I have a complex set of mechanical linkages and components, and I am trying to create a feedback control system for trajectory-tracking with the end-effector. They system is driven by a set of prismatic joints (representing ideal series-elastic actuators). At this stage I am simply trying to establish the control parameters and the required forces to achieve desired positions/velocities in the workspace, and have not yet inserted any realistic actuator characteristics.

An open-loop solution can be found using translational position control on the prismatic joints, this is no problem. However to close the loop properly, I need to convert the system to use a differential input, such as a velocity driver. At this point everything becomes very finicky and highly dependent on initial conditions. Generally Maplesim gives up, claiming that no solution can be found. Due to the complexity of the system, there is no way for me to guess a priori the appropriate initial conditions of all the joints and linkages. I was wondering if anybody had any tips for either pre-determining the initial parameters, or alternatively for relaxing the solver in some way to allow for a wider exploration of possible solutions. Or any other ideas!

So far: I have tried setting up an open loop position-driven system with the same trajectories and using Maple to read all the initial conditions, then transferring them as guesses to my closed-loop system. But this has not worked. As a temporary workaround I am using position drivers on the SEAs, which are in turn driven by a velocity controller (with an integrator term). However this is suboptimal. Any help would be much appreciated!