Development of a detailed canine gait analysis method for evaluating harnesses: A pilot study

Abstract

Dog harnesses are becoming more popular, with their large variety stemming from the idea that different dogs and scenarios require different types of harnesses. While their benefits over collars are self-explanatory, there is a lack of research on their effect on gait, and even the existing studies examine only a limited set of parameters. The goal of present study was to establish a method capable of quantifying canine gait in detail. Based on 3D motion capture, the developed method allows for the examination of 18 joint angles and 35 spatio-temporal parameters throughout multiple gait cycles, and can be used to analyze canine movement in detail in any kind of scenario (e.g. comparing healthy and lame dogs, or measuring the effect of training). The method is presented through the measurement of how different harnesses affect walking kinematics compared to free (unleashed) movements. Four dogs with varying body sizes and breeds and multiple types of harnesses were included. Marker data was filtered using a zero-lag 6th order Butterworth-filter with a cutoff frequency of 20 Hz. The normality of the spatio-temporal and joint range of motion parameters was tested using the Anderson-Darling test (p = 0.05), with most parameters passing in 60+% of test cases. Swing time and range of motion of the sagittal aspect of spinal angle at T1 vertebrae failed more regularly, both resulting from the measurement setup rather than the actual parameters being not normally distributed. Two-sample Kolmogorov-Smirnov tests (p = 0.05) were used to compare each parameter’s distribution between cases, showing that most parameters are significantly altered by the harnesses in about 2/3rd of the cases. Based on the results, there’s no absolute superior harness, however, it is possible to select the best fit for a specific dog and application, justifying their large variety.

Lifelike Agility and Play on Quadrupedal Robots using Reinforcement Learning and Generative Pre-trained Models

“…Next, we quantify how naturalistic are the generated movements from the PMC using gait analysis. Gait analysis is a well-established technique used in motion analysis [29] and has wide range applications in veterinary medicine, e.g., the diagnostic of lameness in dogs [30,31]. Following [31], we record kinematic measurements of each joint and compute the spatiotemporal parameters, including swing time, stance time, step distance and step height of the learned behaviors.…”

“…The swing time is the period while a foot is not in contact with the ground; the stance time is the period when the foot is in contact with the ground; the step distance is the distance of the given limb's heel strike from the opposite side limb's heel strike in the direction of movement; and the step height is the peak of the given limb's toe height for each swing time. For more details please refer to [31]. To achieve fair evaluation, we record these kinematic parameters on ŕat terrain in simulation and setup four comparison groups.…”

“…Gait analysis is a well-established technique used in motion analysis [29] and has wide range applications in veterinary medicine, e.g., the diagnostic of lameness in dogs [30,31]. Following [31], we record kinematic measurements of each joint and compute the spatiotemporal parameters, including swing time, stance time, step distance and step height of the learned behaviors. The swing time is the period while a foot is not in contact with the ground; the stance time is the period when the foot is in contact with the ground; the step distance is the distance of the given limb's heel strike from the opposite side limb's heel strike in the direction of movement; and the step height is the peak of the given limb's toe height for each swing time.…”