Human motion prediction for tasks involving obstacle avoidance is critical for digital human modeling, robotics, and ergonomics. This study compares two approaches for predicting minimum clearance distance during upper extremity reaching tasks: an expanded 3D space version of the Steering Dynamics Model (SDM) and a perceived risk-based optimization motion prediction. The optimization-based method integrates biomechanical constraints and Bayesian Decision Theory to model perceived risk, while the SDM predicts emergent paths based on attractor-repeller dynamics. Both methods were tested using experimental data from fifteen participants, who performed reaching tasks around a 3D obstacle recorded with an IMU-based motion capture system. Results show that both methods improve minimum clearance distance predictions compared to a purely artificial sphere obstacle avoidance constraints approach. The SDM provides a computationally efficient alternative to the optimization-based approach while maintaining accuracy. However, the optimization-based method with perceived risk more closely aligns with experimental data, demonstrating the importance of cognitive modeling. The point cloud obstacle representation proved effective in both approaches. Future work should explore parameter tuning, subject-specific adaptations, and additional cognitive modeling techniques to enhance accuracy. These findings improve digital human simulations and real-time human-robot interaction models by integrating biomechanical and cognitive factors in motion prediction.