Robust systems are designed to deal with some anticipated, though probably unusual, situations such as equipment failure and overloads. They must also be designed to deal with unanticipated situations: no matter how much careful planning is carried out, unexpected events may occur. This paper describes a system, EPL 5.0, for sensing unusual situations and responding appropriately. This paper also describes an event processing language that helps business users specify both conditions that require responses, and the appropriate responses to these conditions. Unanticipated situations are dealt with by specifying normal conditions, and defining an anomaly as a significant deviation from normality. The system also facilitates machine learning of normal and anomalous conditions from examples.