Page - 201 - in Proceedings - OAGM & ARW Joint Workshop 2016 on "Computer Vision and Robotics“

Trajectory planning based on activity recognition and identification of low-level process deviations Sriniwas Chowdhary Maddukuri 1 , Gerald Fritz 1 , Sharath Chandra Akkaladevi 1 , Matthias Plasch 1 , and Andreas Pichler 1 1 Department of Robotics and Assistive Systems Profactor GmbH {Sriniwas,Maddukuri}@profactor.at Abstract Improving work efficiency and ensuring safety of the human worker while the human worker and robot simultaneously perform the tasks in close proximity is one of the key research topics in human-robot cooperation. Given a process which contains a set of tasks or process steps performed within the shared human-robot workspace, a methodology for the robot’s trajectory planning will be mentioned in this concept paper. The methodology will be based on activity recognition and identification of low-level process deviations. Here, the low-level process deviations which occur from the robot assistant side are mainly focussed. 1. Introduction A key requirement in the field of human-robot cooperation is to realize the process execution in a safe and time-efficient manner. Here, process refers to a list of process steps/tasks performed simultaneously by the human worker and robot assistant. To achieve safe execution of shared human-robot tasks, a process monitoring component which identifies low-level process deviations is a pre-requisite. In the context of shared human-robot tasks, deviations are often classified into robot assistant side deviations and human worker side deviations. Robot assistant side deviations are defined as unexpected events like unreachable goal configuration, grasp failure reported by the robot’s tool and high probabilistic existence of collision-prone trajectories with the nearby static or dynamic objects while the robot performs an object manipulation task in the shared workspace. Human worker side deviations are defined as expected events like performing spatial sequence of actions or activities and unexpected transition between the tasks or process steps. Process deviations from the human worker side are not considered within this work. The motivation behind this research work is to come up with a trajectory planning framework which can identify and handle low-level process deviations with respect to the simultaneous recognition of human activities and process steps/tasks. In this research work, the handling of process deviations will also be mentioned. 1.1. Related Work Recent work which deals with trajectory planning is based on prediction of human actions and activities to achieve spatio-temporal synchronization in shared human-robot tasks. The 201