Constructing Fast Hydraulic Robot Models for Optimal Motion Planning

Computing optimal motions for any robot requires a good model, and a method to compute the optimal motions using that model. As research is conducted into automating operations in construction, excavation etc. there arises a need to compute optimal motions for the hydraulic machines used in these areas. Hydraulic machines disallow a simple extension of work done previously on optimal motion planning for electric drive robots. We have constructed a fast model for a hydraulic excavator(HEX) that can capture the non-linear actuator interactions. This model can simulate 75 secs of machine motion in 1 sec. of real time on a Sun Sparc 20. We use a set of neural networks to approximate the actuator response functions. We use the HEX model with a simulated annealing optimization method to compute time-optimal motions for the HEX, for defined start and end states. We demonstrate the efficacy of the constructed model and show results from using it in optimal motion computation. Real testbed results are shown in both cases. This is the first time that such a result has been reported in the literature for hydraulic machines.