go back


Prediction of a Highway Lane Change

We combine two prediction methods: First, a prediction based on the contextual situation of the vehicle, whose behavior we want to predict, which we call context-based prediction. Second, a prediction based on physical evidence of the vehicle’s movement. Mathematically, we combine two probabilities:

The context-based prediction evaluates relations between a group of relevant vehicles to estimate .
It classifies the current situation to predict a possible lane change. Context-based prediction provides early information before any lateral movement can be observed.
To increase reliability, physical prediction uses observations of recent vehicle positions. These positions are compared with typical trajectories to predict behaviors. This prediction is highly accurate, but only capable of predicting behaviors once they started. The combination achieves early and reliable prediction.


Prediction is the estimation of how the future will evolve. It can be as diverse as the trajectory of a thrown ball, the weather, or the discussion between a number of people. Even though most of us have experienced the uncertainty in predicting the weather, the latter example is qualitatively different because of the individual freedom of humans to decide on their future behavior. Systems need to correctly predict the intentions of other traffic participants, cooperation partners or decision makers to behave successfully.

At HRI-EU we decompose the prediction task into a variety of different sequences each following a certain qualitative pattern of assumptions and constraints. These sequences can merge and diverge depending on the continuous perception of the environment. If a decision is required sequences are combined taking their different levels of reliability into account. An illustrative example is the combination of context-based and physical prediction outlined on the previous page.

For more information

J. Schmuedderich et al., “A novel approach to driver behavior prediction using scene context and physical evidence for intelligent adaptive cruise control (i-ACC)”, Proc. 3rd Int. Symp. FAST-zero, 2015, pp. 85-92.