Researchers have developed MPCFormer, an approach for explainable socially-aware autonomous driving. It uses physics-informed and data-driven models to understand social interactions in traffic scenarios. The model is based on a Transformer architecture and learns from naturalistic driving data. It has been tested on various datasets and shown to outperform other approaches in terms of accuracy and safety. MPCFormer's results demonstrate its ability to generate human-like behaviors when interacting with surrounding traffic, while also mitigating potential safety risks associated with learning-based methods. --- Why it matters: This matters because autonomous vehicles still struggle to interact with their environment in a way that is similar to humans. MPCFormer addresses this issue by providing a more accurate and explainable model of social interactions, which can improve the safety and efficiency of autonomous driving systems. Source: https://arxiv.org/abs/2512.03795