Stop Overplanning!

However, overplanning at every turn takes too much time. Instead, you typically plan only the next few turns and rely on a rough estimate of how much time and effort the rest of the journey will take.

The challenge with this approach is that these short-term decisions often lead to inefficient routes due to inaccurate estimates of future travel effort. My research develops a theoretical framework to improve these estimates, so that short-term planning behaves more like a long-term planner. To illustrate this, imagine there is a map, in which a very experienced taxi driver notes how long it takes to reach the airport from each intersection, so now you can reference to this map to improve your route choice, even without full visibility into the entire trip.

This approach reduces the calculation time needed for replanning while ensuring near-optimal travel decisions, and it strikes a balance between efficiency and accuracy. While the traffic example illustrates the core idea, my research focuses on developing a general theoretical framework that can be applied to other problems, such as autonomous driving, robotic arm manipulation, and power grid control. Specifically, I work on improving short-term planning in a data-driven way, by using information from precomputed long-term solutions. The information is extracted with some special mathematical tools to approximate more accurately the future efforts for short-term planning. Therefore, just as a taxi driver uses an accurately marked map to avoid overplanning and navigate efficiently, our framework empowers decision making to blend short-term agility with long-term vision.