Positional Health Assessment of Collaborative Robots Based on Long Short-Term Memory Auto-Encoder (LSTMAE) Network

Calibration is a vital part of ensuring the safety and smooth operation of any industrial robot and this is particularly essential for collaborative robots as any issue pertaining to safety can adversely impact the human operator. Towards this aim, Prognostics and Health Management (PHM) has been widely implemented in the context of collaborative robots to ensure safe and efficient working environments. In this research, as a subset of PHM research, a novel positional health assessment approach based on a Long Short-Term Memory auto-encoder network (LSTMAE) is proposed. An experimental test setup is utilised, wherein the collaborative robot is subject to variations of coordinate system positional error. The operational 3-axis position time-series data of the collaborative robot is collected with the aid of an industrial data acquisition platform utilising influxDB. The experiments show that, with the aid of this approach, manufacturers can assess the positional health of their collaborative robot systems.