Position prediction using a Recurrent Neural Network.
Predicting the walking paths of pedestrians is an important part of realizing autonomous driving. In this research work, 12 future positions of pedestrians should be predicted based on 8 positions. The data was measured with a laser scanner. The pedestrians were moving at a street intersection. To solve this regression problem, a Recurrent Neural Network (RNN) was implemented with a many-to-one architecture. Tensorflow/Keras was used to implement the architecture.
