Managing stress is crucial for maintaining health and improving the quality of life. Consequently, researchers have focused on stress detection in two or three levels. In recent years, efforts have been made to continuously estimate stress, but the introduced methods have limitations in terms of plausibility and automation. This paper presents an automated method for continuous stress estimation using a fuzzy logic approach. To implement the proposed method, a database containing electromyogram, electrocardiogram, hand and foot galvanic skin response, heart rate, and respiration signals was used in conditions of rest, city driving, and highway driving. After preprocessing, the signals were segmented using a sliding window, and features that exhibit monotonic changes with respect to stress were extracted from each window. These features were clustered into five clusters in the feature space using fuzzy clustering. A fuzzy inference system was designed to determine the weight of these clusters, which has three inputs and one output. The design of this fuzzy system is automated and based on the membership values of the clusters for each state and the label of each state. Finally, by combining the membership values obtained from fuzzy clustering and the weight of each cluster, a continuous value called estimated stress was determined. The proposed method showed a correlation of 0.80 between the estimated stress and the average scoring by experts for the test conditions. The unique features of this method include the automatic tuning of the fuzzy inference system using hidden knowledge in the data and the lack of need for experts. The automation of the proposed method can improve human-machine interaction and transfer stress assessment from laboratories to daily life.