Efficient Incremental Adaptation Method for Decision-making Model

The enhancement of decision-making models is of great significance in augmenting the applicability of unmanned systems in complex scenarios. This paper proposes an incremental adaptation method for local trajectory planning decision-making model. The method establishes a Decision Skills Association Model (DSAM), which associates previously learned decision skills with the current scenario to output suitable objective function parameters. Our method actively generates scenarios for further training according to the DSAM’s condition, efficiently expanding the DSAM. A key information extraction method is also employed to enhance the efficiency of decision skills learning. The experimental results show that our method helps unmanned system to achieve generally favorable performance in various scenarios. Moreover, the active exploration strategy and key information extraction method enhance the learning efficiency of both the DSAM and the decision skills learning model.