Can this robot pass through that aperture? Learning to make judgments for a tele-operated robot

Urban Search and Rescue is exceptionally dangerous. To mitigate that danger rescue workers utilize robots to access areas that are too small or too dangerous to send people. However, these robots must be tele-operated, and tele-operation contributes to a host of perceptual and cognitive problems (Chellali, 2009; Chen, Durlach, Sloan, & Bowens, 2005; Cui, Tosunoglu, Roberts, Moore, & Repperger, 2003; Darken, Kempster, & Peterson, 2001; Lichiardopol, 2007; Van Erp, 2000; Vertut & Coiffet, 1985). One such perceptual problem is the judgment of pass-ability, or the width of the robot relative to the width of the aperture. Previous research indicates that operators are good at making judgments of pass-ability (Jones, Johnson, & Schmidlin, 2011; Moore & Pagano, 2006), but that these judgments may not improve with practice. The present study investigated two questions: (1) Do judgments of passability improve with time, and (2), What sources of information in the environment contribute to learning to make better judgments of pass-ability in a tele-operated environment. The results indicated the following. First, judgments of pass-ability do in fact improve with time. However, the detection of that improvement requires the data be examined as percent correct rather than the more traditional threshold analysis. Second, participants learn to make better judgments when they are allowed to explore the environment and receive feedback, when they are only allowed to explore and not receive feedback, and when they are not allowed to explore and instead only receive feedback following a judgment of pass-ability. However, the performance for participants who were allowed to explore was significantly better than that of participants who were not allowed to explore. These results support the importance of exploration when making an affordance judgment. So while the data support that learning can happen without exploration, they also support that the performance with exploration is significantly better.