Cooperative Decision-Making in Shared Spaces: Making Urban Traffic Safer through Human-Machine Cooperation

Balint Varga,Dongxu Yang,Sören Hohmann

DOI: https://doi.org/10.1109/SISY60376.2023.10417908

2023-06-26

Abstract:In this paper, a cooperative decision-making is presented, which is suitable for intention-aware automated vehicle functions. With an increasing number of highly automated and autonomous vehicles on public roads, trust is a very important issue regarding their acceptance in our society. The most challenging scenarios arise at low driving speeds of these highly automated and autonomous vehicles, where interactions with vulnerable road users likely occur. Such interactions must be addressed by the automation of the vehicle. The novelties of this paper are the adaptation of a general cooperative and shared control framework to this novel use case and the application of an explicit prediction model of the pedestrian. An extensive comparison with state-of-the-art algorithms is provided in a simplified test environment. The results show the superiority of the proposed model-based algorithm compared to state-of-the-art solutions and its suitability for real-world applications due to its real-time capability.

Systems and Control

What problem does this paper attempt to address?

The problem that this paper attempts to solve is to achieve cooperative decision - making in shared spaces in order to improve the safety of urban traffic, especially by human - machine cooperation to handle the interactions between highly automated and self - driving vehicles and pedestrians. Specifically, the paper focuses on the possible interactions between these highly automated or self - driving vehicles and vulnerable road users (such as cyclists and pedestrians) when driving at low speeds. Such interactions need to be properly handled by the vehicle's automated functions to enhance the society's trust in these systems.

The main innovation points of the paper are:
1. Adapting a general cooperation and shared - control framework to this new application scenario, that is, handling the interactions between self - driving vehicles and pedestrians in mixed - traffic scenarios.
2. Applying an explicit pedestrian prediction model, which can explicitly handle the prediction of pedestrians' future system states and their interactions with self - driving vehicles.

To verify the effectiveness of the proposed model, the authors carried out extensive comparative experiments in a simplified test environment. The results show that the proposed model - based algorithm is superior to the existing solutions and is suitable for practical applications due to its real - time capabilities.