Responding to Subjective Changes of Customer Requirements in Dynamic Service Execution Environment

When a customer uses a service, he might change his initial requirement due to subjective reasons that could not be predicated by service providers. If this happened, service providers need to adjust the current service solution to adapt the new requirement with the objective of minimizing the change amplitude and cost. Placing this problem into AI-planning based service composition (PSC) scenario, we present two approaches called global re-planning algorithms (RP) and local reinforcement algorithm (LR). RP constructs a virtual requirement according to the composite service´s current execution state and the changed expectation, and invokes PSC algorithm to look for a new solution. A price-rewritten mechanism is used as a heuristic during the re-planning to preferentially reuse those services existing in current solution. In terms of six basic types of requirement changes, LR attempts to make minor repairs to the current composite service to adapt to the new requirement and minimize the change amplitude of the solution. In the experiments, the efficiency, variation cost and amplitude of RP and LR are compared, and how the performance metrics are affected by other factors is preliminarily validated.