quickly generating obstacle avoidance trajectories and accurately executing control commands are the primary tasks to ensure safety. Under relaxed situations, all subtasks can be set to lower criticality levels to optimize resource utilization. The innovation of this step lies in establishing a dynamic mapping relationship between environmental situations and task criticality, enabling resource allocation to accurately match real-time needs. For example, when the robot encounters a pedestrian in a narrow passage, the system prioritizes the computational resources for path planning and motion control to ensure timely generation of obstacle avoidance strategies; after obstacle avoidance is completed, the system restores the normal resource allocation ratio.

[0029] Furthermore, for subtasks with the highest criticality level, the system allocates dedicated computing cores from the computing resource pool, guarantees their memory access bandwidth, and increases their operating frequency to ensure these tasks receive stable and sufficient computing power. For subtasks with lower criticality levels, a resource-sharing mode is adopted, scheduling them to a shared computing core pool. Resource consumption is controlled by limiting maximum CPU utilization and memory bandwidth, and they are allowed to reduce their operating frequency or use frame skipping while meeting basic functional requirements. This differentiated resource allocation strategy is implemented through the operating system's real-time scheduler, ensuring that high-criticality tasks receive priority access to computing resources. In a practical system, when the robot is in a critical situation, the path planning task may be allocated a dedicated GPU core to accelerate trajectory optimization calculations, while the motion control task receives a high-frequency CPU core to ensure control frequency; the localization task is restricted to running on a shared core at a lower frequency. This step achieves flexible allocation of computing power through hardware reconfiguration, effectively solving the efficiency bottleneck problem of the fixed resource allocation mode.

[0030] Furthermore, the dynamic selection of algorithm instances is a key mechanism for achieving resource adaptability. This embodiment prepares multiple algorithm implementations with varying computational complexities for each navigation subtask, forming a gradient distribution in terms of accuracy, speed, and resource requirements. For the localization subtask, when resources are sufficient, the system invokes a graph-optimized full-state SLAM algorithm, which provides high-precision pose estimation and a globally consistent map, but has a high computational cost. When resources are limited, it switches to a filter-based lightweight visual odometry algorithm, which, although less accurate in absolute terms, meets basic trajectory estimation requirements. For the path planning subtask, under critical situations, a search algorithm incorporating spatiotemporal trajectory optimization is invoked to generate an optimal path that balances safety and comfort. Under relaxed situations, a spline interpolation-based local path smoothing algorithm is used to maintain basic navigation functionality with minimal computational cost. This algorithm-level adaptability allows the system to maintain the best performance-efficiency ratio under any resource conditions.

[0031] Furthermore, the feedback optimization mechanism is a core element ensuring continuous system improvement, comprising two processes: forward execution and backward optimization. In the forward execution phase, the system generates motion control commands based on the processing results of algorithm instances, driving the robot to complete navigation tasks. These commands include motion parameters such as velocity and angular velocity. In the backward optimization phase, the system collects data on the execution effectiveness of motion commands through a performance monitoring module, including safety indicators such as path tracking error, obstacle avoidance success rate, and planning delay, while also recording resource utilization efficiency data for each subtask. Based on this operational data, the system automatically adjusts the weight parameters in the dynamic environment risk field using reinforcement learning algorithms. For example, when insufficient obstacle avoidance response is detected, the weight coefficient of the dynamic risk layer is appropriately increased; when low resource utilization is detected, the resource allocation base is automatically adjusted. This closed-loop optimization allows the system to continuously learn from actual operational experience, gradually adapting to specific working environments and task characteristics. For instance, during long-term operation, the system may discover a high frequency of dynamic obstacles in a certain area, and automatically increase the base risk value for that area, enabling the robot to prepare for responses earlier. Through continuous environmental adaptation and parameter optimization, the intelligence level and overall performance of the navigation system can be continuously improved.

[0032] In some embodiments, calculating and generating a dynamic environmental risk field characterizing the level of spatial risk based on environmental perception data includes: Based on static obstacle information in environmental perception data, calculate the Euclidean distance field from the robot to the nearest static obstacle and generate a static risk layer. Based on dynamic object information in environmental perception data, predict the short-term motion trajectory of the object in the future, calculate the spatiotemporal intersection probability of the robot and the predicted trajectory, and generate a dynamic risk layer. Based on semantic information in environmental perception data, predefined fixed risk weights are assigned to different functional areas to generate a semantic risk layer. The static risk layer, dynamic risk layer, and semantic risk layer are weighted and superimposed to generate a unified dynamic environmental risk field.

[0033] Specifically, based on static obstacle information in the environmental perception data, the Euclidean distance field from the robot to the nearest static obstacle is calculated, generating a static risk layer. This static risk layer is constructed using a distance transformation algorithm, with the risk value increasing exponentially as the robot gets closer to the obstacle, accurately reflecting the potential threat distribution of fixed obstacles to the robot in the environment. Based on dynamic object information in the environmental perception data, the short-term trajectory of the object is predicted, and the spatiotemporal intersection probability between the robot and the predicted trajectory is calculated, generating a dynamic risk layer. This dynamic risk layer uses a Kalman filter algorithm for motion prediction and quantifies the immediate threat level of dynamic obstacles through a spatiotemporal collision probability model. Based on semantic information in the environmental perception data, predefined fixed risk weights are assigned to different functional areas, generating a semantic risk layer. This semantic risk layer assigns higher weights to high-risk areas such as elevator entrances and stairwells, and lower weights to open areas, based on environmental semantic understanding. Finally, the static risk layer, dynamic risk layer, and semantic risk layer are weighted and superimposed to generate a unified dynamic environmental risk field. This multi-layered risk field construction method can comprehensively and accurately reflect the overall risk distribution of the environment, providing a reliable environmental perception basis for subsequent navigation decisions.

[0034] In some embodiments, determining the navigation situation category of the robot based on the risk value distribution of the dynamic environmental risk field includes: Set a high-risk threshold and a low-risk threshold; In a dynamic environmental risk field, when the risk value of the robot's location and nearby planned path points is consistently higher than the high-risk threshold, the robot is determined to be in a critical situation. When the overall risk value of the robot's location and surrounding area remains below the low-risk threshold, the robot is considered to be in a relaxed state.

[0035] Specifically, the high-risk threshold and low-risk threshold, derived from extensive experimental data, effectively distinguish environmental conditions with different risk levels. In a dynamic environmental risk field, when the risk value of the robot's location and nearby planned path points consistently exceeds the high-risk threshold, the robot is determined to be in a critical situation. This spatiotemporal continuity-based judgment mechanism avoids misjudgments caused by instantaneous risk fluctuations, ensuring timely upgrades to protection levels in truly dangerous environments. When the overall risk value of the robot's location and surrounding area consistently falls below the low-risk threshold, the robot is determined to be in a relaxed situation. This judgment condition, by comprehensively considering the risk distribution around the robot, ensures that resource allocation is only relaxed after environmental safety is confirmed. This dual-threshold combined with duration-based judgment method ensures that navigation situation determination is both highly sensitive and maintains good stability, providing accurate decision-making basis for subsequent resource scheduling strategies.

[0036] In some embodiments, determining the real-time criticality level of multiple navigation subtasks of the robot includes: The navigation subtask includes at least the localization subtask, the path planning subtask, and the motion control subtask. When a critical situation is determined, the real-time criticality level of the path planning subtask and the motion control subtask is set to the highest level. When the situation is deemed to be relaxed, the real-time criticality level of the positioning subtask, path planning subtask, and motion control subtask is set to low.

[0037] Specifically, the navigation subtask includes at least three core functional modules: localization, path planning, and motion control. When a critical situation is identified, the real-time criticality level of the path planning and motion control subtasks is set to the highest level. This is because in hazardous environments, quickly replanning a safe path and accurately executing obstacle avoidance control are the primary tasks to ensure robot safety, requiring priority to be given to the computational resources of these two subtasks. When a relaxed situation is identified, the real-time criticality level of the localization, path planning, and motion control subtasks is set to a low level. At this time, the environmental risk is low, and the resource allocation of each subtask can be appropriately reduced to optimize system energy efficiency while meeting basic navigation requirements. This dynamic adjustment mechanism of criticality level based on navigation situation ensures that computational resources can be intelligently allocated according to changes in environmental risk, ensuring safety in high-risk environments while improving resource utilization efficiency in safe environments.

[0038] In some embodiments, dynamically allocating corresponding computing resources includes: For subtasks with the highest real-time criticality level, allocate dedicated computing cores, priority memory access channels, and the highest processing frequency. For subtasks with a low real-time criticality level, they are assigned to a shared pool of computing cores, with maximum CPU utilization and memory bandwidth limited, and allowed to run at base frequency or in downclocked mode.

[0039] Specifically, for subtasks with the highest real-time criticality level, dedicated computing cores, priority memory access channels, and the highest processing frequency are allocated. This dedicated resource allocation mode ensures that high-priority tasks receive stable and sufficient computing power to meet real-time requirements in complex environments. For subtasks with lower real-time criticality levels, they are allocated to a shared computing core pool, with maximum CPU utilization and memory bandwidth limited, and allowed to run at base frequency or in downclocked mode. This restrictive resource allocation effectively prevents low-priority tasks from excessively consuming computing resources, while further optimizing energy efficiency through dynamic frequency adjustment. Through this differentiated resource allocation strategy, the system can achieve reasonable allocation and efficient utilization of computing resources while ensuring the performance of critical tasks, ensuring optimal system performance under different navigation situations.

[0040] In some embodiments, invoking and executing an algorithm instance that matches the computing resource level includes, based on computing resources: For the localization subtask with the highest real-time criticality level, call the full state graph optimization SLAM algorithm instance; For the localization subtask with a low real-time criticality level, call the lightweight visual odometry instance; For the path planning subtask with the highest real-time criticality level, invoke an A or D algorithm instance that includes spatiotemporal trajectory optimization. For the path planning subtask with a low real-time criticality level, call the local path smoothing algorithm instance based on uniform B-splines. For motion control subtasks with the highest real-time criticality level, invoke a high-frequency iterative model predictive control algorithm instance. For motion control subtasks with a low real-time criticality level, call a regular cycle proportional-integral-derivative control algorithm instance.

[0041] Specifically, for the localization subtask with the highest real-time criticality level, the full-state graph optimized SLAM algorithm instance is invoked. This algorithm achieves centimeter-level high-precision localization by maintaining all feature points and pose nodes of the global map and using nonlinear optimization methods. Although the computational complexity is high, it can provide reliable pose estimation under critical situations, ensuring the robot's localization accuracy in complex environments. For the localization subtask with the lowest real-time criticality level, the lightweight visual odometry method instance is invoked. This method is based only on feature matching and motion estimation between adjacent image frames, reducing the computational load by about 60%. Although the long-term cumulative error is large, it is sufficient to meet the basic localization requirements for short-distance navigation under relaxed situations, effectively saving computational resources. For path planning subtasks with the highest real-time criticality level, instances of A or D algorithms that include spatiotemporal trajectory optimization are invoked. These algorithms not only consider the safety of spatial paths but also introduce a temporal dimension for trajectory optimization, effectively avoiding dynamic obstacles and generating smooth and dynamically feasible motion trajectories, making them particularly suitable for use in dynamic and complex environments. For path planning subtasks with a low real-time criticality level, instances of local path smoothing algorithms based on uniform B-splines are invoked. This algorithm generates smooth paths that meet continuity requirements with relatively low computational cost by interpolating and optimizing the initial path, making it suitable for static or simple dynamic environments. For motion control subtasks with the highest real-time criticality level, a high-frequency iterative model predictive control algorithm instance is invoked. This algorithm performs forward prediction based on the system dynamics model and solves for the optimal control sequence through online optimization. It can effectively handle various constraints, with a control frequency of up to 1kHz, ensuring accurate trajectory tracking and rapid disturbance suppression under critical situations. For motion control subtasks with a low real-time criticality level, a conventional periodic proportional-integral-derivative (PI-DE) control algorithm instance is invoked. This algorithm operates at a fixed control frequency of 100Hz, has a simple structure, and low computational cost. Although its ability to handle complex dynamic constraints is limited, it can meet basic path tracking requirements under relaxed situations. This dynamic switching mechanism of algorithm instances based on criticality level achieves the best match between computational resources and algorithm performance, ensuring navigation safety and reliability under critical situations while significantly improving the system's energy efficiency under relaxed situations.

[0042] In some embodiments, feedback optimization of the computational parameters and computational resource allocation strategy for the dynamic environmental risk field includes: Monitor the frequency of path tracking error exceeding limits or planning delays caused by insufficient computing resource allocation under critical situations; The resource surplus is calculated by measuring the difference between the actual computing resources consumed by each low-level subtask and the allocated resource limit under a relaxed situation. Adjust the weight of the dynamic risk layer in the dynamic environmental risk field by increasing or decreasing the frequency of occurrence. Based on the available resources, the upper limit of computing resources allocated to lower-level subtasks can be reduced or increased.

[0043] Specifically, the system monitors the frequency of path tracking errors exceeding limits or planning delays caused by insufficient computing resource allocation under critical situations. This monitoring, based on actual performance, accurately reflects the effectiveness of resource allocation strategies and promptly identifies system bottlenecks. Under relaxed conditions, the system calculates the difference between the actual computing resources consumed by each low-level subtask and the allocated resource limit, yielding resource reserves. This calculation helps assess the rationality of resource allocation schemes and identify potential resource waste. The weight of the dynamic risk layer in the dynamic environment risk field is dynamically adjusted upwards or downwards based on the frequency of events, making risk assessment more consistent with actual environmental characteristics. Based on the resource reserve statistics, the upper limit of computing resources allocated to low-level subtasks is intelligently adjusted, achieving refined resource management and adaptive optimization. This feedback optimization mechanism based on operational data enables the system to continuously improve its decision parameters, gradually enhancing its adaptability and overall performance under different environments.

[0044] All of the above-mentioned optional technical solutions can be combined in any way to form the optional embodiments of this application, and will not be described in detail here.

[0045] The following are embodiments of the apparatus described in this application, which can be used to execute the embodiments of the method described in this application. For details not disclosed in the apparatus embodiments of this application, please refer to the embodiments of the method described in this application.

[0046] Figure 3 This is a schematic diagram of a flexible resource scheduling device for robot navigation provided in an embodiment of this application. Figure 3 As shown, the flexible resource scheduling device for robot navigation includes: The data acquisition module 301 is configured to acquire the robot's environmental perception data and task status data; The dynamic environmental risk field generation module 302 is configured to calculate and generate a dynamic environmental risk field that characterizes the level of spatial risk based on environmental perception data. The navigation situation category determination module 303 is configured to determine the navigation situation category of the robot based on the risk value distribution of the dynamic environment risk field. The navigation situation category includes at least critical situation and relaxed situation. The real-time criticality level determination module 304 is configured to determine the real-time criticality level of multiple navigation sub-tasks of the robot based on the navigation situation category and task status data. The computing resource allocation module 305 is configured to dynamically allocate corresponding computing resources based on the real-time criticality level. Execution module 306 is configured to invoke and execute an algorithm instance that matches the computing resource level, based on computing resources, to perform the navigation task; The optimization module 307 is configured to output motion control commands for the robot based on the execution results of the algorithm instance; at the same time, it performs feedback optimization on the calculation parameters and resource allocation strategy of the dynamic environment risk field based on the execution effect of the motion control commands and the resource usage.

[0047] In some embodiments, Figure 3 The dynamic environment risk field generation module 302 is specifically configured as follows: Based on static obstacle information in environmental perception data, calculate the Euclidean distance field from the robot to the nearest static obstacle and generate a static risk layer. Based on dynamic object information in environmental perception data, predict the short-term motion trajectory of the object in the future, calculate the spatiotemporal intersection probability of the robot and the predicted trajectory, and generate a dynamic risk layer. Based on semantic information in environmental perception data, predefined fixed risk weights are assigned to different functional areas to generate a semantic risk layer. The static risk layer, dynamic risk layer, and semantic risk layer are weighted and superimposed to generate a unified dynamic environmental risk field.

[0048] In some embodiments, Figure 3 The navigation situation category determination module 303 is specifically configured as follows: Set a high-risk threshold and a low-risk threshold; In a dynamic environmental risk field, when the risk value of the robot's location and nearby planned path points is consistently higher than the high-risk threshold, the robot is determined to be in a critical situation. When the overall risk value of the robot's location and surrounding area remains below the low-risk threshold, the robot is considered to be in a relaxed state.

[0049] In some embodiments, Figure 3 The real-time criticality level determination module 304 is specifically configured as follows: The navigation subtask includes at least the localization subtask, the path planning subtask, and the motion control subtask. When a critical situation is determined, the real-time criticality level of the path planning subtask and the motion control subtask is set to the highest level. When the situation is deemed to be relaxed, the real-time criticality level of the positioning subtask, path planning subtask, and motion control subtask is set to low.

[0050] In some embodiments, Figure 3 The computing resource allocation module 305 is specifically configured as follows: For subtasks with the highest real-time criticality level, allocate dedicated computing cores, priority memory access channels, and the highest processing frequency. For subtasks with a low real-time criticality level, they are assigned to a shared pool of computing cores, with maximum CPU utilization and memory bandwidth limited, and allowed to run at base frequency or in downclocked mode.

[0051] In some embodiments, Figure 3 Execution module 306 is specifically configured as follows: For the localization subtask with the highest real-time criticality level, call the full state graph optimization SLAM algorithm instance; For the localization subtask with a low real-time criticality level, call the lightweight visual odometry instance; For the path planning subtask with the highest real-time criticality level, invoke an A or D algorithm instance that includes spatiotemporal trajectory optimization. For the path planning subtask with a low real-time criticality level, call the local path smoothing algorithm instance based on uniform B-splines. For motion control subtasks with the highest real-time criticality level, invoke a high-frequency iterative model predictive control algorithm instance. For motion control subtasks with a low real-time criticality level, call a regular cycle proportional-integral-derivative control algorithm instance.

[0052] In some embodiments, Figure 3 The optimization module 307 is specifically configured as follows: Monitor the frequency of path tracking error exceeding limits or planning delays caused by insufficient computing resource allocation under critical situations; The resource surplus is calculated by measuring the difference between the actual computing resources consumed by each low-level subtask and the allocated resource limit under a relaxed situation. Adjust the weight of the dynamic risk layer in the dynamic environmental risk field by increasing or decreasing the frequency of occurrence. Based on the available resources, the upper limit of computing resources allocated to lower-level subtasks can be reduced or increased.

[0053] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0054] Figure 4 This is a schematic diagram of the electronic device 4 provided in an embodiment of this application. Figure 4As shown, the electronic device 4 of this embodiment includes a processor 401, a memory 402, and a computer program 403 stored in the memory 402 and executable on the processor 401. When the processor 401 executes the computer program 403, it implements the steps in the various method embodiments described above. Alternatively, when the processor 401 executes the computer program 403, it implements the functions of each module / unit in the various device embodiments described above.

[0055] Electronic device 4 can be a desktop computer, laptop, handheld computer, cloud server, or other electronic device. Electronic device 4 may include, but is not limited to, processor 401 and memory 402. Those skilled in the art will understand that... Figure 4 This is merely an example of electronic device 4 and does not constitute a limitation on electronic device 4. It may include more or fewer components than shown, or different components.

[0056] The processor 401 may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc.

[0057] The memory 402 can be an internal storage unit of the electronic device 4, such as a hard disk or RAM of the electronic device 4. The memory 402 can also be an external storage device of the electronic device 4, such as a plug-in hard disk, Smart Media Card (SMC), Secure Digital (SD) card, Flash Card, etc., equipped on the electronic device 4. The memory 402 can also include both internal and external storage units of the electronic device 4. The memory 402 is used to store computer programs and other programs and data required by the electronic device.

[0058] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is merely an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiments can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0059] If integrated modules / units are implemented as software functional units and sold or used as independent products, they can be stored in a readable storage medium (e.g., a computer-readable storage medium). Based on this understanding, all or part of the processes in the methods of the above embodiments can also be implemented by a computer program instructing related hardware. The computer program can be stored in a readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program may include computer program code, which may be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable storage medium may include: any entity or device capable of carrying computer program code, recording media, USB flash drives, portable hard drives, magnetic disks, optical disks, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc.

[0060] The above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application 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 this application, and should all be included within the protection scope of this application.

Claims

1. A flexible resource scheduling method for robot navigation, characterized in that, The method includes: Acquire the robot's environmental perception data and task status data; Based on the environmental perception data, a dynamic environmental risk field characterizing the level of spatial risk is calculated and generated; Based on the risk value distribution of the dynamic environmental risk field, the navigation situation category of the robot is determined, and the navigation situation category includes at least critical situation and relaxed situation; Based on the navigation situation category and the task status data, determine the real-time criticality level of multiple navigation sub-tasks of the robot; Based on the real-time criticality level, the corresponding computing resources are dynamically allocated; Based on the computing resources, invoke and execute algorithm instances that match the computing resource level to perform the navigation task; Based on the execution results of the algorithm instance, motion control commands for the robot are output; simultaneously, based on the execution effect of the motion control commands and resource usage, the calculation parameters of the dynamic environment risk field and the allocation strategy of the computing resources are optimized.

2. The method according to claim 1, characterized in that, The step of calculating and generating a dynamic environmental risk field characterizing the level of spatial risk based on the environmental perception data includes: Based on the static obstacle information in the environmental perception data, the Euclidean distance field from the robot to the nearest static obstacle is calculated, and a static risk layer is generated. Based on the dynamic object information in the environmental perception data, predict the object's short-term trajectory in the future, calculate the spatiotemporal intersection probability between the robot and the predicted trajectory, and generate a dynamic risk layer. Based on the semantic information in the environmental perception data, predefined fixed risk weights are assigned to different functional areas to generate a semantic risk layer. The static risk layer, the dynamic risk layer, and the semantic risk layer are weighted and superimposed to generate a unified dynamic environmental risk field.

3. The method according to claim 1, characterized in that, The process of determining the navigation situation category of the robot based on the risk value distribution of the dynamic environmental risk field includes: Set a high-risk threshold and a low-risk threshold; In the dynamic environmental risk field, when the risk value of the robot's location and nearby planned path points is continuously higher than the high-risk threshold, the robot is determined to be in the critical situation. When the overall risk value of the robot's location and surrounding area remains below the low-risk threshold, the robot is determined to be in the relaxed state.

4. The method according to claim 3, characterized in that, The real-time criticality level of the robot's multiple navigation sub-tasks is determined by: The navigation subtask includes at least a positioning subtask, a path planning subtask, and a motion control subtask. When the critical situation is determined, the real-time criticality level of the path planning subtask and the motion control subtask is set to the highest level; When the situation is deemed to be relaxed, the real-time criticality level of the positioning subtask, the path planning subtask, and the motion control subtask is set to low.

5. The method according to claim 4, characterized in that, The computing resources corresponding to the dynamic allocation include: For the subtasks with the highest real-time criticality level, allocate dedicated computing cores, priority memory access channels, and the highest processing frequency. For subtasks with a low real-time criticality level, they are allocated to a shared computing core pool, with maximum CPU utilization and memory bandwidth limited, and allowed to run at base frequency or in downclocked mode.

6. The method according to any one of claims 4 or 5, characterized in that, The step of invoking and executing an algorithm instance that matches the computing resource level based on the computing resources includes: For the localization subtask with the highest real-time criticality level, call the full-state graph optimized SLAM algorithm instance; For the localization subtask with a low real-time criticality level, invoke a lightweight visual odometry instance; For the path planning subtask with the highest real-time criticality level, invoke an A or D algorithm instance that includes spatiotemporal trajectory optimization; For the path planning subtask with a low real-time criticality level, an instance of the local path smoothing algorithm based on uniform B-splines is invoked. For the motion control subtask with the highest real-time criticality level, a high-frequency iterative model predictive control algorithm instance is invoked. For the motion control subtask with a low real-time criticality level, a regular cycle proportional-integral-derivative control algorithm instance is invoked.

7. The method according to claim 1, characterized in that, The feedback optimization of the calculation parameters of the dynamic environmental risk field and the allocation strategy of the computing resources includes: Monitor the frequency of path tracking error exceeding limits or planning delay events caused by insufficient allocation of computing resources under the aforementioned critical situation; The difference between the actual computing resources consumed by each low-level subtask and the allocated upper limit of resources under the aforementioned relaxed conditions is used to obtain the resource surplus. The weight of the dynamic risk layer in the dynamic environment risk field is adjusted up or down according to the occurrence frequency. Based on the resource surplus, the upper limit of computing resources allocated to lower-level subtasks can be reduced or increased.

8. A flexible resource scheduling device for robot navigation, characterized in that, include: The data acquisition module is used to acquire the robot's environmental perception data and task status data; The dynamic environmental risk field generation module is used to calculate and generate a dynamic environmental risk field that characterizes the level of spatial risk based on the environmental perception data. The navigation situation category determination module is used to determine the navigation situation category of the robot based on the risk value distribution of the dynamic environment risk field. The navigation situation category includes at least critical situations and relaxed situations. The real-time criticality level determination module is used to determine the real-time criticality level of multiple navigation sub-tasks of the robot based on the navigation situation category and the task status data. The computing resource allocation module is used to dynamically allocate corresponding computing resources according to the real-time criticality level; An execution module is used to invoke and execute an algorithm instance that matches the computing resource level, based on the computing resources, to perform a navigation task; The optimization module is used to output motion control commands for the robot based on the execution results of the algorithm instance; at the same time, it performs feedback optimization on the calculation parameters of the dynamic environment risk field and the allocation strategy of the computing resources based on the execution effect of the motion control commands and the resource usage.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the method as described in any one of claims 1 to 7.

10. A readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method as described in any one of claims 1 to 7.

Images