Shared Autonomy For Kinesthetic Tools

Lead Research Organisation: University of Edinburgh
Department Name: Sch of Informatics


Remote control of manipulators in industry, such as concrete spraying, is generally repetitive and requires a high level of concentration and manual dexterity. Excessive cognitive loads can lead to fatigue that can become dangerous. The development of sensing devices (eg RGB-D cameras) and compliant feedback will allow shared control robotic systems to improve workflow and reduce operator burden leading to a safer working environment for workers, improved joy quality, and reduced costs.

The intention of this work is to address the scientific problem of developing computationally efficient approaches for shared autonomous control systems that blend kinesthetic operator input and sensory data that modulate the motion of a robotic device performing a manipulation task. We will develop methods on the KUKA LWR arm (fixed-base arm) and Dual UR5 Clearpath Husky (dual arm mobile-base) platforms to demonstrate the platform independence of this work.

Studentship Projects

Project Reference Relationship Related To Start End Student Name
EP/N509152/1 01/10/2015 30/09/2020
1795793 Studentship EP/N509152/1 01/09/2016 30/11/2020 Christopher Mower
Description In areas such as concrete spraying, remote control of these machines is generally repetitive and requires a high level of concentration and manual dexterity. Excessive cognitive loads can lead to fatigue becoming dangerous for the operator and his or her colleagues working in shared workspaces with these devices. The operator controls the machine remotely through a joystick controller manipulating every joint simultaneously to perform a task, for example concrete spraying. This type of human-system interaction can be unintuitive for the operator and since there is no concept of collision avoidance or task optimality built into the control the danger posed to workers can be exacerbated by random errors or oversights in safety regulations from over confidence. The advent of higher quality and cheaper multi-modal sensing apparatus will facilitate the development of shared autonomous methods that reduce the cognitive load on the operator. Our work consists of three main strands. The first strand explores what control modes are most intuitive for an operator to supply motion commands. The second strand explores shared autonomy approaches for real-time motion adaptation for remote controlled robots that simultaneously ensure coverage and collision avoidance. The third strand explores methods to ultimately minimize the human-system
interaction whilst achieving high task performance. The idea is that the operator will act more like a supervisor by influencing the motion constraints rather than supplying motion commands to the robot. Subsequent innovation implementing this work could have many important benefits to the construction sector, among others, such as improved safety for workers, increased quality and efficiency, a reduction in cost for specialized training,
minimized wasted material.
The first contribution has been a study into the various modes of teleoperation for unskilled users for target acquisition tasks. We have found that lower dimensional task spaces elicit highest performance for users with little experience. We have also found that users who are deemed to play computer games on a regular basis elicit higher performance. This work is being submitted to IEEE CASE 2019 conference.
A shared autonomous system has been developed to demonstrate how to handle the mth-order Markovian nature of robotics teleoperation problems. This work will be submitted to a future robotics conference.
Exploitation Route * developers of shared autonomous systems can use the results generated by my work to develop safer and more efficient systems for assistive teleoperation.
Sectors Aerospace, Defence and Marine,Construction,Education

Description NCE TechFest 2018 submission 
Form Of Engagement Activity Participation in an activity, workshop or similar
Part Of Official Scheme? No
Geographic Reach National
Primary Audience Professional Practitioners
Results and Impact * submitted to NCE techfest 2018 with title "Conducting research in robotics for innovation in construction"
* submission was a finalist
Year(s) Of Engagement Activity 2018