Intelligent robot control algorithm scheduling method and system based on dynamic directed acyclic graph

Abstract

The invention discloses an intelligent robot control algorithm scheduling method and system based on a dynamic directed acyclic graph. The scheduling method comprises the following steps: receiving changes of a robot state or environmental data, traversing a conditioned directed acyclic graph from a root node, and dynamically generating an intelligent robot control algorithm scheduling instance when execution is performed to an end node; wherein the conditional directed acyclic graph comprises at least one conditional node, and is used for activating the corresponding model branch according to the input data. When the scheduling method is oriented to different scenes, the deployment of a robot control algorithm can be effectively accelerated, and the complexity of deployment and implementation is reduced.

Description

Technical Field

[0001] The present invention relates to the field of robot control technology, and more particularly to an intelligent robot control algorithm scheduling method and system based on a dynamic directed acyclic graph. Background Art

[0002] With the development of embodied intelligence technology, robots are making the leap from "digital intelligence" to "physical intelligence." In this process, reinforcement learning strategy models empower robots with motion control capabilities, while multimodal visual detection models enable them to perceive their environment.

[0003] However, in engineering practice, due to limitations in edge computing power and detection accuracy, a single model cannot cover all motion control or visual inspection scenarios. Currently, motion control strategy models typically employ a state machine approach, switching between different robot states. When the robot switches to a different state, the corresponding strategy model is invoked to control the robot's motion behavior. Visual inspection often employs custom development.

[0004] However, when controlling the robot's motion behavior based on a state machine, only the state changes triggered by the robot's own motors and sensor data such as the IMU are considered. When switching the algorithm, the switching is still effective. Incorporating visual inspection needs requires considering more complex scenarios, which will lead to the following problems:

[0005] State explosion: Complex scenarios require the definition of numerous state combinations (e.g., "walking + face detected + obstacle present + low battery"), leading to bloated code.

[0006] Poor flexibility: When adding new sensor dimensions, the state transition logic needs to be restructured;

[0007] Weak parallel capabilities: It is difficult to execute multiple tasks concurrently in the same state (for example, running face recognition and obstacle avoidance algorithms simultaneously while walking requires custom development).

[0008] Therefore, how to solve the above problems is an urgent problem that those skilled in the art need to solve. Summary of the Invention

[0009] In view of this, and in response to the shortcomings of traditional robots in controlling motion behavior based on state machines, the present invention provides an intelligent robot control algorithm scheduling method and system based on a dynamic directed acyclic graph.

[0010] In order to achieve the above object, the present invention adopts the following technical solutions:

[0011] First, the present invention provides an intelligent robot control algorithm scheduling method based on a dynamic directed acyclic graph, the steps comprising:

[0012] Receive changes in robot status or environmental data, traverse the conditional directed acyclic graph starting from the root node, and dynamically generate an intelligent robot control algorithm scheduling instance when executing to the end node;

[0013] The conditional directed acyclic graph includes at least one conditional node, which is used to activate a corresponding model branch according to input data.

[0014] Optionally, the model branch includes a policy model and a visual model.

[0015] Optionally, sub-condition nodes are connected to the sub-graph branches of the policy model and the visual model, and semantics are executed in parallel between the sub-condition nodes.

[0016] Optionally, the execution semantics include at least one of sequential execution, parallel execution and nested execution; wherein the nested execution is only applicable to visual model branches.

[0017] Second, the present invention provides an intelligent robot control algorithm scheduling system based on a dynamic directed acyclic graph. The system has a built-in scheduling configuration file, and the scheduling configuration file is used to implement any of the intelligent robot control algorithm scheduling methods based on a dynamic directed acyclic graph as described above.

[0018] Preferably, the system further comprises:

[0019] A configuration file editing unit, used to edit and modify the scheduling configuration file as needed;

[0020] The configuration file loading unit is used to interpret the edited scheduling configuration file and load it into the memory;

[0021] The configuration file management unit is used to trigger the scheduling configuration file in the memory and generate a scheduling instance when receiving changes in the robot status or environment data.

[0022] Preferably, the system further comprises:

[0023] An instance parsing unit, configured to expand the scheduling instance according to a topological structure and generate a scheduling instruction according to node semantics;

[0024] The execution engine unit is used to receive the scheduling instruction and realize unified scheduling execution.

[0025] Through the above technical solutions, it can be seen that the present invention discloses a method and system for scheduling an intelligent robot control algorithm based on a dynamic directed acyclic graph. Compared with the existing technology, under the same hardware conditions, this application can speed up the deployment and implementation of robot models, improve implementation efficiency, and save implementation costs when facing different scenarios; at the same time, it can reduce the deployment and implementation complexity of robot models, and promote the application and implementation efficiency of robots in actual scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] 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 merely embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying any creative work.

[0027] Figure 1 This is a flow chart of the intelligent robot control algorithm scheduling method based on a dynamic directed acyclic graph of the present invention;

[0028] Figure 2 This is a structural diagram of the intelligent robot control algorithm scheduling system based on a dynamic directed acyclic graph of the present invention. DETAILED DESCRIPTION

[0029] The following will clearly and completely describe 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 the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0030] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0031] The biggest challenge for intelligent robots compared to traditional robots is how to autonomously adapt to dynamically changing scenarios, overcoming the limitations of traditional robots, which are limited to handling deterministic scenarios using rigid rules. A changing scenario means that the robot's state and the data it acquires about its surroundings also change. It is precisely these changes in data that drive changes in the scenario.

[0032] Traditionally, directed acyclic graphs (DAGs) are often used to describe the dependencies and execution order of tasks in deterministic scenarios. Therefore, if traditional DAGs can be expanded to include conditional nodes, then conditional DAGs can be used to generalize the set of tasks that intelligent robots should perform in different scenarios, as well as the dependencies between tasks.

[0033] Based on this, the present invention expands the classic directed acyclic graph (DAG), proposes the idea of ​​a dynamic finite acyclic graph, and constructs a method and system for scheduling algorithms such as motion control algorithms and visual detection algorithms of intelligent robots.

[0034] Example 1:

[0035] The present invention provides an intelligent robot control algorithm scheduling method based on a dynamic directed acyclic graph. By expanding the traditional DAG and introducing conditional nodes, a conditional DAG model scheduling configuration file is formed, which replaces the previous state machine and realizes the need to schedule the execution of various types of model algorithms under different conditions into specific execution nodes activated by conditional nodes. At the same time, according to the actual situation of intelligent robot algorithm scheduling, three different execution semantics are formed: concurrent execution, sequential execution, and nested execution.

[0036] Specifically, the scheduling steps include:

[0037] Receive changes in robot status or environmental data, traverse the conditional directed acyclic graph starting from the root node root_node, and dynamically generate an intelligent robot control algorithm scheduling instance when executing to the end node end_node;

[0038] The conditional directed acyclic graph in the present invention includes at least one conditional node, which is used to activate the corresponding model branch according to input data.

[0039] In one embodiment, Figure 1 , Figure 1 This is a flowchart of the scheduling method for this application;

[0040] root_node: represents the root node of the entire conditional DAG. Whenever the robot state or the data obtained from the environment changes, the data is input into the root_node, and the entire conditional DAG is traversed from the root_node to dynamically create a DAG-based model scheduling run instance.

[0041] type_condition_node: represents the conditional node of the model type. In the present invention, different types of nodes are classified, such as the policy model based on reinforcement learning, the visual model, etc. Currently, in the model scheduling configuration file of the conditional DAG, this embodiment supports two types of policy models (RL_graph_node) and visual models (VI_graph_node) based on reinforcement learning by default. When the model to be called is a policy model, it will enter the branch of RL_graph_node according to the judgment of the condition; similarly, when the model to be called is a visual model, it will enter the branch of VI_graph_node according to the judgment of the condition.

[0042] RL_graph_node: represents the subgraph branch of the policy model. All conditional nodes under it are conditions for the policy model.

[0043] VI_graph_node: represents the subgraph branch of the visual model. All conditional nodes under it are conditions for the visual model.

[0044] It should be noted that this application supports the configuration and scheduling of two models, the policy model and the visual model, by default. When other types of algorithms also need to be scheduled, you only need to define a new graph_node.

[0045] In this embodiment, in the execution of a specific model, there are two semantics: one is that multiple models are executed in parallel (parallel), and the other is that multiple models are executed in sequence (sequence).

[0046] For visual models, there's a third semantics: nested execution. This semantics targets the resulting subgraph obtained after the previous parent node is recognized. Subsequent nodes then re-recognize and inspect this resulting subgraph. This approach addresses the challenge of multiple recognition in visual inspection. To address these three semantics, the model scheduling configuration file of this invention defines these three distinct semantics.

[0047] In the defined semantics, condition_node: represents the condition under which a model node can be activated. Through condition_node, different state combinations in the state machine can be turned into independent discrete conditions, and the corresponding model can be activated by activating the condition node.

[0048] execution_node: represents a model node with sequential execution semantics. When this node is reached, the model corresponding to this node will be scheduled for execution. All subsequent nodes between each execution_node and this node also form sequential execution semantics.

[0049] nested_execution_node: represents the semantics of a nested execution. Once activated, the execution_node or nested_execution_node under the subsequent condition node connected to the nested_execution_node can use the subgraph identified by the nested_execution_node as input. The nested_execution_node can only exist in the DAG of the VI_graph_node.

[0050] In this application, multiple condition_nodes can be connected from RL_graph_node and VI_graph_node. In this case, the semantics of each condition_node are parallel execution, and there is no order between them. When an execution_node or nested_execution_node completes execution, multiple condition_nodes can be connected to it. At this time, the semantics of these condition_nodes are concurrent execution.

[0051] When all nodes have executed to end_node, the dynamically created DAG instance is completed.

[0052] In this example, all conditions and models are defined in the configuration file, forming a complete conditional DAG covering different scenarios. Each time the robot's state or data acquired from the environment changes, the conditional DAG is traversed from the root_node and dynamically calculated based on the actual data of each device. This creates the DAG instance to be executed, and all models in this DAG instance are scheduled for execution, completing the complete model scheduling.

[0053] Example 2:

[0054] This embodiment provides an intelligent robot control algorithm scheduling system based on a dynamic directed acyclic graph. The system has a built-in scheduling configuration file, and the scheduling configuration file is used to implement any of the intelligent robot control algorithm scheduling methods based on a dynamic directed acyclic graph as described above.

[0055] In a preferred embodiment, the system structure is as follows Figure 2 As shown;

[0056] like Figure 2 , the system further comprises:

[0057] A configuration file editing unit is used to edit and modify the scheduling configuration file as needed. In one embodiment, initially or in subsequent robot applications, when the execution conditions of a model in a certain scenario change or a new model needs to be added, the user can use the editing unit to edit or modify the model scheduling configuration file based on the conditional DAG. The present invention supports both online and offline editing.

[0058] The configuration file loading unit is used to interpret and load the edited and modified scheduling configuration file into memory. Specifically, when the system starts or the conditional DAG is edited or modified online, the unit automatically detects file changes, automatically interprets the file, loads it into memory, and forms a static structure of the model scheduling based on the conditional DAG and stores it in memory.

[0059] The configuration file management unit is used to trigger the scheduling configuration file in memory and generate a scheduling instance when it receives changes in the robot's state or environmental data. During the robot's operation, as the scene and operating conditions change, the robot's state and sensor data will also change. When these changes are transmitted to this unit, they trigger the calculation of the variables and conditions of the entire DAG in the static structure of the conditional DAG-based model scheduling in memory, starting from the root node. Branches that do not meet the conditions are automatically discarded, and only branches with True conditions are retained, thus forming the DAG that needs to be actually executed, that is, the DAG-based model scheduling operation instance.

[0060] To further optimize the above technical solution, the system also includes:

[0061] The instance parsing unit is responsible for expanding the scheduling instance according to the topological structure and generating scheduling instructions based on node semantics. After obtaining the running instance, this unit will be responsible for expanding the entire instance graph from the rootnode of the DAG instance from top to bottom and sending the corresponding model call commands to the unified model execution engine. For concurrent execution, multiple commands will be sent to the unified model execution engine simultaneously. For sequential execution, the model execution command for the next node will be sent after the previous model has completed execution. For model calls in parent-child relationships, the detection results of the parent node model will be automatically sent to the child model when the child model is called for execution, thus realizing the execution semantics of parent and child nodes.

[0062] An execution engine unit is configured to receive the scheduling instructions and implement unified scheduling execution. Through this unified execution engine unit, the present invention schedules and executes the policy model and the visual model according to the model specified in the model call command, and feeds the results back to the DAG-based model scheduling execution engine after execution.

[0063] The above-mentioned embodiments disclosed in this application can, when facing different scenarios, speed up the deployment and implementation of robot models, improve implementation efficiency, and save implementation costs under the same hardware conditions; at the same time, reduce the deployment and implementation complexity of robot models, and promote the application and implementation efficiency of robots in actual scenarios.

[0064] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. Reference can be made to the common and similar parts between the various embodiments. For the devices disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple, and the relevant parts can be referred to the method description.

[0065] The above description of the disclosed embodiments is intended to enable one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be readily apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention is not limited to the embodiments shown herein but is intended to conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for scheduling intelligent robot control algorithms based on dynamic directed acyclic graphs, characterized in that: include: Receive changes in robot status or environmental data, traverse the conditional directed acyclic graph starting from the root node, and dynamically generate an intelligent robot control algorithm scheduling instance when executing to the end node; The conditional directed acyclic graph includes at least one conditional node, which is used to activate a corresponding model branch according to input data.

2. The intelligent robot control algorithm scheduling method according to claim 1, characterized in that: The model branch includes a policy model and a visual model.

3. The intelligent robot control algorithm scheduling method according to claim 2, characterized in that: Sub-condition nodes are connected under the sub-graph branches of the strategy model and the visual model, and semantics are executed in parallel between the sub-condition nodes.

4. The intelligent robot control algorithm scheduling method according to claim 3, characterized in that: The execution semantics include at least one of sequential execution, parallel execution and nested execution; wherein, The nested execution described above only applies to the vision model branch.

5. An intelligent robot control algorithm scheduling system based on dynamic directed acyclic graph, characterized in that: A built-in scheduling configuration file is used to implement the intelligent robot control algorithm scheduling method based on a dynamic directed acyclic graph as described in any one of claims 1-4.

6. The intelligent robot control algorithm scheduling system according to claim 5, characterized in that: Also includes: A configuration file editing unit, used to edit and modify the scheduling configuration file as needed; The configuration file loading unit is used to interpret the edited scheduling configuration file and load it into the memory; The configuration file management unit is used to trigger the scheduling configuration file in the memory and generate a scheduling instance when receiving changes in the robot status or environment data.

7. The intelligent robot control algorithm scheduling system according to claim 6, characterized in that: Also includes: An instance parsing unit, configured to expand the scheduling instance according to a topological structure and generate a scheduling instruction according to node semantics; The execution engine unit is used to receive the scheduling instruction and realize unified scheduling execution.

Images