KAAPLAX is the result of a student research project at Chalmers University of Technology, Gothenburg.
The goal of this project work is to develop a solution for an assistance system for an autonomous vehicle. The unique and novel challenge of this active safety system is to predict bicyclist’s behavior by using existing sensors and technology. Implementing an algorithm in a virtual environment which prevents crashes in a critical defined situation is the main objective of this project.
The critical defined situation is a non-signalized T-intersection with a bicycle lane crossing the path of the side road. Participants in the chosen scenario are the autonomous vehicle, which is fully equipped with newest sensors and the developed threat assessment algorithm. A bicycle driving on the bicycle lane will enter the intersection from the same direction as the autonomous vehicle. A non-autonomous car coming towards with a given trajectory from naturalistic data studies will enter the intersection from the opposite direction.
The autonomous vehicle will turn left into the intersection with a real trajectory from naturalistic data studies. Before the turning maneuver the autonomous car will switch on the indicators. The bicycle will recognize the indicators and will start braking after a literature based reaction time. Based on experimental datasets from traditional and E-bicycles, the autonomous car takes action to avoid a crash with the bicycle or the oncoming traffic. A decision making algorithm is developed to show possible actions of an autonomous vehicle in a real world intersection to avoid a crash on different stopping points.
The main objective for the project was to ensure that the autonomous emergency brakes (AEB) on autonomous cars will work in order to prevent the crash between a bicyclist and a car. These kinds of projects are necessary with the focus rapidly shifting towards building autonomous cars. The results of the KAAPLAX algorithm have been very positive and promising. The threat assessment and decision making algorithm worked flawlessly in every scenario as was defined by the team for the project. A total of 65 simulations each were carried out for each of the scenario listed with codes written for the algorithm and the results then visualized in a virtual environment. The AEB was actuated in every case where the autonomous vehicle assessed that there was a potential crash.
Keywords: Autonomous vehicle, T-intersection, traditional and e-bicycles, threat assessment and decision making algorithm, Autonomous Emergency Brake.