Modeling expertise and adaptability in virtual operator models

Abstract To advance construction machine design and testing, model-based design and virtual operator models (VOMs) can be used to explore machine designs virtually. However, current VOM efforts have been restricted to mimicking known trajectories, recorded from actual machine operations. Previous work developed a VOM to use in closed-loop simulation with an excavator model. To advance the utility of model-based machine testing, the fidelity of the VOM was enhanced along three lines: 1) representation of expert work cycle operation, 2) adaptation to changes in work site environment and 3) adaptation to changes when operating different machines. To represent expertise, work cycle task overlap was modeled – a hallmark of expert human operator performance. A mental model was developed to adapt to changes in the work site environment. Finally, the VOM was generalized to adapt to changes in excavator dimensions, eliminating the need for time intensive “tuning” typical of trajectory-dependent models. Three case studies demonstrated task overlap modeled productivity gains typical of expert operators, VOM control outputs adapted as trench depth and pile height increased, and the VOM adapted to different excavator models automatically. An additional case study compared VOM results to human-recorded data. This work advances the ability to integrate human expertise and adaptability in virtual operator modeling, resulting in a more realistic simulation of operations.