This course provides an overview of current methods for planning. These include so-called task planning methods, where the problem is to compute actions or policies that achieve certain goals; the sub-problem of scheduling, that is, to compute when actions should take place such that temporal constraints are upheld and/or resource conflicts are avoided; and motion planning, that is the problem of finding a sequence of motions for a mobile platform that brings into a desired pose. These three problems are related in that they are often all sub-problems of a given robotic planning problem, hence the course also illustrates methods for solving these problems jointly.
The specific topics covered in the course are:
- classical planning, STRIPS as a representation and algorithm, planning as search, Graphplan and planning as satisfiability
- planning with Hierarchical Task Networks
- planning under uncertainty, MDPs and POMDPs
- constraint-based resource scheduling
- motion planning for non-holonomic robots
- coordinated motion (integrated motion planning and scheduling)