Deep Learning-Based Acceleration-Aware Trajectory Forecasting
Author Information
Author(s): Ali Asghar Sharifi, Ali Zoljodi, Masoud Daneshtalab
Primary Institution: Mälardalen University
Hypothesis
Can incorporating acceleration data improve trajectory forecasting for autonomous vehicles?
Conclusion
The DAT model significantly enhances trajectory forecasting accuracy, especially for objects with nonlinear motion patterns.
Supporting Evidence
- DAT outperformed state-of-the-art methods in trajectory forecasting.
- The model achieved up to a 2× improvement in forecasting accuracy.
- Extensive experiments were conducted on the NuScenes dataset.
Takeaway
This study shows that using acceleration data helps predict where cars and pedestrians will go, making self-driving cars safer.
Methodology
The DAT model processes sequential sensor data to detect objects and forecast their future trajectories using a novel forecasting module that incorporates acceleration data.
Limitations
The model's performance may vary with different window lengths for acceleration calculation.
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website