The increasing proportion of elderly individuals poses significant challenges for mobility assistance and independent living. Rollators are widely used mobility aids, yet conventional devices remain largely passive and provide no feedback on improper usage that may lead to discomfort or musculoskeletal strain. This paper presents the design and evaluation of a smart posture detection system integrated into a commercially available rollator as part of the AktiMuW project. The proposed system combines multiple sensors - including inertial measurement units, ultrasonic distance sensors, and strain gauges - with machine learning techniques to assess rollator usage and user posture in real time. A two-stage classification approach is employed. First, the operational state of the rollator is identified using supervised learning methods. Feed-forward neural networks, convolutional neural networks, and random forest classifiers are evaluated, with the random forest model demonstrating the best balance of accuracy and computational efficiency, achieving over 97% validation accuracy across all device states while significantly reducing inference time and resource usage. Second, user posture is analyzed using unsupervised k-means clustering. Different posture granularities are investigated, ranging from five detailed posture classes to simplified configurations. A three-class posture model (“Comfortable,” “Too Close,” and “Too Far”) is selected as an optimal compromise between classification accuracy and actionable feedback, achieving validation accuracies of up to 99%. The complete system is deployed on an embedded NVIDIA Jetson Orin Nano platform and integrated via MQTT-based communication. Real-time benchmarking confirms that the combined models operate within acceptable computational limits while maintaining reliable posture detection. The presented approach demonstrates the feasibility of lightweight, sensor-based posture monitoring for rollator users and provides a foundation for future assistive feedback systems aimed at improving safety and comfort for elderly individuals.