In many business applications, workflows are used to describe business processes. Employees and machines get instructions from a plan (the workflow) to be guided or controlled. The workflows make it easier to create and manage business processes. Therefore, using workflows is the standard procedure in the business area today. The distributed execution of workflows plays an important role as almost all nodes are connected to a network today. The importance even increases with the emerging of pervasive environments. Because these systems are prone to failures, it is important to develop reliability methods that ensure that the system works properly even if failures occur. When the robustness of a system in a distributed environment shall be increased, the service that has to be executed is usually replicated and executed by two or more nodes. This means that the exact same behavior is executed by multiple nodes and thereby increases the reliability of the system by being able to cope with node failures. Changing the order of the activities or using alternative activities to increase the robustness is promising because when each node receives a different workflow that achieves the same goal, the possibility of failures should be further reduced by decoupling the replicas in respect of time and hardware dependencies. We developed a robustness metric that evaluates the robustness of a set of workflow replicas. We also developed methods and algorithms that generate workflows with different orders and alternative tasks within reasonable time. Our evaluations show that our proposed methods work significantly better than deploying a brute-force method to achieve the same behavior.