This repo contains the implementation of algorithm IO-TAD from the paper Online Trajectory Anomaly Detection Based on Intention Orientation. It includes code for data processing, Inverse Reinforcemnet Learning(IRL) and online anomaly detection.
- Python 3.8
- Tensorflow
- Scikit-learn
We will transfer the orginal GPS trajectories to grid trajectories with state-action pairs, and derive the destination-based trajectory clusters. For synthetic anomalies, we provide the code of anomaly generation for three types of anomaly: Wrong-destination anomalies, detour anomalies and random walk anoamlies.
The code implements deep maximum entropy IRL.
To implement IO-TAD on Chengdu dataset, please run main.py from the file online detection.
Chen Wang, Sarah Erfani, Tansu Alpcan, Christopher Leckie
This work was supported in part by the Australian Research Council Linkage Project under the grant LP190101287, and by Northrop Grumman Mission Systems’ University Research Program.