Real-Time Sit-To-Stand Phase Classification with a Mobile Assistive Robot from Close Proximity Utilizing 3D Visual Skeleton Recognition
Anas Mahdi, Zonghao Dong, Jonathan Feng-Shun Lin, Yue Hu, Yasuhisa Hirata, Katja Mombaur
AI summary
Problem
Assistive robots require real-time user state monitoring to provide safe support, but vision-based pose estimation becomes highly inaccurate and noisy when users are within one meter of the camera.
Approach
The system uses a wide-angle depth camera and Google's MediaPipe framework to extract 14 kinematic features from a 3D skeleton model, which are then fed into machine learning classifiers to identify movement phases in real time.
Key results
- Validated MediaPipe 3D skeleton tracking accuracy against Vicon ground truth at close range
- Identified 14 critical kinematic features for reliable sit-to-stand phase classification
- Compared SVM, Decision Tree, Random Forest, and XGBoost classifiers for phase estimation
- Demonstrated real-time phase classification on the SkyWalker robot despite close-proximity distortion
Why it matters
Enables safer, adaptive mobility assistance for older adults by allowing low-cost assistive robots to accurately interpret user movement phases without expensive motion capture hardware.
Abstract
Sit-to-stand (STS) transfer is a fundamental but challenging movement that plays a vital role in older adults’ daily activities. The decline in muscular strength and coordination ability can result in difficulties performing STS and, therefore, the need for mobility assistance by humans or assistive devices. Robotics rollators are being developed to provide active mobility assistance to older adults, including STS assistance. In this paper, we consider the robotic walker SkyWalker, which can provide active STS assistance by moving the handles upwards and forward to bring the user to a standing configuration. In this context, it is crucial to monitor if the user is performing the STS and adapt the rollator’s control accordingly. To achieve this, we utilized a standard vision-based method for estimating the human pose during the STS movement using Mediapipe pose tracking. Since estimating a user’s state from extreme proximity to the camera is challenging, we compared the pose identification results from Mediapipe to ground truth data obtained from Vicon marker-based motion capture to assess accuracy and reliability of the STS motion. The fourteen kinematic features critical for accurate pose estimation were selected based on literature review and the specific requirements of our robot’s STS method. By employing these features, we have implemented a phase classification system that enables the SkyWalker to classify the user’s STS phase in real-time. The selected kinematics from vision-based human state estimation method and trained classifier can be furthermore generalized to other types of motion support, including adaptive STS path planning and emergency stops for safety insurance during STS. Manuscript received: October, 7, 2024; Accepted December, 21, 2024. This paper was recommended for publication by Editor Angelika Peer upon evaluation of the Associate Editor and Reviewers’ comments. Anas Mahdi, Jonathan Lin, and Katja Mombaur are with the Department of System Design Engineering, CERC Human-Centred Robotics and Ma- chine Intelligence, University of Waterloo, Waterloo, Canada {a7mahdi, jf2lin, katja.mombaur}@uwaterloo.ca Yue Hu is with the Department of Mechanical Engineering, Active & Interactive Robotics Lab (A.I.R.), University of Waterloo, Canada yue.hu@uwaterloo.ca Katja Mombaur is with the Karlsruhe Institute of Technology (KIT), Insti- tute of Anthropomatics and Robotics (IAR), Optimization and Biomechanics for Human-Centred Robotics, Karlsruhe, Germany Zonghao Dong and Yasuhisa Hirata are with the Department of Robotics, Tohoku University, Sendai, Japan {z.dong, hirata}@srd.mech.tohoku.ac.jp 1 Anas Mahdi and Zonghao Dong are equal first authors of this paper.We gratefully acknowledge funding from the Canada Excellence Research Chair program, the Japan Science and Technology Agency Moonshot R&D Program (JPMJMS2034), the Japan Society for the Promotion of Science KAKENHI (22J10961), and the Tohoku University Graduate Program for Integration of Mechanical Systems. We have obtained ethics approval from the University of Waterloo ethics board to conduct human-involved experiment with SkyWalker. Digital Object Identifier (DOI): see top of this page.