Patrolling strategies for multiple autonomous vehicles Giuseppe Casalino, Gianluca Antonelli, Alessandro Marino Integrated Systems for Marine Enviroment Ancona - Cassino - Genova - Lecce - Pisa - Trieste - Verona http://www.isme.unige.it
ISME in brief Italian Interuniversity research center established in 1999 Sites: Ancona Cassino Genova Lecce Pisa Trieste Verona Infrastructures from all departements > 30 researchers (not full time)
ISME main research areas Underwater robotics ROV AUV UW manipulation guidance navigation & control Underwater acoustics geoacoustics acoustics tomography imaging sonar Signal & data processing geographical information systems decision support systems classification & data fusion
ISME fundings Within FP5-FP6-FP7: 6 European projects Several regional and national projects Technology transfer Formal agreements with NATO Undersea Research Center Italian Navy National University of Singapore Private companies
UAN fundings : FP7 - Cooperation - ICT/Security kind : Collaborative Project (STREP) duration : 3 years start : Oct. 2008 budget : 3Me http://www.ua-net.eu/
TRIDENT fundings : FP7 - Cooperation - ICT - Challenge 2 Cognitive Systems, Interaction, Robotics kind : Collaborative Project (STREP) duration : 3 years start : March 2010 budget : 3.2 Me http://www.irs.uji.es/trident/
CO 3 AUVs fundings : FP7 - Cooperation - ICT - Challenge 2 Cognitive Systems, Interaction, Robotics kind : Collaborative Project (STREP) acronym : CO 3 AUVs duration : 3 years start : Feb 2009 budget : 2.5 Me http://robotics.jacobs-university.de/projects/co3-auvs/index.htm
CO 3 AUVs - Abstract The aim of the CO 3 AUVs project is to provide a significant improvement on the cooperation control for multiple AUVs. Several aspects will be investigated, the situation awareness, the deliberation and navigation, the behavioral control, strictly linked with the communications issues. As a result, the team of AUVs will cooperate, communicate, it will be robust with respect to failures and environmental changes; the key challenge of the project will be an harbor scenario where additional difficulties with respect to open sea applications arise.
Patrolling Merriam-Webster The action of traversing a district or beat or of going the rounds along a chain of guards for observation or the maintenance of security Two problems afforded with different approaches Border Area
Patrolling Constraints Totally decentralized Robust to wide range of failures communications vehicle loss vehicle still Flexible/scalable to the number of vehicles add vehicles anytime Possibility to tailor wrt communication capabilities Not optimal but benchmarking required To be implemented on a real set-up obstacles...
Patrolling Mathematically strong overlap with (dynamic) coverage deployment resource allocation sampling exploration monitoring slight differences depending on assumptions and objective functions
One page on NSB Null-Space-based Behavioral control is yet another behavioral control We like it because: Based on a popular task priority algorithm for industrial robots Exact decomposition of the task errors no corruption Rigorous stability analysis Easy parameter tuning basically decoupled, embed in priority We developed it... Best performance with large DOFs (not really with planar vehicles thus...)
Second of the one page NSB A simple experimental outcome: competitive method 10 5 start 0 5 10 15 obstacle goal 0 10 20 30 40
Second of the one page NSB A simple experimental outcome: cooperative method 10 5 start 0 5 10 15 obstacle goal 20 0 10 20 30 40
Second of the one page NSB A simple experimental outcome: 10 5 start 0 5 10 15 NSB obstacle goal 20 0 10 20 30 40
Border patrolling Solution based on a kind of behavioral control Patrolling embed in the tasks definition and coordination No communication allowed by assumption Experimental results with ground robots Supervisor Level Supervisor Action Level Actions NSB Behaviors Actuators Sensors Robot Level Real Robot Simulator
Border patrolling Definition of several elementary behaviors Reach frontier Patrol (counter)clockwise Teammate avoidance Friend avoidance Basically move-to-goal tasks and obstacle-avoidance
Border patrolling Actions, a frozen priorization of elementary behaviors sensing/perception commands elementary behaviors actions supervisor
Border patrolling Different techniques for the supervisor, in charge of switching among actions Finite State Automata MS1: friend in visibility area Fuzzy logic Action Friend Avoidance 1 low medium DistancefromBorder high low 1 medium DistanceFromNeighbor high!c1 C1 MS2: border not in the visibility area!c1 c1 c1 Action Reach Action Teammate Frontiet Avoidance!c1 Degree of membership 0.8 0.6 0.4 0.2 0 0 2 4 6 8 10 [m] Degree of membership 0.8 0.6 0.4 0.2 0 0 2 4 6 8 10 [m] C2 MS3: border in the visibility area!c2 onmyleft 1 NeighborPosition notpatrolling onmyright 1 ReachBorderLevel, KeepGoing, PatrolCW, PatrolCCW low medium high!c1 c1 c1 Action Keep Action Teammate Going Avoidance!c1 Degree of membership 0.8 0.6 0.4 0.2 Degree of membership 0.8 0.6 0.4 0.2 C3 MS4: robot patrolling!c3 c2 c3 Action Patrol Action Patrol c4 CW CCW!c2!c3 c2 Action Keep c3 Going -!=not operator &=and operator C1=dist. from friend < saft. area C2=dist. from border < Visibility Range C3=dist. from border <= Threshold c1=team. in saft. area c2=team. on left side c3=team. on right side c4=!c2 &!c3. proofs of consistency and completeness 0 1 0.5 0 0.5 1 [] 0 0 0.2 0.4 0.6 0.8 1 []
Border patrolling - numerical validation Dozens of numerical simulations by changing the key parameters: crowded/solitary patrolling friends faults simultaneous events sensor noise border shape y [m] y [m] 200 time = 0.4s 150 100 50 0 50 100 150 200 200 100 0 100 200 x [m] 200 time = 7s 150 100 50 0 50 100 150 200 200 100 0 100 200 x [m] y [m] y [m] 200 time = 4s 150 100 50 0 50 100 150 200 200 100 0 100 200 x [m] 200 time = 25s 150 100 50 0 50 100 150 200 200 100 0 100 200 x [m]
Border patrolling - experimental validation Distributed Intelligence Laboratory University of Tennessee
Border patrolling - benchmarking Instantaneous Segment Idleness (ISI): time elapsed while the single segment is not visited; Instantaneous Border Idleness (IBI): average of the ISI for each segment; Border Idleness (BI): average of the IBI 15 10 Border Idleness NSB Optimum [s] 5 6 9 12 15 18 21 24 27 30 33 Number of Robots
Border patrolling - benchmarking Coverage index (ratio between the patrolled and the empty border) in case of friends and faults Coverage Index 1 0.8 0.6 0.4 0.2 1 5 10 15 20 25 30 [s] Coverage Index 0.8 0.6 0.4 0.2 5 10 15 20 25 30 [s]
Area patrolling - introduction Probabilistic approach Voronoi-based Map-based Communication required only to exchange key data of the maps Motion computed to increase information Framework to handle Spatial variability patrol more often gates Time-dependency come back to visited places Asynchronous spot visiting demand Originally developed as adaptive sampling strategy
Voronoi partitions I The Voronoi partitions (tessellations/diagrams) are a subdivisions of a set S characterized by a metric with respect to a finite number of points belonging to the set the cells union gives back the set the cells intersections is null computation of the cells is a decentralized algorithm without communication needed
Voronoi partitions II Spontaneous distribution of restaurants
Voronoi partitions III Voronoi in nature
Voronoi partitions IV Voronoi in art: Escher
Area patrolling - background I Variable of interest is a Gaussian process Given the points of measurements done... and one to do... Sa = { (x a 1,t a 1),(x a 2,t a 2),...,(x a n a,t a n a ) } Sp = (x p,t) Synthetic Gaussian representation of the condition distribution { ˆµ = µ(x p,t)+c(xp,t) T Σ 1 Sa (y a µa) ˆσ = K(f(xp,t),f(xp,t)) c(xp,t) T Σ 1 Sa c(x p,t) c represents the covariances of the acquired points vis new one
Area patrolling - description I The variable to be sampled is a confidence map Reducing the uncertainty means increasing the highlighted term ˆµ = µ(xp,t)+c(xp,t) T Σ 1 Sa (y a µa) ˆσ = K(f(xp,t),f(xp,t)) c(xp,t) T Σ 1 Sa c(x p,t) }{{} ξ > ξ example
Area patrolling - description II Distribute the computation among the vehicles each vehicle in its own Voronoi cell Compute the optimal motion to reduce uncertainty several choices possible
Area patrolling - accuracy Based on: communication bit-rate computational capability area dimension
Area patrolling - numerical validation Dozens of numerical simulations by changing the key parameters: crowded/solitary patrolling faults obstacles sensor noise area shape/dimension communication bit-rate
Area patrolling - experimental validation Laboratory of Robotics and Systems in Engineering and Science IST, Technical University of Lisbon
Area patrolling - experimental validation 3 Medusas switched off only for low battery virtual obstacle Laboratory of Robotics and Systems in Engineering and Science IST, Technical University of Lisbon
Area patrolling - some benchmarking With a static field the coverage index always tends to one Coverage Index 1 0.8 [ ] 0.6 0.4 0.2 0 200 400 600 step 800 1000
Area patrolling - some benchmarking Comparison between different approaches 2 1.5 [ ] 1 0.5 Lawnmower Proposed Random Deployment time varying case same parameters lawnmower rigid wrt vehicle loss deployment suffers from theoretical flaws 0 0 200 400 600 800 1000 1200 step
Area patrolling - future active behavior against intruder
Area patrolling - challenge reuse pieces of code for different set up
Where do the equations are? I G. Antonelli and S. Chiaverini. Kinematic control of platoons of autonomous vehicles. IEEE Transactions on Robotics, 22(6):1285 1292, Dec. 2006. G. Antonelli, F. Arrichiello, and S. Chiaverini. The Null-Space-based Behavioral control for autonomous robotic systems. Journal of Intelligent Service Robotics, 1(1):27 39, online March 2007, printed January 2008. G. Antonelli. Stability analysis for prioritized closed-loop inverse kinematic algorithms for redundant robotic systems. IEEE Transactions on Robotics, 25(5):985 994, October 2009.
Where do the equations are? II A. Marino, L. Parker, G. Antonelli, and F. Caccavale. Fuzzy behavioral control for multi-robot border patrol. In 17th Mediterranean Conference on Control and Automation, Thessaloniki, GR, June 2009. A. Marino, L. Parker, G. Antonelli, and F. Caccavale. Behavioral control for multi-robot perimeter patrol: A finite state automata approach. In Proceedings 2009 IEEE International Conference on Robotics and Automation, pages 831 836, Kobe, J, May 2009. A. Marino, L. Parker, G. Antonelli, F. Caccavale, and S. Chiaverini. A fault-tolerant modular control approach to multi-robot perimeter patrol. In 2009 IEEE International Conference on Robotics and Biomimetics (ROBIO 2009), Guilin, PRC, December 2009.
Where do the equations are? III G. Antonelli, F. Arrichiello, and S. Chiaverini. The NSB control: a behavior-based approach for multi-robot systems. Paladyn Journal of Behavioral Robotics, 1(1):48 56, 2010. G. Antonelli, S. Chiaverini, and A. Marino. Decentralized deployment with obstacle avoidance for AUVs. In 18th IFAC World Congress, Milan, I, August 2011.
Acknowledgements in rigorous casual order Lynne Parker Pedro Aguiar Fabrizio Caccavale Antonio Pascoal Stefano Chiaverini Filippo Arrichiello
Patrolling strategies for multiple autonomous vehicles Giuseppe Casalino, Gianluca Antonelli, Alessandro Marino Integrated Systems for Marine Enviroment Ancona - Cassino - Genova - Lecce - Pisa - Trieste - Verona http://www.isme.unige.it