Influence of Prior Knowledge on Cyclist's Trajectory Forecast

The ability to predict the movement of other traffic participants is crucial for ensuring safety and comfort in shared traffic spaces. While human drivers can often identify behavioral indicators of other road users to anticipate their actions, automated vehicles lack this ability. This is especially important in interaction situations with vulnerable road users, such as cyclists, which are in the focus of this work. Trajectory predictions through machine learning models are a common method to estimate movements of those traffic participants. This paper investigates the impact of incorporating prior knowledge about cyclists' driving intention on the accuracy of trajectory forecasts. The study builds on previous research that utilized neural networks for trajectory predictions and proposes a method to incorporate prior knowledge as an additional input to an existing model. The influence of prior knowledge in different scenarios has been evaluated using different methods. The results indicate that incorporating prior knowledge yields improvements in the accuracy of trajectory predictions, demonstrating the potential of this approach for enhancing the safety and effectiveness of cyclist-vehicle interactions.