Trajectory prediction for autonomous driving: Progress, limitations, and future directions

As the potential for autonomous vehicles to be integrated on a large scale into modern traffic systems continues to grow, ensuring safe navigation in dynamic environments is crucial for smooth integration. To guarantee safety and prevent collisions, autonomous vehicles must be capable of predicting the future trajectories of surrounding traffic agents. Over the past decade, significant efforts from both academia and industry have been dedicated to designing solutions for precise trajectory forecasting. These efforts have produced a vast range of approaches, raising questions about how they differ, which challenges they address, and what limitations still remain in trajectory prediction. This paper reviews a substantial portion of recent trajectory prediction methods, proposing a taxonomy to classify existing solutions. It also provides a general overview of the prediction pipeline, covering input and output modalities, modeling features, performance evaluation metrics, and existing prediction paradigms. In addition, the paper discusses active research topics within the forecasting realm, addresses the posed research questions, and highlights the remaining research gaps and challenges. • An extensive overview of the prediction pipeline, which includes discussions of input modalities, modeling features and approaches, output modalities, evaluation metrics, and prediction paradigms. • Taxonomy of modeling approaches, offering a structured classification of methods. Each category includes a comparative table summarizing key models and their characteristics for quick reference. • Literature review of active research areas in the trajectory prediction domain, covering topics such as uncertainty estimation, integration of driving knowledge, planning-conditioned prediction, vision-language model (VLM)-guided prediction models, and collaboration between vehicles. • An open code repository featuring two prediction pipelines to facilitate experimentation, benchmarking, and the development of new methods.