Recently aerial robots have become a popular platform for
load transportation. In this paper we are interested in
solving the problem trajectory tracking with the suspended
load without previously knowing or defining the trajectory
of the quadrotor. Exploiting the knowledge that this system
is differentially flat we use it to control the trajectory
of the suspended load. As a solution we propose a method
based on reinforcement learning algorithm called
least-square policy iteration (LSPI). The control input for
the quadrotor is obtained from the LSPI algorithm in order
to drive the quadrotor in the direction that ensures that
the load tracks the given trajectory. This method is model
free which makes it robust to model uncertainties and it
can be implemented as an online learning algorithm. The
proposed method is verified through simulation and
experiments.