I am interested in sequential decision-making problems of the type that appear in the areas of AI Planning and Heuristic Search. Often these problems involve one or more agents moving through a task environment. Our job in this case is to find a collision-free trajectory that brings each agent from an initial start location and to a desired target position. Such problems appear in many settings including Robotics, GPS Navigation and in Computer Games. They are challenging for a variety of reasons: (i) we need high quality solutions; (ii) we need those solutions fast; (iii) there are often operational variables and constraints that need to be accounted for (e.g., the world might be dynamically changing, the agent might have incomplete information); (iv) even if we can find a plan, its successful execution might be derailed (e.g., agents can be affected by unexpected delays or even failures).
In my research I aim to develop fast algorithms that are useful in practice but which also come with guarantees. For example "completeness", which is an algorithmic promise to return a solution if one exists. Also "bounded sub-optimality" which is an algorithmic promise that says the cost of any returned solution is no more than some fixed multiplicative factor larger than the best possible cost.
For more info about what I do (and links to papers and source code), please check out my homepage: