Tracking Human Pose During Robot-Assisted Dressing Using Single-Axis Capacitive Proximity Sensing

Zackory Erickson, Maggie Collier, Ariel Kapusta, Charles C. Kemp

TL;DR

This paper introduces a novel capacitive proximity sensing method that allows robots to accurately track human arm movement and adapt to pose estimation errors during dressing assistance. This approach overcomes limitations of vision-based methods, especially when the body is occluded by clothing, enabling more robust and responsive robotic assistance for individuals with motor impairments.

What Problem Does This Paper Address?

Robots hold great promise for assisting individuals with motor impairments in daily tasks like dressing, which can significantly enhance independence and privacy. However, current robotic dressing systems often struggle with two key challenges: accurately estimating and tracking the human user's pose in real-time, and adapting to errors or changes in that pose during the dressing process. Vision-based pose estimation methods, commonly used in robotics, are particularly problematic during dressing because clothing visually occludes the body, and the robot's own arms can further obstruct the view. This paper tackles the problem of enabling robots to reliably assist with dressing by providing a sensing method that can overcome these visual limitations and dynamically adjust to human motion and pose uncertainties.

Methods

The core of this paper is a single-axis capacitive proximity sensing approach. Unlike vision-based methods, this technique directly estimates the distance between the robot's end-effector and a person's arm. The researchers derived a closed-form equation based on the parallel plate capacitor model to convert capacitance measurements into distance estimates. This equation, $d(\Delta C) = \frac{\alpha}{\Delta C + \beta}$, relates the change in capacitance ($\Delta C$) to the distance ($d$).

The capacitive sensor itself consists of a copper foil electrode attached to an acrylic sheet, connected to a Bare Conductive Touch Board microcontroller. This setup allows for high-frequency (over 200 Hz) sampling with low latency and minimal computational requirements. The sensor is mounted on the robot's end-effector, enabling it to "feel" the proximity of the human arm.

For control, a Cartesian controller on the PR2 robot was used, with a PD controller specifically designed to regulate the distance between the end-effector and the arm in the vertical (Z) direction. The desired distance was set to 5 cm, and the controller aimed to maintain this gap, allowing the robot to follow the arm's contours and movements.

Datasets

The paper does not describe the use of pre-existing datasets for training or evaluation. Instead, the researchers collected their own data through experiments with human participants. This involved measuring capacitance and distance as the robot's end-effector moved towards and along the participants' arms at various initial distances. This experimental data was used to fit the capacitance-to-distance function and evaluate the control strategy.

Key Insights and Ideas

  • Capacitive Sensing as a Solution to Visual Occlusion: The most significant insight is that capacitive sensing provides a robust alternative to vision for tasks like dressing where visual occlusion is a major hurdle. The sensor can "see" through clothing, a capability that traditional cameras lack.
  • Real-time Adaptation to Pose Errors: The method allows the robot to dynamically adjust its trajectory based on real-time distance measurements, effectively compensating for initial pose estimation errors (up to 15 cm) and ongoing human movement. This is crucial for safe and effective assistance.
  • Low-Latency, Low-Computational Sensing: The capacitive approach offers a significant advantage in terms of speed and computational load compared to complex vision processing, making it suitable for real-time control and responsive interaction.
  • Generalizability Across Individuals: The capacitance-to-distance relationship was found to be remarkably consistent across all 10 participants, suggesting that the method does not require individual calibration and can generalize well to a wider population.
  • Contact Detection and Force Mitigation: The system can also detect when contact occurs, and the control strategy ensures that any incidental contact during movement involves low forces, enhancing user safety.

Key Results

  • The robot successfully pulled the sleeves of a hospital gown and a cardigan onto the arms of 10 human participants.
  • The system could adapt to initial pose estimation errors of up to 15 cm.
  • The capacitive sensing method was unaffected by visual occlusion from clothing, including long-sleeve shirts.
  • The robot maintained an average tracking error of approximately 1.5 cm from the participant's arm during dressing, even with arm motion.
  • Capacitance measurements showed low variability across participants, indicating good generalization.
  • The fitted capacitance-to-distance function achieved a high coefficient of determination ($R^2$) of 0.969.

Limitations and Future Work

  • Single-Axis Sensing: The current system only estimates distance along a single axis. To track 3D human motion (vertical, lateral, and rotational), multiple capacitive sensors would be required.
  • Electromagnetic Interference (EMI): The unshielded sensor is susceptible to EMI, which can affect accuracy, particularly when the robot's arm is fully extended. Active shielding is suggested as a future improvement.
  • Clothing Material Sensitivity: Certain clothing types with metal threads or decorations might interfere with capacitance measurements.
  • Limited Sensing Range: The maximum effective sensing range is approximately 10 cm.

Future work suggested includes extending the system to 3D motion tracking, incorporating active shielding to mitigate EMI, and exploring more advanced signal processing and learning techniques to handle complex interactions and multiple sensors.

Summary updated

Abstract

Dressing is a fundamental task of everyday living and robots offer an opportunity to assist people with motor impairments. While several robotic systems have explored robot-assisted dressing, few have considered how a robot can manage errors in human pose estimation, or adapt to human motion in real time during dressing assistance. In addition, estimating pose changes due to human motion can be challenging with vision-based techniques since dressing is often intended to visually occlude the body with clothing. We present a method to track a person's pose in real time using capacitive proximity sensing. This sensing approach gives direct estimates of distance with low latency, has a high signal-to-noise ratio, and has low computational requirements. Using our method, a robot can adjust for errors in the estimated pose of a person and physically follow the contours and movements of the person while providing dressing assistance. As part of an evaluation of our method, the robot successfully pulled the sleeve of a hospital gown and a cardigan onto the right arms of 10 human participants, despite arm motions and large errors in the initially estimated pose of the person's arm. We also show that a capacitive sensor is unaffected by visual occlusion of the body and can sense a person's body through cotton clothing.
BibTeX
@article{erickson2018tracking,
  author = {Zackory Erickson and Maggie Collier and Ariel Kapusta and Charles C. Kemp},
  title = {{Tracking Human Pose During Robot-Assisted Dressing Using Single-Axis Capacitive Proximity Sensing}},
  journal = {IEEE Robotics Autom. Lett.},
  year = {2018},
  url = {https://openreview.net/forum?id=qR8J_pjo0TR},
}
Figure