A New AI-Based Approach for Contextualization and Prediction of Human Activities in Industrial Robot Applications

Human activity prediction in industrial environments with content semantics using 3D point clouds can ensure human safety and process efficiency. This work aims to train and optimize context-based AI models to develop a framework for anticipating human group activities and estimating human motion. Main Contributions of this work: 1. Design of multi-layer structure of various DNN Classifiers to segment, detect and track dynamic agents (e.g., humans, robots, or AGVs) 2. Training of context-based sequential relational anticipation model to predict human activities and positions in the early stage 3. empirical experiment with 60 subjects for collecting datasets of human actions and activities in on industrial setting. All these procedures are merged in the proposed framework and evaluated in six scenarios of one industrial Use-Case.