Single-session estimation of cardiorespiratory fitness, lower-limb strength, and core strength using a wearable hip-joint exoskeleton

Laboratory criterion tests for cardiorespiratory fitness and muscular strength are valid but require specialised equipment and trained personnel, limiting scalability in real-world settings. Wearable hip exoskeletons already deployed for exercise and rehabilitation can standardise short bouts while capturing kinetic and physiological signals, enabling compact “smart tests” for multi-domain fitness estimation without additional laboratory infrastructure. In healthy adults ( \(n = 42\) ), a single-session knee-up protocol with a hip-joint exoskeleton yielded heart-rate–derived and kinetic features to estimate (i) cardiorespiratory endurance (estimated VO \(_{2}\) max), (ii) lower-limb strength, and (iii) core strength. We prioritised simple multivariable linear regression and used polynomial+LASSO only as a comparator. Targets were validated by (i) benchmarking treadmill-estimated VO 2 max against directly measured VO 2 max in a feasibility subset ( \(n = 30\) ) and (ii) correlating isometric abdominal crunch with plank hold time. Agreement was examined via Bland–Altman analysis. We also introduced absolute percentile-weighted and relative Pareto-based composite scores. Linear models explained 62–88% of the variance across the three fitness targets within the development sample (VO 2 max \(R^2\) up to \({\sim }0.88\) ), with in-sample Bland–Altman analyses showing negligible mean biases (VO 2 max: 0.00 ml \(\cdot\) kg -1 \(\cdot\) min -1 ; lower-limb: 0.00 Load/BW; core: 0.00 Load/BW). Polynomial+LASSO did not provide consistent improvements over linear regression. Target-level checks supported convergent validity (treadmill-estimated VO 2 max vs. directly measured VO 2 max: \(r = 0.92\) –0.94) and core capacity (crunch vs. plank: \(r \approx 0.80\) , \(\rho \approx 0.78\) ). Pairwise correlations among the three targets were strong ( \(r = 0.78\) –0.82). Bootstrap resampling ( \(B = 1000\) ) and Gaussian noise perturbation provided supplementary evidence regarding model stability within the development sample. The absolute composite summarised standing within sex–age peers, while the Pareto-based ranking highlighted balanced profiles. These proof-of-concept findings show that a brief, exoskeleton-guided single-session protocol can provide promising, internally consistent estimates of multi-domain fitness using simple linear regression, supporting interpretable, low-overhead deployment as a structured monitoring tool within existing exoskeleton-assisted sessions pending independent external validation.