Statistics

We use box-plots to represent the performance of the control software we produce. For each method, we report the performance obtained in simulation (thin boxes) and with physical robots (thick ones). In all cases, we support comparative statements with an exact binomial test, at 95% confidence [66]: statements like "*A* performs *significantly* better/worse than *B*" imply that the comparison is supported by a an exact binomial test, at 95% confidence. In addition, we estimate the overall performance of TuttiFrutti with respect to EvoColor. To this purpose, we aggregate the results by comparing the performance of the two design methods across each mission. In the context of the overall performance of the design methods, any statement like "*A* performs *significantly* better/worse than *B*" also implies that the comparison is supported by an exact binomial test, at 95% confidence [66].
