Sport climbing has gained popularity over the past decade, which is expected to grow since its inclusion in the Tokyo 2020 and Paris 2024 Olympic Games. This has created increased interest in quantitative climber performance data for coaching, athlete feedback, and ensuring safety. Although coaches currently rely on their trained observational skills, there is a lack of detailed data on the interactions between climbers' limbs and holds. Additionally, beginners often start climbing without guided support, resulting in poor techniques and elevated injury risk. Thus, there is a need for methods that provide a comprehensive evaluation of climbing performance by combining measurements of climbers' interactions with holds and their overall movements. This thesis describes the design conceptualisation, fabrication, and validation of a low-cost instrument for measuring forces exerted on climbing holds.

Previous designs for measuring forces on climbing holds were often limited to laboratory settings due to their high cost and the need for specialised sensors. More recently, low-cost and unobtrusive designs have emerged, but there is a recognised gap in instrumented climbing holds that are affordable, non-intrusive, and easy to install on existing climbing walls without requiring structural modifications.

In this study, a standard M10 socket head cap screw was modified to function as an instrumented mounting screw (IMS). The screw shank was machined to have a smaller diameter, and a channel was milled down the threads. Finite element analysis (FEA) proved the machined IMS was still fit for purpose (FOS = 1.128). Initially, three strain gauges were bonded to the shank at 0°, 120° and 240°, orientated in the axial direction. Benchtop testing determined the optimal strain gauge configuration around the shank to accommodate variable load magnitudes and directions on the climbing hold. These tests also identified the effect of the point of force application on strain measurements, with anteroposterior centre of pressure (COP) changes being the most influential. Consequently, the IMS design was modified to feature a double-triplet strain gauge configuration, with two sets of three strain gauges bonded at 0°, 120° and 240° around the shank and positioned at 15 and 20mm from the screw head. Sensors were connected to a data acquisition system via flexible printed circuit boards (PCBs) routed through the channel to the back of the climbing wall. All sensors and PCBs were mechanically protected with epoxy. An additional washer was inserted between the climbing hold and the wall to enhance strain measurements.

Due to the complex behaviour of the screw, hold, washer, and climbing wall system, a calibration procedure using neural networks (NN) was developed to address non-linearities and minimise crosstalk in the strain gauge measurements. Reference measurement systems with off-the-shelf sensors were used to optimise the NN architecture and hyperparameters. The NN was trained and verified for estimating 3D forces with resulting relative errors of 10%, 17%, and 17.6% in the vertical, lateral, and anteroposterior directions. R2 values for the force measurements were 0.89 for the lateral direction, 0.92 for the vertical direction, and 0.88 for the AP direction. Further testing for force magnitude recognition yielded more accurate results, with a relative error of 7%. Angles measured with the IMS compared to true values in the coronal and sagittal planes achieved R² values of 0.80 and 0.70, respectively. A special calibration rig, utilising a wooden dry tool axe and load cell, was developed for on-wall calibration. Final testing demonstrated that the modified screw could reliably measure force magnitudes above 30N with a mean relative error of <6% across the loading scenarios tested.

The loads from the IMS and two motion capture methods were also tested. The first motion capture method was a wearable IMU-based (inertial measurement unit) approach, and the second method utilised smart device cameras with a human pose estimation (HPE) algorithm. A study with 32 participants across three proficiency groups (beginner, intermediate, and advanced) was conducted. The study was used to identify distinguishing kinetic and kinematic parameters and provide a holistic evaluation of climber performance. Three proficiency groups were tested with equal numbers of male and female climbers. Climbers were given three attempts to climb each of three routes of increasing difficultly. Most beginner climbers were unable to complete the moderate and advanced climbs, which limited comparison of the proficiency groups to the easy climbing route. All experiments were conducted under ethical approval from the University of Canterbury Human Ethics Committee.

Results from the easy climbing route showed that the IMS could identify differences in climbers' proficiency through statistical analysis of performance parameters. Significant differences in contact time, Higuchi fractal dimension (HFD), and pulse count were observed on individual climbing holds (p<0.05), with experienced climbers exhibiting lower values indicative of their more fluent and efficient styles. Novel parameters such as power spectral density (PSD) and force rate of change (FRC) also revealed significant differences (p<0.05), with advanced climbers demonstrating higher FRC values and greater power at lower frequencies, while beginners exhibited more prominent higher frequencies, indicating increased force fluctuations. Additionally, a binary classifier and SHAP (SHapley Additive exPlanations) analysis for all holds’ identified mean normalised force magnitude as the most significant feature to distinguish beginners from advanced climbers. Beginners displayed lower mean normalised force values on both footholds and handholds, indicative of less dynamic climbing, which was confirmed with kinematic observations. SHAP analysis found that FRC and PSD were valuable in distinguishing between intermediate and advanced climbers, with advanced climbers showing higher FRC values on handholds and lower PSD values on footholds. Furthermore, the IMS also provided additional observations, identifying the crux of the climb through increased contact time, HFD, and pulse count, highlighting challenging holds. It also detected changes over three climbs, showing the ability to monitor the effect of familiarisation. The system also analysed dynamics between upper and lower body normalised forces, revealing advanced climbers' more dynamic approaches, confirmed by greater fluctuations of normalised total force.

A Bland-Altman analysis of the IMU and HPE kinematic measurements revealed limits of agreement (LoA) in displacement measurements of ±0.05m, ±0.06m, and ±0.10m for the mediolateral (ML), anteroposterior (AP), and vertical directions, respectively. The study also assessed whether both methods could distinguish proficiency groups based on climbers' centre of gravity (COG) parameters. Results indicated that average position values were more comparable between the methods (relative percentage difference RPD < 1%) than peak value comparisons (RPD > 10%). Average vertical and AP velocities were more consistent (RPD < 6%), whereas the ML direction showed greater differences (RPD > 12%). Similar RPDs were observed in average acceleration. These findings suggest that while both methods are suitable for separate motion analysis, their results are not identical, particularly for peak displacement and dynamic parameters.

The kinematic analysis of climbers across different proficiency levels identified several distinguishing parameters. Advanced climbers displayed greater COG displacement in the AP direction, higher velocities, and increased lateral oscillations as measured by mean absolute deviation (MAD) with p<0.05, indicating a more dynamic and efficient climbing style. Beginners, on the other hand, showed higher geometric index of entropy (GIE), time-normalised jerk coefficients, and immobility ratios, reflecting less fluent and more erratic movements with frequent stops. Advanced climbers also demonstrated greater power throughout the climb, making it a promising parameter for performance evaluation. Significant differences between intermediate and advanced groups were observed in the ML MAD parameter, with advanced females exhibiting greater lateral oscillations than the intermediate male group, suggesting a more conservative climbing style in the latter. Additionally, head orientation observations indicated that beginners had prolonged neck flexion from gazing at their feet rather than handholds, which reduced climbing efficiency.

The last objective was holistic observation of climbers’ performance. By employing Markov models to analyse combined kinematic and kinetic data, differences in climbing behaviours between proficiency groups were detected. The analysis compared normalised forces on footholds during mobility phases, normalised forces on handholds relative to COG lateral displacement, and total normalised forces versus AP displacement. Advanced climbers exhibited higher normalised forces on footholds during the traction phase (15% higher probability), while beginners showed increased foothold forces during the postural regulation phase (6% higher probability). Advanced climbers also demonstrated 4% higher probability to be in a state with greater lateral COG displacement, resulting in higher normalised forces on handholds, compared to beginners and intermediate climbers who had 4% higher probability to be in a state that relied more on their upper bodies when their COG was closer to the midline. Further examination of COG in the AP direction also revealed that advanced climbers had higher probability (9%) to be in a state with their COG positioned farther from the wall, contributing to greater total normalised forces in comparison to the state of COG closer to the wall. These findings emphasise the importance of a holistic evaluation, as forces exerted on holds are intricately linked to COG trajectory.

Overall, based on the results of this thesis, the instrumented mounting screw was found to be an effective tool for measuring climbing performance parameters. While it was less precise than other climbing hold instrumentation methods reported in the literature, this reduced precision was offset by its low cost and ease of installation on existing climbing walls, which could lead to widespread adoption. Although each climbing hold required calibration, this could be completed quickly. Hence, the proposed instrumentation will be suitable for use in climbing gyms for assessing recreational climbers and as a coaching tool.