Automated Driving has the potential to profoundly reduce traffic fatalities. However, challenges remain when extending the Operational Design Domain of automated vehicles to urban scenarios, especially in mixed traffic where automated systems interact with human drivers. This thesis investigates behavior planning strategies in the scenario class of Cooperative Ordering Problems where two vehicles have to cooperatively determine their order of passing through a shared road section. This requires close interaction between traffic participants. Since both the intention of the other agent and its exact behavior are unknown, novel approaches are needed to plan safe, efficient and courteous behavior. The aim of this thesis is to identify the required components and build an Automated Driving System which can resolve these Cooperative Ordering Problems safely and efficiently. To that end, it presents three systems that incrementally solve more complex tasks within such problems. The quality of the proposed system is analyzed utilizing prototype vehicles, to demonstrate the behavior in complex traffic scenarios and large-scale empirical experiments in simulation to provide statistical evidence. A first reactive behavior planning system is presented which provides safe solutions to Cooperative Ordering Problems. Based on work in risk theory and uncertainty modeling, a behavior planner is designed which demonstrates intelligent behavior planning in scenarios, where the intention of the other agent is clear. The system is validated on prototype vehicles in various scenarios such as multi-lane traffic, intersections with priority and automated overtaking in urban traffic. In all scenarios, it shows safe and foresighted behavior, albeit in some scenarios, the behavior is overly conservative. To overcome the limitations of the first system, a second system is introduced which extends the previous approach by detecting the intention of the interaction partner regarding the order of passing through the shared traffic space. By observing the asymmetry in the state of both agents, the system is capable of detecting mismatching intentions and resolving them. Large-scale empirical simulation experiments show a significant improvement of resolution time compared to the previous system, drastically increasing the efficiency without compromising safety. However, the system cannot purposefully deviate from matching the other agent's intention, a necessary ability for negotiation. The third system extends the concepts of the previous systems to provide cooperative behavior planning. It allows courteous behavior in situations where a small disadvantage of the automated vehicle leads to a larger benefit for other traffic participants. This system is also capable of negotiation in Cooperative Ordering Problems by influencing the behavior of the other vehicle towards preferred solutions. The negotiation model modulates the decision of when to match the intention of the other agent and when to insist on achieving a better outcome. The system is tested extensively both in prototype vehicles and in simulative experiments. The system with negotiation can robustly resolve Cooperative Ordering Problems in a safe, efficient and courteous manner. Overall, this thesis provides an overview on the complexity and diverse implications of Cooperative Ordering Problems. It proposes methods to solve individual requirements and how these methods can be combined. The presented systems integrate with existing system components and can therefore extend the cooperative capabilities of existing behavior planning systems. This enables various mobility concepts, such as automated vehicles, automated micro mobility solutions, and mobile robots to interact safely with human drivers and expand their autonomy.
Cooperative Behavior Planning for Automated Vehicles in Ordering Situations
2025
Sonstige
Elektronische Ressource
Unbekannt
Recognition of dangerous situations within a cooperative group of vehicles
BASE | 2009
|Towards Cooperative Motion Planning for Automated Vehicles in Mixed Traffic
ArXiv | 2017
|