Task scheduling for dual-arm industrial robots through Constraint Programming

Sammanfattning: In a society where more and more production becomes automated it demands robots that are as flexible and versatile as humans. Such flexibility demands automatic scheduling of tasks. In this thesis we approach the problem using Constraint Programming and through a case study we present a model for a dual-armed robot that is able to deal with a more flexible workload. We also introduce filters to cut down the runtime of the solver. To evaluate the model we tested it on 6 solvers; G12/FD, JaCoP, Gecode, or-tools, Opturion CPX and Choco3. The results show that the model can produce a solution as good as the one manually implemented for the case study. We introduce filters on the domains of some of the variables and they made an improvement on the runtime for many of the solvers. We also found that the runtime of the solvers varied a lot and could range from several hours to just a few milliseconds using the same data. Unfortunately, in many of the tests the solvers did not complete their searches within the time limit of 4 hours. In some cases when using MiniZinc version 2.0.1, the solvers were not able to read the FlatZinc files. The fastest solver in our tests was Gecode using MiniZinc version 2.0.1.