[Paper Review] Characterizing Lifting and Lowering Activities with Insole FSR sensors in Industrial Exoskeletons

Luis A. Mateos|arXiv (Cornell University)|Jun 16, 2017

Prosthetics and Rehabilitation Robotics13 references3 citations

TL;DR

This paper proposes a low-cost, modular insole FSR sensor system with eight force sensors (two modules per foot) to detect dynamic weight changes during lifting and lowering tasks in industrial exoskeletons. By monitoring heel and metatarsal pressure variations, the system enables real-time detection of lifting and lowering activities with 15% average error, allowing the exoskeleton to trigger assistive force without user hand control.

ABSTRACT

This paper presents an insole FSR (Force Sensing Resistor) to dynamically detect weight variations in an exoskeleton system. The proposed methodology is intended for tasks of lifting and lowering heavy weights with an industrial exoskeleton to substantially reduce spinal loads during these manual handling activities. Instead of extensively placing high-dense force sensors by covering the whole plantar area, as most of commercial applications do, we integrate only a few force sensors in specific plantar area, so that the sensory system is not restricted to an individual foot size and shape, and on the other hand has relatively lower material cost. Industrial exoskeletons are intended to assist workers when handling heavy goods. With this in mind, wearers are not able to use their hands to control the exoskeleton since they use them to handle the goods. Therefore, the exoskeleton controller is required to indirectly infer how much and when the wearer requires assistance for lifting or lowering a heavy weight. Our approach of dynamically detect and characterize the increment/decrement of weight, as well as the rising/falling edge, enables the exoskeleton's controller to trigger the request of assistive force to the actuators.

Motivation & Objective

• Develop a low-cost, non-intrusive sensory system for industrial exoskeletons that does not restrict foot size or shape.
• Address the challenge of hand-free control in industrial settings where workers need both hands free for lifting.
• Enable real-time detection of lifting and lowering activities using minimal force sensors instead of full-plantar coverage.
• Reduce material cost and improve portability compared to high-density commercial plantar pressure systems.
• Integrate the sensor system into common footwear or exoskeletons for industrial deployment.

Proposed method

• Use 8 thin-film FSR sensors (4 per foot) arranged in two modular units: one under the heel and one under the metatarsals.
• Integrate sensors into a flexible insole that fits standard footwear and exoskeletons without customization.
• Implement a real-time algorithm that detects weight changes over 0.25-second intervals to identify lifting and lowering phases.
• Use base weight (user's standing weight) as reference to detect pre-lifting (weight drop) and pre-lowering (weight rise) states.
• Trigger lifting detection when weight increases above base weight after a pre-lift flag; trigger lowering when weight drops below base weight after a pre-lower flag.
• Calibrate each sensor individually due to unique force-resistance curves to improve accuracy.

Experimental results

Research questions

• RQ1 Can a minimal set of FSR sensors in strategic plantar locations reliably detect lifting and lowering activities in industrial exoskeletons?
• RQ2 How accurate is the weight estimation using only heel and metatarsal sensors compared to ground truth?
• RQ3 Can dynamic weight changes (e.g., pre-lift drop, pre-lower rise) be used to reliably infer user intent without direct control inputs?
• RQ4 To what extent does the modular sensor design maintain performance across different foot sizes and shapes?
• RQ5 What is the achievable accuracy and cost-effectiveness of such a system compared to full-plantar coverage systems?

Key findings

• The insole FSR system successfully detected lifting and lowering activities with an average error of ±15% in weight estimation across all test subjects.
• The system achieved reliable detection of lifting and lowering phases by identifying dynamic weight shifts: a drop before lifting and a rise before lowering.
• The modular sensor configuration (heel and metatarsal modules) allowed the system to be non-restrictive to individual foot size and shape.
• The system demonstrated portability and low material cost by using only eight FSR sensors per foot, significantly reducing cost compared to full-plantar coverage systems.
• The algorithm correctly identified four lifting/lowering cycles in Experiment I and three in Experiment II, with consistent detection of pre-lift and pre-lower states.
• The FSR system measured forces from 0 to 160 kg per foot, with a maximum capacity of 720 kg, and showed consistent performance across repeated trials.