Predicting Ego-Bicycle Trajectory : An LSTM-based Approach Using Camera and IMU

In order to make bicycles with driver assistance systems a reality, a suitable trajectory prediction must be developed. This research will investigate sensor and trajectory prediction models that are suitable for the task, and develop such a model. In addition, two datasets consisting of sensor data collected with bicyles are created to train the model. Two different types of Long Short Term Memory (LSTM) trajectory prediction models are evaluated, one using a Convolutional Neural Network LSTM hybrid architecture, and one using only LSTM by itself. The performance of these different models is then evaluated and compared.