Individual Localization and Tracking in Multi-robot Settings with Dynamic Landmarks (Extended Abstract)
Anousha Mesbah, Prashant Doshi · 2011
Our focus is on the subject robot’s localization at its own level in the presence of others who may not be cooperative. Consequently, our perspective and approach differs from previous work in multi-robot settings, which has predominantly focused on joint localization by multiple cooperating robots [3,4]. In this context, we introduce a nested set of particles to track the subject robot and others, and recursively project these particles as the subject robot moves and makes observations. Consequently, the subject robot attributes a behavioral model to the other in order to predict its actions. Extending Rosencratz et al.’s laser tag approach [5] for our experimentation, we generalize the problem by assuming that the subject robot is itself not localized. Motivated by the challenges faced in search and rescue, we require the subject robot to tag the other and then seek to reach the opponent’s base. On being tagged, the opponent robot may move the nearest landmarks in order to confound the subject’s localization. This is analogous to independent rescue robots moving obstacles while searching for victims. This has the effect of delaying the robot’s approach toward the base. We adopt the perspective of a robot i whose task is to tag another robot j and then proceed to reach j’s base within a certain amount of time steps. As we focus on localization in this paper, we assume that robots i and j know the exact locations of the landmarks. However, i is unaware of its own and j’s location in the environment. Observation of the landmarks could be used by robot i for localizing itself in the environment. Unfortunately, the presence of multiple objects with the same color means that the localizing information is often ambigious. Each robot is equipped with a standard laser range sensor, a camera and a bump sensor. Tagging is accomplished by identifying robot j when it is in close proximity of i by fusing readings from the laser range and camera sensors. Robot j utilizes a mixed strategy behavioral model that is in play until j is tagged. The behavioral model takes as input the hypothesized pose of a robot