8.4.1. Evaluation Criterion

To evaluate the system performance, 40 key points were set along the experimented path, which served as the ground truth for comparison. During the experiment, the participant was asked to stop a little at each key point for the purpose of recording the markers. Let *xie*, *yie* and *xig*, *yig* be the estimated location of an unknown target, and the ground truth location at the *i*th key point, respectively; then the error at *i*th key point is

$$
\varepsilon\_{\mathcal{S}^c}^i = \sqrt{\left(\mathbf{x}\_{\mathcal{S}}^i - \mathbf{x}\_c^i\right)^2 - \left(y\_{\mathcal{S}}^i - y\_c^i\right)^2} \tag{26}
$$

This study reports the mean error, median error, standard deviation, 25th, 75th, and 90th percentile errors, and a cumulative distribution function (CDF) plot to analyse the localisation performance of the proposed system.
