| Optimal motion planning for hydraulic robots is an important problem that must be solved as the traditional domains of hydraulic machines, e.g. construction, excavation, forestry etc., embrace varying levels of automation and robotics. There is almost no previous work on optimal motion computation for hydraulic robots. Most work done to date on optimal motion planning has focused on electric drive manipulators, which are fundamentally different from hydraulic robots due to the physical nature of fluid power actuation versus electric power. Also, hydraulic robots suffer from significant interactions between different joint actuators due to their coupling through the hydraulic system; this is in addition to the non-linear linkage dynamic coupling. These differences preclude adapting previous approaches for hydraulic robots.; In this thesis, we propose and demonstrate an approach to computing optimal motions for hydraulic robots; this problem has never been addressed before in the literature. Although we only optimize free-space motions (due to the difficulty in modeling soil-tool interaction) we allow different end-effector loads. We use a robust search technique (Simulated Annealing) to search the robot's discretized command space for the temporal command sequence that minimizes an objective function while performing a given task (with specified start and end states) subject to obstacle avoidance and kinematic and dynamic constraints. The objective function for the optimization can be composed of any quantifiable measures such as time, energy, joint forces etc. Our approach is the first to allow the use of a diverse set of measures in the objective function. All existing methods focus almost exclusively on time-optimality.; During the search, we use a robot model to evaluate the cost of a candidate command sequence. This hydraulic robot model was constructed using a novel approach, also developed in this research. This modeling approach allows the construction of computationally inexpensive hydraulic robot models that capture the significant actuator interactions that are typical of hydraulic robots.; We demonstrate our optimization approach by computing optimal motions for a Caterpillar 325A hydraulic excavator (HEX) robot. We present examples of computing time-optimal motions for 7 different tasks, as well as energy-optimal motions for the same 7 tasks. The time-optimal motions for all 7 tasks are demonstrated on a HEX testbed. The optimal motions computed for 4 of the 7 tasks are compared to that of a human expert performing the same tasks. The comparisons show that the computed optimal motions are as good as, or better, than the expert's motions. The energy-optimal motions offer some interesting insights into the operation of the HEX, while demonstrating the power of our method in handling different measures in the objective function. |
