Towards a Robotic Ecology Briefing September 23, 1999 Rodney Brooks Pottie (MIT) (UCLA) Greg Robot Ecologies Where we are: Single robot that has as its intellectual metaphor a lone animal that perhaps can interact with people. Where we are going now: Swarms of identical robots based on social insect metaphors, perhaps with augmented communication. Where we want to go:
ISAT Self deploying, and self sustaining ecologies of plant-like robots and animal-like robots that symbiotically interact across many species, in order to carry out complex missions without logistical support. DARPA The Robot Ecologists COMMITTEE ITINERANTS Rod Brooks, ISAT Greg Pottie, UCLA Dick Urban, DARPA Elana Ethridge, SPC Polly Pook, IS Robotics Sarita Thakoor, JPL David Gerrold, writer Russ Frew, ISAT Al McLaughlin, ISAT Chuck Taylor, UCLA Maja Mataric, USC ISAT
DARPA GUEST PRESENTERS Brian Wilcox, JPL Paul MacCready, AeroVironment Doug Stetson, JPL, Helen Greiner, IS Robotics, Ian Waitz, MIT Dave Shaver, Lincoln Lab Steve LaFontaine, MIT Steve Leeb, MIT Erik Syvrud, OST John Blitch, DARPA Mark Swinson, DARPA Bob Nowak, DARPA Keith Holcomb, Marines (ret) Warfare in an Asymmetrical Situation The game is changing--we must change our response. ENGAGEMENT
SURVEILLANCE Stay outside of detection circle depends on cross section (self) Within circle want to: sense what is happening maintain long term presence tag things and infiltrate surgically and outfiltrate(!) maintain covertness detection/lethality circle ISAT robots DARPA Stay outside of lethality circle depends on weapons (of opponent) Want numerical advantage
Within circle want to: sense what is happening provide targeting information disrupt the opponents cohesion and will Logistics chain people Why Using Robots Is Hard, Yet Good ENGAGEMENT SURVEILLANCE Need covert deployment Need occasional mobility Need long term operation energy supply logistics possibly resupply (bio sensors) Need covert information return
Robots can move Robots can be very small Robots can carry variety of sensors Robots wait patiently We know where you are ISAT DARPA Need rapid deployment Need rapid mobility Need logistics chain Need reliable, rapid information processing and transmission Need active responses Robots can move
Robots are expendable Robots can carry a variety of sensors Robots can provide many andviewpoints what you are doing. Solution: The Robot Ecology Build an ecology of animal- and plant-like robots Go beyond the idea of single mobile robots Develop the collective as a super-organism where no single part understands the whole The Robot Ecology is a self-constructing infrastructure supports diverse individual tasks and enables more complex missions handles system degradation gracefully is self-sustaining throughout mission life ISAT DARPA How The Components Combine
seed sensors stationary sensor ISAT mother plant DARPA caterpillar (mobile sensor) What new capabilities? Precondition the battlefield for timely and precise targeting of enemy assets Know the environment scout, search, collect, penetrate, filter, report Tag enemy assets reduce fog; trace and target Weaken enemy infrastructure disrupt, confuse, attack cohesion and will
Deploy friendly infrastructure communication, navigation, supplies, weapons High-quality low-cost real-time intelligence available to small tactical units ISAT DARPA Symbiosis Between People and Robots The robot ecology needs to intermesh with the human organization in a symbiotic relationship People are better at some things Robots are better at some things Robots will be the remote extension of people Robots must support people rather than force people to support robots People are freed to make the higher level judgements in command without having to control The currencies of the self-sustaining robot ecology are energy and information
they trade against each other and between themselves they need to be supplied at the right places and times ISAT DARPA 0 Application Scenarios Remote exploration Tagging of people/trucks/ships/submarines Self-deploying communications/power network Search and rescue Battlefield surveillance, mine countermeasures Response to bio/chem attack Monitoring (infesting) a building Monitoring remote site for underground facilities (UGF) Support for military operations in urban terrain (MOUT) ISAT DARPA 1
UGF Threats: missile sites, weapons factories (e.g. biochem), command facilities, storage, weapons research What needs to be done: covertly characterize the facility (activity and structure) and possibly disrupt it Task List: monitor input/output of facility (roads, vents, effluent), sense nearby, sense inside, guide weapons, disrupt facility Steps: locate, infiltrate/disrupt, infestation, gather information; establish logistical chain for communication, sample retrieval and/or facility disruption ISAT DARPA 2 Underground Facility Characterization (maybe satellite detect)
UAV follows; releases microflyers, seeds pods, creepers, burrs, mobile burrowing device from mother plant down to buried targets ISAT communication relay to hill DARPA creeper down air vent; burr placed inside;
set up sensor net (vibrations, gases, etc.) [not to scale] 3 MOUT Threats: snipers, suicide bombers, biohazards, traps/mines; complication of neutrals as shields, chaos and confusion What needs to be done: avoid entering circle of lethality while establishing order and control Task List: navigation, communication, clearing, securing cleared areas, security in crowded/cluttered areas Steps: long-range deployment (e.g. to rooftops), local selfdeployment, sense assess and reposition cycle, weapons use; diversity and numbers to overcome countermeasures ISAT DARPA 4 Military Operations in Urban Terrain
Microflyers harvest biosamples Sensors defend secured areas Camouflaged devices for tracking, scanning, extracting bio-samples Creeper/ climbers gather indoor /outdoor info; form comm relay ISAT Robo-insects gain access inside doors/windows, around corners, DARPA not to scale
5 There are some key systems challenges Scaling 10s (now) to 100s and 1000s Heterogeneity Symbiotic relationships of plantbots, mobots, and people Adaptivity Context-aware self-organizing systems Some holes in base technology research areas Mobility Self-configuring networks Sensors Energy sources Cooperative behavior
ISAT DARPA System issues supported by technologies Why Cant We Just Do This Today? 6 NA 1 1 1 0 1 2
Self-configuring 1 networks Sensors 2 Energy sources Cooperative behavior ISAT Adaptability Heterogeneous Mobility Scaling Systems Issues Relate to Technologies NA 1
0 1 0 1 DARPA Each of these systems issues can only be pushed forward with adequate support from the underlying technologies. The technologies have certain levels of development as they relate to the systems issues. Evaluation Scale: 0 = no idea 1 = fragile lab demo 2 = solid lab demo
9 Plantbots Capabilities Accumulate/convert energy, information, provide shelter (e.g., for short-lived bio-sensors), resupply; no selflocomotion for whole plant Benefits Limited mobility (seeds, creepers) can lead to advantage in information or energy collection Will provide the infrastructure for the mobile ecology components Challenge: requires extensive new research to devise appropriate forms and interoperation ISAT DARPA Communications SelfDeployment air drop
spreads over tree climbs up, establishes new nettwork climbs down 0not to scale ISAT sends out network on ground DARPA
'bots crawl mobile on jungle floor 1 Sensor State of the Art Current: Lots of low-power compact sensors exist acoustic, magnetic, seismic, pressure, IR, and visible Other sensors require considerable development to meet reliability/size requirements, e.g. bio/chem In general, cost dominated by communications and signal processing, rather than the sensor itself Imaging (IR or visible) costly in signal processing and (especially) communications Active sensors (e.g. radar) costly in power; require energy support network, cueing by other sensors for sustainability Future - Systems Approach: Exploit large numbers of sensors via self-organizing mobile networks ISAT
DARPA 2 Self Configuring Networks General-Purpose Networks wont work: set-up is labor-intensive, even for military field command posts cant be deployed in denied areas pushing the limits result in high energy/complexity costs Future Mobile Sensor Networks by contrast are relaxed in all aspects if processing is done locally exploitation of application and mobility allows energyefficient and scalable design ISAT DARPA 3 Benefits of Mobile Sensor Networks Current: static distributed sensor net
provides dense data gathering but, taxes information management through large numbers Small motion can dramatically improve detection and communication e.g., maximize field of view, line-of-sight, form synthetic apertures with better signal need many fewer elements Larger motion enables dynamic network deployment repair network failures, track and investigate threats beyond initial region of sensors extend or change detection region ISAT DARPA 4 Energy Generation/Extraction/Distribution Many methods 1. battery exchange 2. wires (incl. telephone and power grid) 3. solar
4. wind/water/waves 5. beaming (incl. concentrator mirrors) 6. hydrocarbon/fuel cells 7. convoys/depot system 8. animals (burrs and lures) 9. vehicles (burrs; exploit vibrations) 10. hybrid, e.g., both capacitors and batteries for high currents Research required into how to best combine methods for particular systems and missions ISAT DARPA 5 Energy Conversion / Sustainment micro-flyer moves battery
plugs in creeper comes out ISAT DARPA 6 Future Energy Management Sustainment through ecology Design of energy system has large impact on sustainability; e.g. plantbot energy network for energy accumulation and distribution Efficient use through distributed information Network provides global information to minimize energy waste navigation assistance, actuation/mobility avoidance, resource discovery and management, exploitation of heterogeneity of ability/location
ISAT DARPA 7 Cooperation: The Lessons of Ants Specialization and castes enable range of tasks to be performed Cooperative behaviors enlarge the set of tasks Main benefits of colonies however are: parallelism of tasks collective reliability with individual unreliability Ants apply distributed algorithms for collective control Much more research is needed to enable robot colonies to get these kinds of benefits ISAT DARPA 8 networking, competing, cooperating, distributing, sweepin
Current cooperative robots are mostly homogeneous, and never more than 20 robots ISAT DARPA 9 Robot Cooperation Challenges Centralized systems are brittle and require excessive communications resources. Must identify effective heuristics for distributed coordination Communications and energy network self-organization cannot be general purpose Cooperation must be pursued in applications context Lack of operational data Field tests to discover the needed behaviors for particular missions, and integrate human operators and larger military/industrial infrastructure
Lack of general theory of cooperation With a better understanding, can reduce number of experiments ISAT DARPA 0 Today: Robot Ecology Today & Tomorrow Factory automation: adjust environment for convenience of robots Battlefield: unpredictable environment and no infrastructure, and thus many people to sustain each robot Tomorrow: Scaling More than 20 robots
Heterogenous robots Diverse sets of robots working together in sustained missions Adaptivity Context-aware adaptation among members of the ecology for operation in unplanned environments ISAT DARPA 1 A Program Outline Pick a scenario The union of all the scenarios lacks focus--build out from one E.g., installation perimeter security Can start in a non-covert way, and over time introduce covert aspects Can be operationally tested in parallel to existing methods Pursue a set of large scale experiments More than 20 robots working together Dynamic deployment Dynamic task exchange in the case of failures
Heterogeneous robots - including plant and animal robots Pursue maintenance of energy economy as example of autonomous resupply Standardized parts and interfaces for plug and play with large community Integrate situation assessment with autonomous tagging and dynamic networking Pursue mathematical understanding of the systems level in order to get the right language to analyze how to generalize ISAT DARPA
Gametic Isolation: Even if the gametes of two species do meet, they might be unable to fuse to form a zygote. ... Nondisjunction during meiosis results in organisms with extra sets of chromosomes. Example: plants may have a 4n chromosome...
CTC / MTC 222 Strength of Materials Final Review Final Exam Wednesday, December 12, 10:15 -12:15 30% of grade Graded on the basis of 30 points in increments of ½ point Open book May use notes from first two tests...
Graph Drawing There are many methods to draw untangled pictures of graphs. Two such methods: GIOTTO method Force-directed method GIOTTO Batini et al. began to investigate drawing ER diagrams in the early 1980s.
• De nouveaux instruments seront mis en œuvre : groupe permanent de liaison avec le ministre pour la concertation militaire, mise en œuvre des accords de Bercy et poursuite de l'agenda social dans le cadre d'un dialogue nourri et respectueux...
Certain lipophobic molecules (e.g., glucose, amino acids, and ions) use _____, both of which: Exhibit _ Are _____; rate is determined by number of carriers or channels. Can be regulated in terms of activity and quantity
An Introduction to CBT Using Blended Learning (Days 1-2) M. Zachary Rosenthal, Ph.D. Dept. of Psychiatry & Behavioral Sciences (Duke University Medical Center). Dept. of Psychology and Neuroscience (Duke University) Director, Dept. of Psychiatry & Behavioral Sciences Adult Outpatient Psychiatric...
Two nights with a ~6 m or larger telescope between 2017-03-21 and 04-03 needed. Questions/Issues: What will be our priorities for the above objectives? Should we try to get telescope time on some 4+ m telescopes through the normal channels...