Online learning for an individualized lane-change situation recognition system applied to driving assistance

Situation recognition is a significant part of supervision to advance human operator decision making. It is a process for identification of occurred situations as the result of a sequence of actions. Situation recognition process could be individualized for an assistance system by considering exclusive behaviors of human operators individually. Accordingly, the assistance system should be provided with an online learning process to explore new experiences by modeling and labeling the occurred situations and adapt the knowledge base. In this paper, an improved Case-Based Reasoning (CBR) approach is proposed and applied for lane-change driving situation recognition. The proposed CBR is able to model event-discrete situations using Situation-Operator Modeling (SOM) approach. In addition, human operator experiences are learned online and reused for situation recognition by integration of fuzzy logic. Additional processes need to be carried out in the proposed fuzzy-SOM based CBR to support online learning for data reduction and knowledge indexing. As an experiment, the proposed approach is implemented to recognize lane-change situations for a driving assistance system. According to fundamental evaluation results, the proposed approach is able to improve lane-change situations recognition performance for individual human operators.