Guaranteed Computation in Robotics
M. Kieffer (L2S – CNRS - Supélec, Univ Paris-Sud, France), R. Pepy (L2S – CNRS - Supélec, Univ Paris-Sud, France) and É. Walter (L2S – CNRS - Supélec, Univ Paris-Sud, France)
Robotics is a challenging and important field of application for validated numerical methods. Problems are usually solved using local or random search algorithms, with no guarantee as to their results. A definite advantage of interval analysis is that it provides tools to obtain approximate but guaranteed results that take into account the uncertainty in the data and errors resulting from the finite nature of the representation of real numbers on a computer. This talk will present recent results and open problems in the application of guaranteed methods to robot localization and tracking, simultaneous localization and mapping (SLAM), or path planning.