Waist rotation angle as an indicator of probable human collision avoidance direction for autonomous mobile robots

Currently, the probability of pedestrians on the streets of smart cities encountering autonomous mobile robots (AMRs) is increasing. Previous studies have discussed collision avoidance between humans as they cross paths. Their avoidance behavior toward AMRs, however, remains unclear. To address this, we experimentally investigated the avoidance direction a human would choose against an AMR approaching head-on. This experiment included recording human locomotion under such a scenario. Further, an AMR was programmed to approach from various starting points, including directly through the participants (head-on collision). The participants were directed to evade it by moving rightward or leftward, and their paths were tracked. We found that the participants did not strongly prefer either direction, suggesting that the avoidance direction is not solely determined by the participants’ attributes, such as their adherence to the traffic rules of their region. The probability of rightward evasion when the AMR approached head-on indicated that humans use different avoidance strategies when encountering other humans and obstacles. Moreover, the participants’ motion analysis revealed that they involuntarily twisted their waists in the avoidance direction before they evaded the AMR. These results suggest that this twist is the most important predictor of the avoidance direction. These findings could be encoded into the programs of AMRs to adapt these vehicles to our locomotory responses more organically.