Online path parameterization for manipulators with input/state constraints

Summary form only given. Tracking a given geometric path r(s) in the presence of physical constraints is a task which often occurs in robotic applications. The paper introduces a device called the path governor (PG) which, according to a prediction of the evolution of the robot from the current state, generates online a suitable time-parameterization s(t) of the path to be tracked, by solving at fixed intervals a constrained scalar optimization problem. We assume that a feedback controller has been already designed in order to guarantee, in the absence of constraints, nice stability and tracking properties. The PG attempts to reduce the computational complexity in two ways: first, only a portion of the desired path is considered at a time; second, the resulting sub-trajectory depends only on a scalar parameter-its end-point. As for predictive controllers, these simpler planning processes evolve according to a receding horizon strategy. Then, a new parameterisation is evaluated which replaces the previous one. This provides robustness against both model and measurement disturbances. The selection of the end-point is performed by considering two objectives: (i) minimize the traversal time, i.e. the time required to track the desired path, and (ii) guarantee that the constraints are and will be fulfilled i.e., no blind-alley is entered. Higher level switching commands are also taken into account by simply associating a different optimization criterion to each mode of operation.