Dynamic Process Control System Based on Artificial Intelligence Agent

Abstract

A dynamic process control system based on artificial intelligence agents includes a communication interface, memory, and a processor. The processor is electrically connected to the communication interface and memory, and performs semantic analysis of prompt data through an artificial intelligence reasoning model to dynamically determine the tool calling order and execution strategy. The system continuously updates the data status it holds through a state feedback loop until the artificial intelligence reasoning model autonomously determines that the requirements of the target task have been met based on the updated state data, and then outputs a completion instruction to end the process. This achieves the technical effect of improving the adaptability, verifiability, and robustness of dynamic process control.

Claims

1. A dynamic process control system based on artificial intelligence agents, the system comprising: a communication interface; a memory for storing tool catalog data and status data; and a processor electrically connected to the communication interface and the memory, the processor being configured to perform the following steps: (A) receiving input data, the input data including information describing a target task; (B) combining the input data, the status data, and the tool catalog data into prompt data; (C) providing the prompt data to an artificial intelligence reasoning model, and receiving decision data generated by the artificial intelligence reasoning model after semantically parsing the prompt data; (D) When the decision data is determined to be a tool call instruction, a target tool identification code and a target parameter set are parsed from the tool call instruction. Then, a corresponding tool is located from the tool catalog data based on the target tool identification code. The target parameter set is transmitted to the corresponding tool through the communication interface to execute the corresponding tool. An execution result is received from the corresponding tool through the communication interface, and the execution result is written into the status data to update the status data to an updated status data. (E) Return to execute steps (B) to (D) until the decision data generated in step (C) is a completion indication. The completion indication is output by the artificial intelligence reasoning model based on the updated status data, which determines that the requirements of the target task have been met. (F) When the decision data is the completion indication, a final execution result is output.

2. The dynamic process control system based on artificial intelligence agents as described in Request 1, wherein the processor is further configured to perform the following steps between step (D) and step (E): providing the execution result to the artificial intelligence inference model, and receiving an evaluation result output by the artificial intelligence inference model after evaluating the validity of the execution result; when the evaluation result indicates that the execution result does not meet the requirements of the target task, merging the evaluation result into the updated status data so that the prompt data combined in the subsequent step (B) includes the evaluation result.

3. The dynamic process control system based on artificial intelligence agents as described in request item 1, wherein the artificial intelligence reasoning model outputs the completion indication based on a judgment criterion, the judgment criterion including that at least one predetermined field in the updated status data corresponding to the target task has been filled with data conforming to a predetermined format.

4. As in Request 1, a dynamic process control system based on an artificial intelligence agent, wherein the tool catalog data includes a sub-agent calling tool, and the processor is further configured to provide the target parameter set to another artificial intelligence inference model when the target tool identification code corresponds to the sub-agent calling tool, and to receive a sub-execution result output by the other artificial intelligence inference model based on the target parameter set, wherein the execution result includes the sub-execution result.

5. The dynamic process control system based on artificial intelligence agent as claimed in claim 1, wherein the processor is further configured to perform the following steps after step (A): providing the input data to the artificial intelligence inference model and receiving a preliminary task plan output by the artificial intelligence inference model, the preliminary task plan including multiple steps estimated to be required to perform the target task, and the prompt data combined in step (B) further including the preliminary task plan.

Images