TruST (Trunk Support Trainer) is a cable-driven robot where a thoracic belt is pulled by wires by actuators mounted on a fixed frame. The wires are controlled to apply a force field on the upper body in response to the body’s motion. The wires provide assist-as-needed forces for rehabilitation, training, and strengthening of the upper body.
Recently, we introduce deep learning into TruST's real-time controller design. We collected 4320 trials of 3D trunk movement data (about 1.5 million data frames) during the postural star sitting test from 45 healthy subjects in two conditions (with or without foot support). It is the largest 3D trunk movement dataset in the literature and might be helpful in other seated postural control research.
The dataset will be available to the research community upon request consistent with the IRB guidelines, please contact with Xupeng Ai ([email protected]).
The Trunk Support Trainer (TruST) is a cable driven robotic system for postural training of individuals with postural instabilities. It allows controlling of forces and moments along a region of the human trunk. TruST utilizes four cables attached to each corner of an adaptable but rigid torso belt, while the other end of the cable is routed through a pulley and connected to an AC servo motors attached to the bottom of a fixed frame. A load cell and spring are attached in series with each cable, for measuring and controlling the tension in each cable.
TruST is capable of applying assistive and resistive forces and perturbations along any region of the trunk, in the transverse plane. In addition, one of the unique capabilities of the TruST is its force field control strategy for posture training. This allows an individual to train at their individualized boundary of stability. By training at one's stability limits, he or she can explore their workspace boundary and improve their intersegmental coordination.
The dataset includes trunk movement data collected from 45 healthy young adults, 14 females and 31 males (age 23.9±2.9 yrs, height 169.7±7.2 cm, weight 63.6±11.5 kg). Subjects were randomly assigned to two groups, 23 conducted the postural star sitting test with foot support and the other 22 without. The dataset contains data from 4320 trials (45 subjects × 12 rounds × 8 trials). Our collected data are time sequenced. The dataset is structured as a design matrix with 1476463 rows and 15 columns. Rows denote time frames and columns denote features. The 15 features are raw signals sensed by the infrared motion capture system and are the x, y, z coordinates of the five markers in TruST’s global reference frame.