Research at the Boundary of Robotics and AI: Challenge Problems - PowerPoint PPT Presentation

Research at the Boundary of Robotics and AI: Challenge Problems Prof: Peter Stone Department of Computer Science The University of Texas at Austin

Features of Challenges • Good performance metrics Peter Stone

Features of Challenges • Good performance metrics • Lead to data sets Peter Stone

Features of Challenges • Good performance metrics • Lead to data sets − Tension between data sets and embodiment/interactivity Peter Stone

Features of Challenges • Good performance metrics • Lead to data sets − Tension between data sets and embodiment/interactivity • Open world aspect, real-time aspect Peter Stone

Features of Challenges • Good performance metrics • Lead to data sets − Tension between data sets and embodiment/interactivity • Open world aspect, real-time aspect • Standard platform, software Peter Stone

Features of Challenges • Good performance metrics • Lead to data sets − Tension between data sets and embodiment/interactivity • Open world aspect, real-time aspect • Standard platform, software − Build on platforms and domains that are already working robustly Peter Stone

Features of Challenges • Good performance metrics • Lead to data sets − Tension between data sets and embodiment/interactivity • Open world aspect, real-time aspect • Standard platform, software − Build on platforms and domains that are already working robustly − Also can stick us in local minimum Peter Stone

Features of Challenges • Good performance metrics • Lead to data sets − Tension between data sets and embodiment/interactivity • Open world aspect, real-time aspect • Standard platform, software − Build on platforms and domains that are already working robustly − Also can stick us in local minimum • Something that people actually want done Peter Stone

Features of Challenges • Good performance metrics • Lead to data sets − Tension between data sets and embodiment/interactivity • Open world aspect, real-time aspect • Standard platform, software − Build on platforms and domains that are already working robustly − Also can stick us in local minimum • Something that people actually want done • Learning Peter Stone

Features of Challenges • Good performance metrics • Lead to data sets − Tension between data sets and embodiment/interactivity • Open world aspect, real-time aspect • Standard platform, software − Build on platforms and domains that are already working robustly − Also can stick us in local minimum • Something that people actually want done • Learning − Commonsense knowledge from humans and/or big data Peter Stone

Features of Challenges • Good performance metrics • Lead to data sets − Tension between data sets and embodiment/interactivity • Open world aspect, real-time aspect • Standard platform, software − Build on platforms and domains that are already working robustly − Also can stick us in local minimum • Something that people actually want done • Learning − Commonsense knowledge from humans and/or big data • Detect and adjust to human preferences Peter Stone

Features of Challenges • Good performance metrics • Lead to data sets − Tension between data sets and embodiment/interactivity • Open world aspect, real-time aspect • Standard platform, software − Build on platforms and domains that are already working robustly − Also can stick us in local minimum • Something that people actually want done • Learning − Commonsense knowledge from humans and/or big data • Detect and adjust to human preferences • Push state of the arts both in robotics and in AI Peter Stone

Challenges • Robot scavenger hunt (Amazing Race) Peter Stone

Challenges • Robot scavenger hunt (Amazing Race) • Human-robot interaction for assembly Peter Stone

Challenges • Robot scavenger hunt (Amazing Race) • Human-robot interaction for assembly − Intent/activity recognition Peter Stone

Challenges • Robot scavenger hunt (Amazing Race) • Human-robot interaction for assembly − Intent/activity recognition • RoboCup with 4 robots and 1 person Peter Stone

Challenges • Robot scavenger hunt (Amazing Race) • Human-robot interaction for assembly − Intent/activity recognition • RoboCup with 4 robots and 1 person • Adjustable autonomy wheelchair Peter Stone

Challenges • Robot scavenger hunt (Amazing Race) • Human-robot interaction for assembly − Intent/activity recognition • RoboCup with 4 robots and 1 person • Adjustable autonomy wheelchair • Device disassembly Peter Stone

Challenges • Robot scavenger hunt (Amazing Race) • Human-robot interaction for assembly − Intent/activity recognition • RoboCup with 4 robots and 1 person • Adjustable autonomy wheelchair • Device disassembly • Kitchen and household tasks Peter Stone