Multiclass-SGCN: Sparse Graph-based Trajectory Prediction with Agent Class Embedding
This work solves trajectory prediction for mixed road user types, which is important for autonomous driving and robotics, but it is incremental as it builds on existing graph-based methods with specific improvements.
The paper tackled the problem of predicting trajectories for multiple types of road users (e.g., pedestrians, cars, cyclists) by addressing issues like superfluous interactions in dense graphs, and it resulted in significantly outperforming state-of-the-art methods on the Stanford Drone Dataset with more realistic predictions.
Trajectory prediction of road users in real-world scenarios is challenging because their movement patterns are stochastic and complex. Previous pedestrian-oriented works have been successful in modelling the complex interactions among pedestrians, but fail in predicting trajectories when other types of road users are involved (e.g., cars, cyclists, etc.), because they ignore user types. Although a few recent works construct densely connected graphs with user label information, they suffer from superfluous spatial interactions and temporal dependencies. To address these issues, we propose Multiclass-SGCN, a sparse graph convolution network based approach for multi-class trajectory prediction that takes into consideration velocity and agent label information and uses a novel interaction mask to adaptively decide the spatial and temporal connections of agents based on their interaction scores. The proposed approach significantly outperformed state-of-the-art approaches on the Stanford Drone Dataset, providing more realistic and plausible trajectory predictions.