Dimensionality
& Complexity: Revoking the Two Curses
Ameet Deshpande, UCSD
Advisor: Bill McEneaney
Recently,
a curse-of-dimensionality-free algorithm for solution of
Hamilton-Jacobi-Bellman PDEs was developed at UCSD. This suffered from
a curse of complexity, as the number of quadratic bases used to
approximate the value function grew exponentially with propagation
horizon. Through use of better pruning techniques, this curse of
complexity has been partially abated, making the algorithm more useful
for practical problems.