A Review on Scene Prediction for Automated Driving

Abstract

Towards the aim of mastering level 5, a fully automated vehicle needs to be equipped with sensors for a 360∘ surround perception of the environment. In addition to this, it is required to anticipate plausible evolutions of the traffic scene such that it is possible to act in time, not just to react in case of emergencies. This way, a safe and smooth driving experience can be guaranteed. The complex spatio-temporal dependencies and high dynamics are some of the biggest challenges for scene prediction. The subtile indications of other drivers’ intentions, which are often intuitively clear to the human driver, require data-driven models such as deep learning techniques. When dealing with uncertainties and making decisions based on noisy or sparse data, deep learning models also show a very robust performance. In this survey, a detailed overview of scene prediction models is presented with a historical approach. A quantitative comparison of the model results reveals the dominance of deep learning methods in current state-of-the-art research in this area, leading to a competition on the cm scale. Moreover, it also shows the problem of inter-model comparison, as many publications do not use standardized test sets. However, it is questionable if such improvements on the cm scale are actually necessary. More effort should be spent in trying to understand varying model performances, identifying if the difference is in the datasets (many simple situations versus many corner cases) or actually an issue of the model itself.

Self-evaluation of automated vehicles based on physics, state-of-the-art motion prediction and user experience

“…For a quantitative assessment of , we choose a Deep Learning-based supervised approach, in particular, two models with different degrees of complexity: a simple LSTM-based model which is a common approach for handling data sequences, and VectorNet 20 , a more sophisticated model that employs graph representations as a fundamental component. The latter was selected based on its performance (good trade-off between low displacement error and large prediction horizons 5 ) and availability of open-source implementations, as it is a prominent model frequently cited in the motion prediction community.…”

“…It has to be distinguished between data-driven and model-driven approaches. The former use Machine Learning and Deep Learning techniques achieving prediction horizons of up to 8 s while the latter, often physics-based models, can reliably predict less than 2 s 5 , 12 . A commonly used approach of data-driven models in the field of vehicle motion planning is Deep Reinforcement Learning (DRL).…”

“…The future traffic situation can be predicted well with Deep Learning models. Such data-driven models show the best performance in anticipating the future situation for several seconds, clearly outperforming physical models 5 . Still, a prediction time of 10 s is challenging and only possible in very few situations, such as straight forward highway driving with moderate vehicle speeds.…”