Skip to content


Switch branches/tags

Name already in use

A tag already exists with the provided branch name. Many Git commands accept both tag and branch names, so creating this branch may cause unexpected behavior. Are you sure you want to create this branch?

Latest commit


Git stats


Failed to load latest commit information.
Latest commit message
Commit time

Maneuver-based Anchor Trajectory Hypotheses at Roundabouts

We extend existing recurrent encoder-decoder models to be advantageously combined with anchor trajectories to predict vehicle behaviors on a roundabout. Drivers’ intentions are encoded by a set of maneuvers that correspond to semantic driving concepts. Accordingly, our model employs a set of maneuver-specific anchor trajectories that cover the space of possible outcomes at the roundabout. The proposed model can output a multi- modal distribution over the predicted future trajectories based on the maneuver-specific anchors. We evaluate our model using the public RounD dataset.

model image

RounD Dataset Pre-processing

The RounD Dataset is a new dataset of naturalistic road user trajectories recorded at German roundabouts. Download the dataset to the 'rounD' directory, then run the following MATLAB script:


This will do the required pre-processing, split the dataset into train, validation and test subsets, and save such subsets into the 'data' directory.

Model Arguments

The default network arguments are in: 

You can set the required experiment arguments in this script. For example:

  • args['ip_dim'] selects the input dimensionality (2D or 3D).

  • args['use_intention'] and args['use_anchors'] are Boolean variables that choose whether using intention prediction and anchor trajectories or not.

Model Training and Evaluation

The model structure is coded in ''. After setting the required experiment arguments, create a 'trained_models' directory to save the trained models. You can start model training by running:

To test a trained model, first create an 'eval_res' directory, then run:

which will load and test the trained model defined by the selected model arguments. The RMSE results will be saved as csv files to the 'eval_res' directory.


If you find this code useful for your research, please cite our work:

  • Mohamed Hasan, Evangelos Paschalidis, Albert Solernou, He Wang, Gustav Markkula and Richard Romano, "Maneuver-based Anchor Trajectory Hypotheses at Roundabouts", preprint 2021.


This project is licensed under the MIT License - see the file for details.


No releases published


No packages published