You should Look at papers in the planning field and the mathematical tools used behind them. Look at open source codes such as Apollo, and write some code yourself. Set up simulation environments such as Carla to verify whether your code is effective. Make sure you have some understanding of what is being done in the field.
It is best to participate in some mass production projects to learn how companies give and achieve R&D goals based on application scenarios, product definitions, regulations and other requirements. The focus of learning at this time is to make yourself professional and systematic, know the environment for running PnC algorithm code and how to design and implement this environment, know the advantages and disadvantages of mainstream PnC algorithms in mass-produced cars, and how professional teams make choices and optimizations, know how car companies and suppliers can collaborate on a large scale, and independently undertake a small part of the R&D tasks.
The third stage may be for students who are interested in studying for a doctorate, trying to solve the current industry problems in the field of autonomous driving decision-making and planning. For example, how to ensure that decisions are complete and accurate in complex scenarios that cannot be exhaustively enumerated; how to use the minimum cost to complete the actual vehicle verification of hundreds of millions of kilometers required for L4, etc. Many students skip the second stage and go directly to the third stage. This is a way of writing papers. In the short term, they may gain something and get a doctoral offer from a prestigious university, but in the long run, due to the lack of engineering background, they will still take a detour after work. If they stay in academia, they will not be able to truly promote the development of autonomous driving and planning. They will just continue to write papers and evaluate their professional titles.
When choosing a new direction as a future career, we need to start from the demand side and see what knowledge and skills are needed by mainstream companies in the recruitment of positions in this direction. The following are the recruitment requirements for autonomous driving planning positions in most companies. Knowledge required for planning positions (1) Basic planning algorithms (A*, RRT*, Lattice based, MPC, POMDP, etc.); (2) Parametric curve construction (polynomial, Bezier, spline curves); (3) Commonly used numerical optimization methods; (4) Familiarity with C++ programming. The above is the basic and necessary knowledge for planning positions, and it is also the goal. After determining the goal, the next step is how to do it. I estimate that if you study hard, it will take 1 year to reach a level of familiarity. Of course, this depends on your personal ability. After you have mastered the necessary basics, find an internship.
For more informtion, you can visit Robotic Search Engine to find latest research papers, AI Agent Marketplace and AI Agent Search Engine to find AI Agents related topics reviews and latest product information.