Arkin abridged

Category: Entertainment

Presentation Description

No description available.


Presentation Transcript

Situated Multi-robot Systems: A Team is More than Just Robots : 

Situated Multi-robot Systems: A Team is More than Just Robots Ronald C. Arkin Mobile Robot Laboratory Georgia Tech

What are some needs for real Cooperative Control?: 

What are some needs for real Cooperative Control? Thinking AND Acting: GOFAI revisited 1 more chance for Deliberation/Cognition System-level Cooperative Control RAPs (Robot-Agent-People Teams) Human-Robot-Interaction Robustness/Reliability/Predictability

One Take-home Lesson: 

One Take-home Lesson Cooperative Control includes people as well as robots: Understanding Human-robot interaction is essential for success

A Bit of our Multirobot History: 

A Bit of our Multirobot History

Multiagent Communication: 

Multiagent Communication Tasks Forage (retrieval) Consume (mine-clearing) Graze (reconnaissance) Information Content No communication State communication Goal Communication None State Goal

Formation Control: 

Formation Control Formation Types Based on Military Protocols for Scouts Line Column Wedge Diamond

Formation References: 

Formation References Unit-center: average position of all robots Leader: displacement relative to designated robot Neighbor: displacement relative to neighboring robot Unit-center Leader Neighbor

Formation Control: 

DARPA UGV Demo C Ported to Lockheed-Martin HMMWVs Demonstrated Summer 1995 before live military audience Column -> Wedge -> Line -> Column Formation Control

Team Teleautonomy: 

Team Teleautonomy MOTIVATION To reduce cognitive overload on operators To permit behavioral reconfiguration with limited operator knowledge

Team Teleautonomy: 

Team Teleautonomy Trapped in box canyon Helped out of trap

MARS Learning Techniques: 

MARS Learning Techniques CBR Wizardry Guide the operator Probabilistic Planning Manage complexity for the operator RL for Behavioral Assemblage Selection Learn what works for the robot CBR for Behavior Transitions Adapt to situations the robot can recognize Learning Momentum Vary robot parameters in real time THE LEARNING CONTINUUM: Deliberative (premission) . . . Behavioral switching . . . Reactive (online adaptation) . . .

Learning Momentum Multirobot Scenario: 

Learning Momentum Multirobot Scenario New scenario was invented involving soldiers protecting a target object in the presence of 2 types of enemies Combines MoveToTarget, AvoidSoldiers, and InterceptEnemy assemblage outputs via vector summation LM alters behavioral gains based on the current situation Intercept enemies Avoid Robots

Intelligent Role Switching: 

Intelligent Role Switching A Q-learner on each robot is used to switch between three well-tested roles Forager Searches for and collects landmines In an obstacle and hazard free environment, a single forager can easily clear the map. Once a patch of landmines are detected, places a marker so other robots can move to the same area. Soldier Uses the intercept, stop, terminate policy learned by previous scenario. 3. Mechanic When a dead robot is detected, it moves to and fixes the damaged robot.

DARPA/NRL MARS 2020: Adaptive Autonomous Robot Teams for Situational Awareness: 

DARPA/NRL MARS 2020: Adaptive Autonomous Robot Teams for Situational Awareness Provide communication-sensitive planning and behavioral control algorithms in support of network-centric warfare, that employ valid communications models provided by BBN Provide an integrated mission specification system (MissionLab) spanning heterogeneous teams of UAVs and UGVs Demonstrate warfighter-oriented tools in three contexts: simulation, laboratory robots, and in the field Joint work with UPenn, USC, and BBN

\ End Digression \: 

\ End Digression \

Give GOFAI a chance! (Why you ask? Top 5 reasons): 

Give GOFAI a chance! (Why you ask? Top 5 reasons) 5. OK, it’s a second chance…. 4. Much more mature underpinnings in agents/robotics this time around 3. Put up or shut up time 2. DARPA IPTO Thrust raises opportunities 1. Minsky’s upset! “The worst fad has been these stupid little robots. Graduate students are wasting 3 years of their lives soldering and repairing robots, instead of making them smart. It’s really shocking.” (Wired 5/13/03)

RAPS – Teams of Robots/Agents/People: 

RAPS – Teams of Robots/Agents/People Truly Heterogeneous teams Right RAP for the Job - Planning Dynamic Reassignment - Reacting Managing communication over varying timescales Dealing with limitations of each RAP type

Additional Perspectives on Human-robot Interaction:: 

Additional Perspectives on Human-robot Interaction: DESIGN: Usability Issues for Designing Multi-Robot Missions MissionLab Mission Specification Software INTERACTION: Software Architecture Designed for Human Interaction Sony AIBO and SDR

Perspective 1: Usability Issues for Designing Multi-Robot Missions  : 

Perspective 1: Usability Issues for Designing Multi-Robot Missions   Will robots be able to complete useful and meaningful tasks in conjunction with people? Will the results being produced in academic laboratories have impact in real-world everyday robotics? Will end-users of these systems be required to have a Ph.D. in robotics in order for them to be of any value?

Usability Objectives: 

Usability Objectives To provide effective methodologies that evaluate the performance of multiagent robotics systems from an end-users perspective. To provide methods and tools in support of cognitive modeling of the interaction of users with multiagent robotic systems. To create meaningful applications for robotic teams, that can serve as prototypical tasks for the research community.

MissionLab – Mission Specification System: 

MissionLab – Mission Specification System Objective: To empower users to specify, evaluate, and execute complex robot team missions


MissionLab Problem Statement Constructing robot control configurations is ad hoc and tedious Configurations are difficult to retarget for new vehicles Component reuse is difficult yet needed at all levels of abstraction Support is needed for evaluation of multirobot configurations


MissionLab Example: Scout Mission




MissionLab Example: Trashbot (AAAI Robot Competition)


MissionLab Reconnaissance Mission Developed by University of Texas at Arlington using MissionLab as part of UGV Demo II Coordinated sensor pointing across formations

GT Robotics Usability Studies: 

GT Robotics Usability Studies Mission Specification for Robot Teams DARPA Demo II -Mission Specification (MacKenzie Dissertation) TMR - Tactical scenarios for Urban Warfare MARS - CBR Wizard for user assistance Team Teleautonomy Run-time Interface design (Ali Dissertation) Human Factors Study – TMR Team of 4 robots

Perspective 2: Interaction Software Architecture Designed for Human-Robot Interaction: 

Perspective 2: Interaction Software Architecture Designed for Human-Robot Interaction Joint Work with M. Fujita, T. Takagi and R. Hasegawa, Sony Digital Creatures Lab, Tokyo Goals: Incorporation of high-fidelity ethological models of behavior to allow human to relate predictably to a robotic artifact Generation of motivational behavior to support existing conceptions of living creatures to encourage bonding between the human and artifact

Sony Entertainment Robots: 

Sony Entertainment Robots

Animals -> Robots: 

Animals -> Robots

Dog Behavior  Ethological Controller: 

Dog Behavior  Ethological Controller Ethologically Inspired Design Starting Point: Scott/Fuller ethogram Timberlake-Lucas systems modeling approach

12 Behavioral Subsystems for Dog Model: 

12 Behavioral Subsystems for Dog Model Investigative Epimeletic Et-epimeletic Allelomimetic Agonistic Sexual Eliminative Ingestive Comfort-seeking Miscellaneous Play Maladaptive

Entire Ethological Behavioral Tree Design: 

Entire Ethological Behavioral Tree Design

Emotional Space: 

Emotional Space Homeostasis Regulation Rule: Internal variables (hunger, thirst, etc.) are regulated to remain within certain acceptable ranges Takanishi’s model: Pleasantness, Arousal, and Confidence define emotional space Pleasantness maximized when homeostatic variables are within acceptable range Arousal axis controlled by circadian rhythm and unexpected stimuli Confidence determined by certainty in perception

SDR Humanoid Preliminary Design: 

SDR Humanoid Preliminary Design

SDR Emotional Expression: 

SDR Emotional Expression Associates different emotional state with different people (Attitudes)

Model Framework (TAME): Traits, Attitudes, Moods and Emotions (w/ L. Moshkina): 

Model Framework (TAME): Traits, Attitudes, Moods and Emotions (w/ L. Moshkina) Environment

How TAME fits in: 

How TAME fits in Composed of four interrelated components: Personality Traits, Attitudes, Moods, and Emotions Emotions and moods constitute dynamically changing robot’s affective state Traits and attitudes are more or less time-invariant, and define general dispositions The environment is continuously scanned for relevant cues The module modifies behavioral parameters, which affect currently active behaviors

Psychological Foundations: 

Psychological Foundations Each serves a distinct adaptive function: Traits serve as an adaptation mechanism to specialized tasks and environments Emotions provide a fast response to significant environmental stimuli Moods bias behavior according to environmental conditions Attitudes facilitate decision-making process by reducing decision space

Exploratory Experimental Study: 

Exploratory Experimental Study Overall goal: to identify aspects of effective human-robot collaboration In particular, specific affective phenomena to include into the framework Two-part study: Longitudinal study – at least 5 thirty-minute sessions to allow the subjects to bond with the robot One-time short study – to assess participants response to an affective robot


Summary Coordinated Control, yes, but what is being coordinated? Coordination includes greater system issues: humans, softbots, etc., not just robots HRI, Human-robot teams, and related design issues should be considered from the beginning

For further information . . .: 

For further information . . . Mobile Robot Laboratory Web site Contact information Ron Arkin:

authorStream Live Help