You are on page 1of 29

Distributed Robotic Teams

Nikos Papanikolopoulos Professor, Department of Computer Science and Engineering Digital Technology Center University of Minnesota npapas@cs.umn.edu

AI, Robotics, and Computer Vision Laboratory, CSE, UMN

Faculty
Maria Gini (Artificial Intelligence, Robotics, Intelligent Agents) Nikos Papanikolopoulos (Robotics, Computer Vision, Transportation Systems) Stergios Roumeliotis (Robotics, Robot Localization and Mapping) Rich Voyles (Robotics, Computer Vision, MultiAgents) Paul Schrater (Computer Vision, Pattern Recognition, Haptic Interfaces)

AI, Robotics, and Computer Vision Laboratory, CSE, UMN

Mission Statement
The goal of the laboratory is to combine algorithm development with strong system building efforts that validate our algorithms. With respect to teaching, the goal is to educate top researchers/teachers and use AI, robotics, and computer vision as the vehicle to promote science/engineering principles to our students.

Rangers and Scouts


Heterogeneous robotic team Rangers
General purpose off-road robots

Scouts
Custom portable sensor platforms

Technical Overview of the Project: Goals

Develop distributed robots (main unit is roughly the size of a soda can) with various mobility and sensory modes for exploration of structures in an urban warfare scenario. Emphasize the development of innovative mobility modes for these robots. Data fusion and analysis are to be at a remote operator station. Effective combination of software and hardware design.

Some Applications

Special Unit Operations Law Enforcement Agencies:


Hostage situations Counter terrorist units in large cities Fire and rescue Disaster relief

Coast Guard: vessel boarding and search FBI Fugitive Task Forces, FBI Crisis Response Team NASA space exploration programs Toy and Entertainment industries

Scout Designs

Grasshopper

Rolling Ball Scout 2001

Scout 2001 vs. Scout 1999

Complete electronic and electrical redesign Significantly improved power management Encoders have been added to the drive motors. Scout 2001

The internal components have all been modularized to decrease assembly and repair time from hours to minutes. The jumping spring and winching motor have been improved to increase overall mechanical efficiency and the height of jumps.

Scout 1999

Scout Jumping Trial

(a)

(b)

(c)

(d)

(e)

(f)

Scout Videoclip

Hardware Efforts

Mobility improvements
Variable-size actuated scout wheels Rooftop scout with grappling hook COTSScout and MegaScout

Communication improvements
Scout repeaters

Variable-Size Actuated Scout Wheels

wheel extended

wheel retracted

Rooftop Scout

User feedback indicated the need for a scout that can operate on a roof or a ceiling. The goal again is to achieve the desirable functionality without significant changes to the scout design. The idea was to have an external scout attachment with a spring-loaded mechanism and a grappling hook. Two prototypes have been built.

Rooftop Scout

Grappling Hook Mechanism

Scout with the Pan-Tilt Camera System

Motion Detection and Tracking


A computer vision algorithm was written to detect motion in the field of view and automatically track the center of motion.

Low Cost Alternatives

Commercial OffThe-Shelf (COTS) Scout


Designed to be a low cost, nearly disposable alternative with reduced functionality and slightly improved resilience Physical Characteristics: 44 mm x 97 mm cylinder with 3 wheel sizes (57 mm, 64 mm, 76 mm) Sensing Capability: Camera Locomotion: Rolling: Top speed
57 mm wheel: 0.39 m/s 64 mm wheel: 0.42 m/s 76 mm wheel: 0.50 m/s

Operational Life: 190 minutes in quiescent mode, 60 minutes at top speed broadcasting video Cost: < $300

Low Cost Alternatives

Commercial OffThe-Shelf (COTS) Scout

Modified Scout Platforms


MegaScout

Control and Simulation


Control architecture is layered and based on behaviors. Rangers are the control and communication central units (utility units). Lego-based platforms were used initially for rapid prototyping and experimentation.

Software Implementation
Prioritized Behaviors, Mission Decomposition, Parallel Execution

Backbone Mission Control Resource Pool User Interface

CORBA XML Scouts, Rangers, Radios, C++ Video Frequencies,

Framegrabbers, Launchers, etc. Supervisory Workstation Palm Pilot/Wristwatch TVs

Scheduling, System Configuration, Starting/Stopping Components, Load Balancing, etc.

Behavior: Servo to a Target

Human selects target Robot handles servoing process Assumption: known target geometry Deployment and retrieval of robot Navigation between waypoints

Examples of Vision-Based Behaviors (Image Mosaics)

The scout rotates 360 deg and blends the individual images in order to create mosaics of urban spaces.

General Innovations

Design and functionality of the scouts. Design of the ranger-scout system (launching, communication, navigation). Simple control and communication primitives that can be reconfigured for a variety of robot behaviors. Development of a large heterogeneous distributed robotic system based on these scalable, reconfigurable behaviors and physical components. Miniaturization of the scout hardware peripherals (e.g. sensing, communication, etc.) required significant innovation. Creation of a large variety of mobility modes in order to increase the scout functionality.

Technical Challenges

Communication!!!!! Power requirements and management* Design of the jumping mechanism Hardening* Assembly and disassembly of the scout Miniaturization* Human-scout interaction given a large number of scouts Software issues

*High risk issues

Conclusions

The scout has proven to be a simple and effective design. It has been featured on CNN, BBC, NBC, Comedy Central, and was a finalist for the Discover award. Feedback from military users (Lincolnia exercises) indicates that the scout is the right design and the right size. There is a patent for the overall design. Various innovative mobility enhancements have been developed. The COTSScout has a long life and a satisfactory set of capabilities at a very low-cost. A variety of user interfaces has been developed. Communication remains a challenge.

Acknowledgements
This material is based upon work supported by National Science Foundation through grant #EIA0224363, Microsoft Inc., INEEL, and the Defense Advanced Research Projects Agency, MTO (Distributed Robotics Program), ARPA Order No. G155, Program Code No. 8H20, issued by DARPA/CMD under Contract #MDA97298-C-0008

URL Information
Videoclips and papers can be found at: http://distrob.cs.umn.edu

Scout Coverage by BBC