基于模糊推理和有限状态机的行为决策

Behavioral Decision Based on Fuzzy Inference and Finite State Machine

本文提出了一种应用于单向双车道的规则型行为决策算法,通过将模糊推理与有限状态机相结合,提高了算法的可行性和决策结果的准确性。基于有限状态机的决策方法具备场景遍历广度优势,但对某一特定场景缺乏遍历的深度,其状态转移条件也较为简单。使用层次状态机与模糊推理相结合的决策算法,其不仅能提升场景遍历深度,还能根据速度和距离进行自适应调节,进一步提升换道决策的准确性与合理性。先在Matlab自动驾驶工具箱中建立道路场景,然后提出并构建基于模糊推理和有限状态机的行为决策模型,采用B样条曲线进行轨迹规划、模型预测控制进行路径跟踪,最后通过Simulink仿真验证了决策系统的可行性。

This paper proposes a rule-based behavioral decision algorithm for one-way two lane traffic.By combining fuzzy reasoning with Finite-state machine,the feasibility of the algorithm and the accuracy of the decision results are improved.The decision method based on Finite-state machine has the advantage of scene traversal breadth,but it lacks the depth of traversal for a specific scene,and its state transition conditions are relatively simple.The decision algorithm combining hierarchical state machines and fuzzy reasoning can not only improve the depth of scene traversal,but also adaptively adjust according to speed and distance,further improving the accuracy and rationality of lane changing decisions.First,the road scene is established in the Matlab autopilot toolbox,then a behavioral decision model based on fuzzy reasoning and Finite-state machine is proposed and constructed.B-spline curve is used for trajectory planning,and model predictive control is used for path tracking.Finally,the feasibility of the decision system is verified through Simulink simulation.