Motivation: As human-centric manufacturing is becoming more prevalent in the aftermath of COVID-19, metrics to determine the feasibility of robotic programming methods are becoming more critical. However, mixed reality interfaces are empirically designed and validated using jury studies. This limits the applicability of system-tailored human-robot interfaces in manufacturing environments. 

Approach: This research aims to evaluate native and mixed reality-based programming techniques for assembling a NIST task board. By studying metrics for evaluation of mixed reality interfaces. this research aims to provide a systematic framework for developing mixed reality interfaces.

Partnerships: This research is sponsored by the NIST PREP Program and done in collaboration with the Performance of Human-Robot Interaction Project at NIST.

Motivation: As Industry 4.0 and 5.0 data are becoming more complex and multimodal in nature, AI algorithms for control and decision-making are becoming more complex. Hence, explaining and interpreting advanced AI predictions is becoming more critical. However, a modular approach towards interpreting these complex multimodal AI models are lacking, which limits their applicability in operator-AI teaming.

Approach: This research aims to develop novel algorithms and frameworks for explaining and interpreting AI predictions based on multimodal data including time series, images, text, and video data naturally collected in the manufacturing process.

Partnerships: This work is done in collaboration with Dr. Rui Liu at the Rochester Institute of Technology. 

Motivation: Existing methods for evaluating the sound quality of home appliances are confined to enumerative methodologies. At first glance this appears to be a prudent approach to validate the sound performance. However, this does little to guide the product development team that is working on new products or disruptive technologies due to the lack of competitive sets for precompetitive technologies.

Approach: This research aims to develop novel methods and frameworks to systematically study psychoacoustics of industrial and home product machines to drive optimal product design. However, there is a lack of knowledge of general methods to systematically quantify the sound quality of a design. Therefore, this research aims to fill this gap in knowledge to promote consumer happiness and improve the efficiency of the sound quality design process. 

Motivation: As manufacturing systems and data become more complex and integrated, cyber-physical representations via AI models will be more critical towards analytics, control, and connectivity of the production ecosystem. However, manual preparation and processing of data is required to develop production-acceptable cyber-physical representations and ML models, which is time-consuming to individually conduct for every manufacturing system. In addition, the accessibility of ML developers who can perform such tasks is limited, especially for rural manufacturers.

Approach: This research aims to investigate a Automated Data Preparation (ADP) framework that aims to solve the missing link towards a one-stop-shop cyber-manufacturing software that will automatically develop and deploy AI models for integrated manufacturing processes and data modalities. This research investigates systematic preparation of data based on manufacturing systems and modalities and the selection of the data preparation and AI models from the infinite combination of methods and models. 

Partnership: This work has been supported by Michigan Tech's Institute of Computing and Cybersystems.