Abstract: The trigger-action programming (TAP) is an end-user programming paradigm widely-used in the field of the IoTs, which allows users to define rules for executing actuator actions when sensor events occur. Its correct execution depends on the reliable operation of the device, however, in real environments, the device may have execution abnormalities. Aimed at the execution exception issue of TAP program, a method for handling exceptions in TAP program based on semantic abstraction of devices and MAPE-K loop is proposed. The exception detection was performed by ontology modeling and TAP expansion rules, and a dynamic compensation strategy was executed to repair the detected exceptions. The feasibility, efficiency, and ease of use of the method were verified by the TAP program exception handling framework. The experimental results show that the system can handle the exceptions faced in an efficient way by selecting alternative devices, improving the program robustness and its self-adaptive capability, and guaranteeing that the end-user requirements are realized.