Here, we propose a model for driver-intention recognition that refrains from driver-based input and instead explores the utilization of information about the traffic situation to extend the predictive capabilities of the model and enable the use in highly automated or autonomous driving. The model is explored in three different scenarios: real-world motorway, simulated rural road, and simulated roundabout scenarios.

Chapter Contents:

• 21.1 Introduction
• 21.2 Literature review
• 21.3 A conceptional model for driver-intention recognition
• 21.3.1 Preliminaries
• 21.3.2 Model formulation
• 21.3.3 Learning procedure
• 21.4 Scenarios
• 21.4.1 Motorway scenarios
• 21.4.1.1 Data collection and preparation
• 21.4.1.2 Results and discussion
• 21.4.2 Rural road scenarios
• 21.4.2.1 Data collection and preparation
• 21.4.2.2 Results and discussion
• 21.4.3 Roundabout scenarios
• 21.4.3.1 Data collection and preparation
• 21.4.3.2 Results and discussion
• 21.5 Online learning
• 21.5.1 Approach
• 21.5.1.1 Distribution update methods
• 21.5.1.2 Online sample generation
• 21.6 Conclusion and future work
• Acknowledgements
• References

Inspec keywords: road traffic control; network theory (graphs); Bayes methods

Other keywords: real world motorway; driver-based input; traffic situation; dynamic Bayesian networks; simulated round about scenarios; simulated rural road; driver-intention recognition

Subjects: Traffic engineering computing; Combinatorial mathematics; Other topics in statistics