Title: TAR: Trajectory Adaptation for Recognition of Robot Tasks to Improve Teamwork

Michael Novitzky

School of Interactive Computing

Georgia Institute of Technology

Date: Thursday July 16

Time: 10AM

Location: TBA

Committee:

Dr. Tucker R. Balch, School of Interactive Computing, Georgia Tech (Advisor) Dr. Henrik Christensen, School of Interactive Computing, Georgia Tech Dr. Thomas R. Collins, School of Electrical and Computer Engineering, Georgia Tech Dr. Magnus Egerstedt, School of Electrical and Computer Engineering, Georgia Tech Dr. James M. Rehg, School of Interactive Computing, Georgia Tech Dr. Lora Weiss, Georgia Tech Research Institute

Abstract

One key to more effective cooperative interaction in a multi-robot team is the ability to understand the behavior and intent of other robots. Observed teammate action sequences can be learned to perform behavior recognition which can be used to determine their current task. Previously, we have applied behavior histograms, hidden Markov models (HMMs), and conditional random fields (CRFs) to perform behavior recognition as an approach to task monitoring in the absence of communication. To demonstrate behavior recognition of various autonomous vehicles, we used trajectory-based techniques for model generation and behavior discrimination in experiments using actual data. Beyond recognition of behaviors, we additionally introduced strategies, based on the honeybee’s waggle dance, in which cooperating autonomous teammates could leverage recognition during periods of communication loss. While the recognition methods were able to discriminate between the standard behaviors performed in a typical survey mission, there were inaccuracies and delays in identifying new behaviors after a transition had occurred. Inaccuracies in recognition lead to inefficiencies as cooperating teammates acted on incorrect data. We then introduce the Trajectory Adaptation for Recognition (TAR) framework which seeks to directly address difficulties in recognizing the behaviors of autonomous vehicles by modifying the trajectories the robots follow to perform them. Optimization techniques are used to modify the trajectories to increase the accuracy of recognition while also maintaining task objectives while adhering to vehicle dynamics. Experiments are performed which demonstrate that using trajectories optimized in this manner lead to improved recognition accuracy.