I am a newly enrolled doctoral student, having embarked directly on my PhD after completing my undergraduate studies. The trend towards end-to-end technology as the technical route in autonomous driving seems unstoppable, and it appears that the rapid iteration of this technology will render many jobs of autonomous driving engineers obsolete, as it is poised to replace them. Previously, the research route in autonomous driving was based on perception-decision-planning-control, but end-to-end technology is expected to supplant these intermediate steps, making the job market even more challenging. My doctoral research direction is likely to focus on image processing, fusion perception, and the like. However, the conditions of my research group are limited, and we are unable to conduct research on end-to-end and large models due to their high computational demands (number of GPUs) and the need for extensive data training, which my current research group cannot support. I would like to ask for advice on how a new doctoral student like me, aiming to engage in research on autonomous driving environment perception, should proceed. From which aspects should I develop my research work, and in which direction should I strive to align with the technological iterations?