Effective robot navigation in real-world environments requires an understanding of terrain properties, as different terrain types impact factors such as speed, safety, and wear on the platform. Preference-based learning offers a compelling framework in which terrain costs can be inferred through simple trajectory queries to the user. However, existing query selection methods often suffer from redundant selection due to limited trajectory diversity, as well as query ambiguity, where the user must choose between trajectories with minimal distinguishable differences. These issues lead to inefficient learning and suboptimal terrain cost estimation. In this paper, we introduce a joint optimization framework that increases learning efficiency by improving both the diversity of the trajectory set and the query selection strategy. We used a variational autoencoder (VAE) to encode and group trajectories based on their terrain characteristics. Clusters were used to identify less represented terrain types so that new trajectories can be added to the corresponding cluster to ensure a balanced and representative query set. Additionally, we employ a cluster-aware query selection mechanism that prioritizes diverse trajectory pairs pulled from distinct clusters to maximize information gain. Experimental results demonstrate that our approach significantly reduces the number of queries required to converge to the ground-truth terrain cost assignment, outperforming state-of-the-art query selection techniques.

Robot designers and Human-Robot Interaction (HRI) practitioners can face challenges when people form a mental model of a robot that is not appropriate. Although the field of robotics would benefit significantly from a broad representation of designers, there is currently no comprehensive method of including many people in the design process and no theory of what expectations a robot design feature might elicit. We seek to address these challenges through the creation of a robot design platform, an online tool similar to a character creation interface in a video game, where users create a robot design. By collecting a large number of robot designs from users, we seek to be able to identify aspects of a robot’s design that influence the mental models humans ascribe to the robot. To maximize the universal usability of the platform, we conducted a three-part survey to assess which icons should be used to visually represent the mental states ascribed to the robots created by users on the platform. In our assessment, we found nine icons that met our criteria for use in the platform and others that should be further evaluated.

This work presents our strategy for investigating the fundamental guidelines and theories related to robot mind perception, and for establishing a metric for mental models, using our web-based tool, Build-A-Bot. We also discuss the effectiveness and efficiency of our platform by virtue of its inclusive design and its ability to visualize the user’s intended representation of a mental model for a robot through a 3D game-like interface. We conducted an observational user test study to assess if the website and the embedded robot building tool are effective and efficient to use for users. We found that the design of the robot creation platform and its associated website are considered intuitive and effective by a majority of our survey population. The Build-A-Bot platform successfully provides the ability for users to visualize their ideal representation of their mental model through an interactive game. Based on the obtained data, we propose further steps to optimize the Build-A-Bot platform for universal usability

Blogs are becoming an increasingly important medium -- socially, academically, and politically. Much research has involved analyzing blogs, but less work has considered how such analytic techniques might be incorporated into tools for blog readers. A new tool, metaViz, analyzes political blogs for potential conceptual metaphors and presents them to blog readers. This paper presents a study exploring the types of critical and creative thinking fostered by metaViz as evidenced by user comments and discussion on the system. These results indicate the effectiveness of various system features at fostering critical thinking and creativity, specifically in terms of deep, structural reasoning about metaphors and creatively extending existing metaphors. Furthermore, the results carry broader implications beyond blogs and politics about exploring alternate configurations between computation and human thought.