An Autonomous Decision-Making Method for Unmanned Vehicle Based on the Collaboration of Large Language Model and Dedicated Models

In response to the demand for enhancing the autonomous decision-making capabilities of unmanned vehicle in future military scenarios, this paper addresses the challenges in the traditional decision-making process of unmanned vehicle, such as excessive intervention by commanders, insufficient correctness and poor generalization ability of decision-making schemes. An autonomous decision-making method for unmanned vehicle based on the collaboration of large language model and dedicated models is proposed. By constructing the autonomous decision-making large language model for unmanned vehicle oriented towards typical tasks and establishing a collaborative mechanism between the large language model and dedicated models, the method takes the natural language instructions of commanders as input and uses the autonomous decision-making large language model as the control core to invoke dedicated models such as path planning, reconnaissance planning, and action parameter planning. This enables the unmanned vehicle to accurately understand, reasonably plan, and make intelligent decisions on typical tasks. Finally, the effectiveness of the method is tested and verified in a test environment.