Premature Convergence in Motion Planning of Nonholonomic Systems and how to Counteract it
Arkadiusz Mielczarek, Ignacy Dulȩba · 2019
In this paper an undesirable phenomenon of a premature convergence in nonholonomic motion planning is discussed. It relies on reaching a configuration where energy expensive motions are necessary to approach a goal configuration. An algorithm is proposed to counteract this phenomenon and to choose some sub-goals which allow for energy effective motions. The motion towards the sub-goals is evaluated using some Lie algebraic tools. The algorithm is illustrated on a simple unicycle robot. Simulations reveal that using the proposed algorithm some energy gains can be achieved, resulting paths are shorter and easy to localize within a configuration space.