Ethical Decision-Making Assistance for Autonomous Driving by Hybrid Approach in Collision Imminence with Vulnerable Road Users

In this paper, we propose a hybrid decision aid system for moral decision-making during autonomous driving. Ethical decision-making for an imminent collision situation is a critical topic. First, we focused on the limitations and dilemma conditions of existing methodologies are summarized. We implemented an effective judgment method by using all of the utilitarian risk score calculations, the application of deontological rules, and the learning-based decision tree algorithm. In order to verify the proposed method, performance change was confirmed through various scenarios in multi-pedestrian crosswalks, dimensions of the verification model, and whether laws were applied. We confirmed a high accuracy of our decision-making model, about 93% without overfitting.