Test-Time Training of Trajectory Prediction via Masked Autoencoder and Actor-specific Token Memory

Enhancing trajectory prediction through test-time training with a masked autoencoder and actor-specific token memory.