The process of creating the phonology is described at length in Omar Mubin's PhD thesis "ROILA: RObot Interaction LAnguage" (available from the publications page). It seems to have happened in three stages.
He determined which phonemes are most common in a number of natural and artificial languages (pp. 17-19). The most common vowels were /i, u, o, a, e/ (in IPA).
The final set of phonemes he chose included the vowels a, e, i, o and u. He doesn't indicate that the vowels were given in IPA, and indeed he later gives them the IPA values /ae (sic), ɛ, ɪ, ɑ, ʌ/ (pp. 22-23).
AA /ɑ/ was later altered to AO /ɔ/ in order to avoid confusion with AE /æ/ (p. 35-36).
The only explanation I can give for the incongruity between stage 1 and stage 2 is that he (mistakenly?) put the five vowels /a, e, i, o, u/ into his orthography, but changed their phonetic values to the English orthographic "short" vowels, which are /æ, ɛ, ɪ, ɑ, ʌ/ in American English.
The fact that he had to avoid confusion between /æ/ and /ɑ/ (in step 3), but doesn't consider any confusion between /æ/ and /ɛ/ or between /ɔ/ and /ʌ/ seems very English-centric, but which would be very strange from the perspective of Dutch, which doesn't distinguish either pair of phonemes (considering the fact he was doing his research in Holland).
Unsurprisingly, "American English speakers significantly outperformed other speakers in our setup" (p. 34). They also mention problems with Dutch accents and different regional English accents (p. 86 of this document).