Automated generation of robust error recovery logic in assembly systems using genetic programming

Abstract

Automated assembly lines are subject to unexpected failures, which can cause costly shutdowns. Generally, the recovery process is done "on-line" by human experts or automated error recovery logic controllers embedded in the system. However, these controller codes are programmed based on anticipated error scenarios and, due to the geometrical features of the assembly lines, there may be error cases that belong to the same anticipated type but are present in different positions, each requiring a different way to recover. Therefore, robustness must be assured in the sense of having a common recovery algorithm for similar cases during the recovery sequence.The proposed approach is based on three-dimensional geometric modeling of the assembly line coupled with the genetic programming and multi-level optimization techniques to generate robust error recovery logic in an "off-line" manner. The approach uses genetic programming's flexibility to generate recovery plans in the robot language itself. An assembly line is modeled and from the given error cases an optimum way of error recovery is investigated using multilevel optimization in a "generate and test" fashion. The obtained results showed that with the improved convergence gained by using multi-level optimization, the infrastructure is capable of finding robust error recovery algorithms. It is expected that this approach will require less time for the generation of robust error recovery logic. Keywords: Automated Assembly Systems, Error Recovery, Genetic Programming, Multi-Level Optimization IntroductionAn unexpected failure is an unavoidable phenomenon, which causes the automated assembly lines to halt their operation. These failures can bring out drastic results in economical issues. As indicated in the results of the EUREKA project, 1 initiated to benchmark maintenance in Scandinavian countries in 1992, approximately 30% of the time spent on maintenance is used for unforeseen repairs, 20% for preventive maintenance, and 37% for planned repairs. A similar survey in the United States showed that excessive maintenance costs were approximately $200 billion in 1990.The diagnosis and recovery from such failures are normally handled by on-line investigation of the assembly line by human experts, which means costly shutdown of the assembly lines. Another approach is using controller codes. It is stated in Zhou and DiCesare 2 that in automated systems up to 90% of the control coding effort is based on error recovery by using programmable logic controller (PLC) codes. However, these PLC codes are programmed by humans based on "expected" error scenarios and are deficient in dealing with "unexpected" scenarios, leaving the recovery process to manual labor work. A novel approach 3 to deal with the unexpected failures is off-line synthesis of error diagnosis and recovery logic based on the three-dimensional geometry-based modeling of an entire assembly line. Generation of unexpected error cases can be accomplished by using Monte Carlo simulation of the assembly process...

Off-line error prediction, diagnosis and recovery using virtual assembly systems

“…Then, the next step is the off-line logic synthesis for error diagnosis and recovery from the predicted error scenarios. At this step, Bayesian reasoning is used for identifying the most probable failure while genetic algorithms (GAs) are used to generate recovery logic as discussed in our previous work in detail (Baydar and Saitou, 2001a). The reason for using genetic algorithms of genetic programming arises from the fact that most of the time same error can occur in numerous con®gurations in the workspace (i.e., part jamming).…”

“…The performance of the error recovery logic is tested in a generate and test fashion (Baydar and Saitou, 2001a) such that, several recovery logic algorithms are generated with the GP engine and tested with the commercial software package. Then, the results of this evaluation are supplied to the GP engine and improved recovery logic is generated based on the obtained results Saitou, 2001b, 2001c).…”

“…In our previous work (Baydar and Saitou, 2001a), a GP module was developed because the previous version of the software package (Workspace 4 UserManual, 1998) was not capable of translating collision recovery path to controller codes. However, in this study, rather than GP, GAs are used to create a path for collision error recovery by representing the sequence of reference points given to the robot controller.…”

“…This is natural since these systems are composed of many working parameters such as dimensional variations of the products, ®xtures, sensor capabilities and robot repeatability and when these working parameters are coupled with the 3-D workspace, it is dif®cult to anticipate all of the error conditions with their likelihood of occurrence and 3-D-state in the workspace (Baydar and Saitou, 2001a). One example for these serious error conditions is the propagated errors, which result from the propagation of undetected errors during the assembly process.…”

“…* 3-D-states of the possible errors are not included (Baydar and Saitou, 2001a). * Most of the systems are de®cient in dealing with multiple error conditions (Sampath et al, 1996).…”