Context-Aware Information Agents for the Automotive Domain Using Bayesian Networks

To reduce the workload of the driver due to the increasing amount of information and functions, intelligent agents represent a promising possibility to filter the immense data sets. The intentions of the driver can be analyzed and tasks can be accomplished autonomously, i.e. without interference of the user. In this contribution, different adaptive agents for the vehicle are realized: For example, the fuel agent determines its decisions by Bayesian Networks and rule-based interpretation of context influences and knowledge. The measured variables which affect the driver, the system, and the environment are analyzed. In the context of a user study the relevance of individual measured variables was evaluated. On this data basis, the agents were developed and the corresponding networks were trained. During the evaluation of the effectiveness of the agents it shows that the implemented system reduces the number of necessary interaction steps and can relieve the driver. The evaluation shows that the intentions are interpreted to a high degree correctly.