Development of a Context-Aware Assistive System for Manual Repair Processes - A Combination of Probabilistic and Deterministic Approaches

Abstract

Looking at the trend of mass customization of products, companies are confronted with an increasing number of different products and product variants. Especially for humans working in assembly and rework domain it is increasingly difficult to maintain an overview of all assembly paths and work steps. To tackle this problem, context aware assistance systems are introduced to the field. Although there is a lot of research in the area of context aware assistive systems, most work focuses on fixed work plans or purely probability-based activity recognition. As a result, these systems either restrict the worker’s personal way of working or the modelling of the work plans is complex and time-consuming. The goal of our approach is a context sensitive support for all performable steps at a given time. That system requires an intuitive modelling of work processes including an activity recognition. Therefore, we present a process model consisting of the combination of a petri net and aspects of Hidden Markov Models (HMM). Based on the modelled work process, the system determines performable work steps at the given time, while the activity recognition selects the most likely work step on the subset of feasible work steps. In this paper we describe the structure of the process model and how petri nets and approaches from the area of HMM are combined in this model. The outlook shows an implementation approach for the model.