Robotics and Autonomous Systems 105 (2018) 59–68 Contents lists available at ScienceDirect Robotics and Autonomous Systems journal homepage: www.elsevier.com/locate/robot Realization of human gait in virtual fluid environment on a robotic gait trainer for therapeutic purposes ✩ Tayfun Efe Ertop *, Tolga Yuksel, Erhan ilhan Konukseven Department of Mechanical Engineering, Middle East Technical University, Dumlupinar Bulvari No. 1, Cankaya, 06800, Ankara, Turkey highlights • A novel control strategy for robotic gait trainers was developed using virtual fluid environment. • Precise control over gait assistance and resistance was achieved with adjustable virtual fluid. • A real-time gait phase detection algorithm utilizing only kinematic gait data was developed. • Gait characteristics comparable to the literature were observed with the robotic system. • Significant changes in healthy subject gaits were observed by varying virtual fluid parameters. article info Article history: Received 22 June 2017 Received in revised form 21 December 2017 Accepted 12 February 2018 Keywords: Robotics Rehabilitation Assistive Medical Aquatic therapy Locomotor training abstract Patients with disorders such as spinal cord injury, cerebral palsy and stroke can perform full gait when assisted, which progressively helps them regain the ability to walk. A very common way to create assistive effects is aquatic therapy. Aquatic environment also creates resistive effects desired for strength building. In this study, realization of a virtual fluid environment on a robotic gait trainer is presented as an alternative method. A model was created to determine torques and forces acting on the human body while performing gait in a fluid environment. The developed model was implemented on a robotic gait trainer. By adjusting the virtual fluid model parameters, precise control over assistive and resistive effects during gait was achieved without enforcing any pre-defined gait pattern. The real-time gait phase information required by the fluid model to determine torques was provided with a developed algorithm which only uses kinematic gait data. Experiments with healthy subjects were done using the robotic gait trainer to verify the gait phase algorithm, and to compare gait characteristics obtained in virtual land and water environments with the literature. Additional experiments were performed with the robotic system to assess effects of changing fluid model parameters to healthy subject ga it characteristics. The results show that force and torque effects of virtual fluid environment on robotic gait trainer were achieved. The gait phase algorithm was shown to provide smooth transition between phases. Also, significant changes in gait characteristics were observed by modifying fluid model parameters. © 2018 Elsevier B.V. All rights reserved. 1. Introduction Gait rehabilitation devices aim to help patients with locomotor dysfunction in lower extremities due to conditions like spinal cord injury (SCI), cerebral palsy, and stroke perform gait exercise so that the walking ability can be restored. Different methodologies are used to tackle this problem. One of the earliest solutions was ✩ Research supported by BAMA Technology, Turkey. Patent pending. * Corresponding author. E-mail addresses: ertop@metu.edu.tr (T.E. Ertop), yuksel.tolga@metu.edu.tr (T. Yuksel), konuk@metu.edu.tr (E.i. Konukseven). manually assisted treadmill training in which more than one ther- apist moves patient’s legs by hand [1–4]. Labor intensive nature of this therapy forced researchers to seek for alternative methods. Aquatic therapy and robotic gait training are both viable options. Water can provide safe environment for patients to walk without the fear of falling meanwhile buoyancy effects both assist their lower extremity and decrease the apparent weight [5]. Some pa- tients are able to walk in aquatic environment even if they cannot walk on land [6]. Combined buoyancy and drag forces in aquatic environment deliver assistive and resistive effects needed for the https://doi.org/10.1016/j.robot.2018.02.012 0921-8890/© 2018 Elsevier B.V. All rights reserved.