Objective-based framework for motion planning under sensing and control uncertainties

The authors consider the problem of determining robot motion plans under sensing and control uncertainties. Traditional approaches are often based on methodology known as preimage planning, which involves worst-case analysis. The authors have developed a general framework for determining feedback strategies by blending ideas from stochastic optimal control and dynamic game theory with traditional preimage planning concepts. This generalizes classical preimages to performance preimages and preimage planning for designing motion strategies with information feedback. For a given problem, one can define a performance criterion that evaluates any executed trajectory of the robot. The authors present methods for selecting a motion strategy that optimizes this criterion under either nondeterministic uncertainty (resulting in worst-case analysis) or probabilistic uncertainty (resulting in expected-case analysis). The authors present dynamic programming-based algorithms that numerically compute performance preimages and optimal strategies; several computed examples of forward projections, performance preimages, and optimal strategies are presented.