Search

CN-121978991-A - Task processing method and device and intelligent body

CN121978991ACN 121978991 ACN121978991 ACN 121978991ACN-121978991-A

Abstract

The application provides a task processing method, a task processing device and an intelligent body, which can be used in the technical field of artificial intelligence. The method comprises the steps of obtaining current environment information, carrying out task planning based on the current environment information and a target task to generate a multi-step action instruction sequence, outputting the multi-step action instruction sequence to enable an intelligent body to execute a first step action instruction in the multi-step action instruction sequence, repeatedly executing the following steps A to C until the target task is judged to be completed based on the current environment information, responding to a completion signal of the first step action instruction to obtain the current environment information again, carrying out task planning again based on the obtained current environment information and the target task to generate a new multi-step action instruction sequence, and outputting the new multi-step action instruction sequence to enable the intelligent body to execute the first step action instruction in the new multi-step action instruction sequence.

Inventors

  • Request for anonymity

Assignees

  • 北京千诀科技有限公司

Dates

Publication Date
20260505
Application Date
20251211

Claims (11)

  1. 1. A method of task processing, comprising: acquiring current environmental information, planning a task based on the current environmental information and a target task, and generating a multi-step action instruction sequence; Outputting the multi-step action instruction sequence to enable an intelligent body to execute a first step action instruction in the multi-step action instruction sequence; The following steps A to C are repeatedly executed until the target task is judged to be completed based on the current environment information: step A, responding to a completion signal of the first step action instruction, and re-acquiring current environment information; step B, re-planning the task based on the re-acquired current environment information and the target task, and generating a new multi-step action instruction sequence; and step C, outputting the new multi-step action instruction sequence so that the body-building intelligent agent executes a first step action instruction in the new multi-step action instruction sequence.
  2. 2. The method of claim 1, wherein the environmental information is collected by a perception module of the body-building agent, the multi-step sequence of motion instructions is provided to a control module of the body-building agent, and the control module controls an actuator of the body-building agent to perform the corresponding motion according to a first step of motion instructions in the multi-step sequence of motion instructions.
  3. 3. The method of claim 2, wherein the mission planning is implemented by a decision model that is built based on a large language model or a visual language model.
  4. 4. A method according to claim 3, wherein each step of the action instruction is obtained by instantiating an atomic skill of the body-building agent.
  5. 5. The method of claim 4, wherein the task planning process of the decision model is as follows: Receiving current environment information and a target task; selecting a plurality of atomic skills from an atomic skill set of the body-building agent based on the current environmental information and the target task, and determining parameters required to perform each of the atomic skills; the selected atomic skills are combined with corresponding parameters to generate a multi-step sequence of action instructions.
  6. 6. The method of claim 1, wherein the action instruction of each step is one of a natural language description action instruction, a programming structure instruction, or a JSON format instruction.
  7. 7. A task processing device, comprising: the acquisition module is used for acquiring current environmental information; The task planning module is used for carrying out task planning based on the current environment information and the target task and generating a multi-step action instruction sequence; The output module is used for outputting the multi-step action instruction sequence so that an intelligent body can execute a first step action instruction in the multi-step action instruction sequence; the acquisition module, the task planning module and the output module are further respectively configured to repeatedly execute the following steps a to C until the target task is judged to be completed based on the current environmental information: step A, responding to a completion signal of the first step action instruction, and re-acquiring current environment information; step B, re-planning the task based on the re-acquired current environment information and the target task, and generating a new multi-step action instruction sequence; and step C, outputting the new multi-step action instruction sequence so that the body-building intelligent agent executes a first step action instruction in the new multi-step action instruction sequence.
  8. 8. A body-building agent, comprising: A processor configured to implement the method of any one of the preceding claims 1 to 6; The sensing module is connected with the processor and is configured to acquire environment information; The control module is connected with the processor and is configured to control the executor of the intelligent body to execute corresponding actions according to a first step action instruction in the multi-step action instruction sequence; And the actuator is connected with the control module and is configured to execute corresponding actions under the driving of the control module.
  9. 9. An electronic device comprising a processor and a memory, the memory having stored therein a computer program, the processor implementing the method of any of the preceding claims 1 to 6 when executing the program.
  10. 10. A computer readable storage medium, on which a computer program is stored, characterized in that the program, when being executed by a processor, implements the method of any of the preceding claims 1 to 6.
  11. 11. A computer program product comprising a computer program, characterized in that the computer program, when executed by a processor, implements the method of any of the preceding claims 1 to 6.

Description

Task processing method and device and intelligent body Technical Field The application relates to the technical field of artificial intelligence, in particular to a task processing method and device and an intelligent body. Background With the rapid development of self-contained intelligent technology, robots have received increasing attention for their autonomous decision making capability in multi-step, complex task scenarios. The task planning with body, namely converting natural language instructions into a series of executable actions to complete tasks in a complex dynamic environment, is a core technical problem for constructing intelligent robots. The existing robot control method generally adopts a static task planning mode, namely after the environment awareness is acquired, a complete action sequence is output by a model at one time and is gradually executed according to the sequence. The method models the task planning problem as a static predictive task, and the environment is not changed or dynamic adjustment is not needed in task execution. However, the actual environment often has high dynamic property and uncertainty, so that the static task planning cannot adjust the execution strategy in time when facing environmental change or perception error, and poor robustness and adaptability are shown. Disclosure of Invention Aiming at the problems in the prior art, the embodiment of the application provides a task processing method, a task processing device and an intelligent body, which can at least partially solve the problems in the prior art. According to a first aspect of the application, a task processing method is provided, which comprises the steps of obtaining current environment information, carrying out task planning based on the current environment information and a target task to generate a multi-step action instruction sequence, outputting the multi-step action instruction sequence to enable an intelligent body to execute a first step action instruction in the multi-step action instruction sequence, repeatedly executing the following steps A to C until the target task is judged to be completed based on the current environment information, re-obtaining the current environment information in response to receiving a completion signal of the first step action instruction, re-carrying out task planning based on the re-obtained current environment information and the target task to generate a new multi-step action instruction sequence, and outputting the new multi-step action instruction sequence to enable the intelligent body to execute the first step action instruction in the new multi-step action instruction sequence. In some embodiments, the environmental information is collected by a perception module of the body-building agent, the sequence of multi-step action instructions is provided to a control module of the body-building agent, and the control module controls an actuator of the body-building agent to perform a corresponding action according to a first step action instruction in the sequence of multi-step action instructions. In some embodiments, the mission plan is implemented by a decision model that is built based on a large language model or a visual language model. In some embodiments, each step of the action instructions is obtained by instantiating an atomic skill of the body-building agent. In some embodiments, the task planning process of the decision model includes receiving current environmental information and a target task, selecting a plurality of atomic skills from an atomic skill set of the body-building agent based on the current environmental information and the target task, determining parameters required for executing each of the atomic skills, and combining the selected atomic skills with the corresponding parameters to generate a multi-step action instruction sequence. In some embodiments, the action instruction of each step is one of a natural language description action instruction, a programming structure instruction, or a JSON format instruction. According to a second aspect of the application, a task processing device is provided, which comprises an acquisition module, a task planning module, an output module and a step C, wherein the acquisition module is used for acquiring current environment information, the task planning module is used for carrying out task planning based on the current environment information and a target task to generate a multi-step action instruction sequence, the output module is used for outputting the multi-step action instruction sequence to enable an intelligent body to execute a first step action instruction in the multi-step action instruction sequence, the acquisition module, the task planning module and the output module are respectively used for repeatedly executing the following steps A to C until the target task is judged to be completed based on the current environment information, the current environment information