Real-time continuous workload assessment is important for researchers and developers of tools that aim to reduce human operators' cognitive workload, especially in dynamic environments, as the military environment, where task demands and workload change rapidly. Most workload measurement techniques provide a single retrospective value or require expensive high-end sensing equipment. This study aimed to introduce an affordable continuous machine learning (ML) based workload assessment tool, that can provide real-time workload scores. Using experienced military unmanned aerial vehicle (UAV) operators in a simulated operational setting, muscle behavior represented by their interaction with a joystick was modeled to predict Subjective Workload Assessment Technique (SWAT) scores. Data were obtained from six professional participants. Four machine learning (ML) modeling methodologies were tested on each participant's data. It has been shown that after running an ML setup phase for each participant, an already in use available tool as the UAV joystick controller can be used to predict SWAT scores at any given time. By implementing the approach presented in this study, researchers can more accurately evaluate various aspects of the human operator's cognitive workload, and developers can evaluate the progression of their solutions on operators' cognitive workload over time.