Spatiotemporal Scene-Graph Embedding for Autonomous Vehicle Collision Prediction

In autonomous vehicles (AVs), early warning systems rely on collision prediction to ensure occupant safety. However, state-of-the-art methods using deep convolutional networks either fail at modeling collisions or are too expensive/slow, making them less suitable for deployment AV edge hardware. To address these limitations, we propose SG2VEC, a spatiotemporal scene-graph embedding methodology that uses the graph neural network (GNN) and long short-term memory (LSTM) layers predict future via visual scene perception. We demonstrate SG2VEC predicts 8.11% more accurately 39.07% earlier than method synthesized data sets, 29.47% challenging real-world set. also show is better state of art transferring knowledge from synthetic sets driving sets. Finally, performs inference $9.3\times $ faster with an 88.0% smaller model, 32.4% power, 92.8% energy industry-standard Nvidia DRIVE PX 2 platform, it implementation edge.