Citation Context

...imal distance between two agents. As in the work described here, their approach takes into account the multiple topological structures of the trajectories of the different agents. Trautman and Krause =-=[26]-=- point out that joint collision avoidance is crucial for mobile robot navigation to prevent the robot from “freezing” and getting stuck in densely populated environments. Their work assumes humans to ...

Citation Context

...e human crowds. We thus develop a cooperative navigation methodology and conduct the first extensive (nruns ≈ 500) field trial of robot navigation in natural human crowds. 1These authors are with the California Institute of Technology. 2This author is with NASA/JPL. 3This author is with ETH Zurich. Fig. 1. Overhead still of the crowded university cafeteria testbed. The density of the crowd varies through the day, allowing for diverse experiments. (Figure 1). These experiments quantify the degree to which our cooperation model improves navigation performance; with the theoretical arguments of ([10]), we deduce the importance of a cooperation model for safe and efficient crowd navigation. A. Related Work Naively modeling the uncertainty in dynamic environments (e.g., with independent agent constant velocity Kalman filters) leads to an uncertainty explosion that makes safe and efficient navigation impossible ([10]). Some research has thus focused on controlling predictive uncertainty: in [11] and [12], high fidelity independent human motion models were developed, in the hope that reducing the uncertainty would lead to improved navigation performance. Similarly, [13] holds the individual a...

Citation Context

... equilibrium theory [5], have also been evaluated with robots [6, 7]. Contemporary probabilistic modeling techniques have been applied to person-aware robot navigation in dynamic crowded environments =-=[8]-=-, to goal-oriented navigation behavior with a human partner [9], and to calculating a robot approach trajectory to initiate interaction with a walking person [10]. Copyright is held by the author/owne...

Citation Context

... influence obstacle behavior is also not appropriate in certain settings like navigating amongst pedestrians at intersections. Some recent work has applied learning to pedestrian navigation scenarios =-=[15]-=- but incorporating vehicle dynamics is still an open problem. Finally, we plan to use a commercial driving simulator to observe how human drivers perform and respond to assisted driving using our syst...

Citation Context

...ut lacks our closed-loop simulation of vehicle interactions. Recent work uses Gaussian process (GP) regression to learn typical motion patterns for classification and prediction of agent trajectories =-=[24, 25, 40]-=-, particularly in autonomous driving [1, 38, 39]. Nonetheless, these methods require collecting training data to reflect all possible motion patterns the system may encounter, which can be time consum...

Citation Context

...iration from human social conventions for moving in crowded pedestrian spaces. pursuit, leading to reciprocal algorithms in which an agent can expect its co-agents to cooperate in avoiding each other =-=[13, 23, 26]-=-. Trautman and Krause [26] recognize that if agent trajectory prediction is done for the purpose of robot navigation, then it is vital to perform joint collision avoidance by incorporating the robot’s...

Citation Context

...les in the map, to decide a future path for the robot. In dynamic environments, these algorithms often yield trajectories which are highly suboptimal and often require the robot to stop unnecessarily =-=[27]-=-. We anticipate the current application will require path planning approaches for dynamic environments that anticipate the trajectories and goals of obstacles [28]–[30]. Lastly, the robot will not onl...

Citation Context

...have been evaluated with sociable robots [5, 6]. Contemporary probabilistic modeling techniques have been applied to socially appropriate person-aware robot navigation in dynamic crowded environments =-=[7]-=-, to calculating a robot approach trajectory to initiate interaction with a walking person [8], and to the recognition of averse and non-averse reactions of children with autism to a socially assistiv...

Citation Context

...o account the behavior of the pedestrians in the vicinity of the robot. In particular, our approach implicitly learns how pedestrians typically react when interacting with robots. Trautman and Krause =-=[18]-=- demonstrate that mobile robot navigation fails in densely populated environments unless the robot takes into account the interaction between the robot and the humans. Van den Berg et al. [19] present...

Citation Context

... HRI [25, 26] and immersive virtual social environments [27]. Contemporary machine learning techniques have been applied to sociallyappropriate person-aware navigation in dynamic crowded environments =-=[28]-=- and recognition of positive and negative attitudes of children with autism to an interactive robot [29]. A lack of high-resolution metrics limited previous efforts to coarse analyses in both space an...