A | B | C | D | E | F | G | H | I | J | K | L | M | N | O | P | Q | R | S | T | |
---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|

1 | UNIT | Authors | Title | WP | ||||||||||||||||

2 | ||||||||||||||||||||

3 | ETH, Zurich, Switzerland | Nikolaos Kariotoglou, Sean Summers, Davide M. Raimondo and John Lygeros | Stochastic reachability specifications for self-organizing network surveillance systems | Typical multi-agent network systems require a significant amount of resources to be spent on communication in order to achieve synchronized and cooperative behaviour. In an attempt to reduce this communication overhead we introduce a self-organizing scheme that considers only the ability of each agent to handle individual objectives in assigning system-wide tasks. The proposed approach reduces the required communication data to a set of performance indices encoded by stochastic reachability and reported by each agent. The final task allocation among the employed agents is decided by an autonomous supervisory controller and is based on a hierarchical weight assignment among the possible states of the combined system. We develop and illustrate the performance of these ideas on a Pan Tilt camera surveillance system with robotic evaders. | ||||||||||||||||

4 | EC-Lille, France | C. Fiter, L. Hetel, W. Perruquetti, J.-P. Richard | A polytopic approach for state-dependent sampling | The uprising use of embedded systems and Networked Control Systems (NCS) requires reductions of the use of processor and network loads. In this work, we present a state dependent sampling control that maximizes the sampling intervals of state feedback control. We consider linear time invariant systems and guarantee the exponential stability of the system origin for a given decay rate. The proof of the exponential stability is based on a quadratic Lyapunov function which is computed, thanks to LMIs, so as to optimize some performance criterion on the sampling intervals. A mapping of the state space is then designed offline: it computes for each state of the state space the maximum allowable sampling interval, which makes it possible to reduce the number of actuations during the real-time control of the system. | ||||||||||||||||

5 | University of L'Aquila, Italy | Maria D. Di Benedetto, A. D’Innocenzo, E. Serra | Fault Tolerant Control of Multi-Hop Control Networks | A Multi-hop Control Network (MCN) consists of a plant where the communication between sensors, actuators and computational units is supported by a (wireless) multi-hop communication network, and data flow is performed using scheduling and routing of sensing and actuation data. We address the control design problem on a MCN that is subject to link failures (e.g. malfunctions or battery discharge of a communication nodes, communication link drops, malicious intrusions), where the plant is a SISO LTI system. We first characterize controllability and observability of a MCN, by means of necessary and sufficient conditions on the plant dynamics and on the communication scheduling and routing. We provide a methodology to design scheduling and routing, in order to satisfy controllability and observability of a MCN for any link failure that preserves connectivity of the radio connectivity graph. Then, we characterize the problem of detecting the failure of links of the radio connectivity graph: we provide necessary and sufficient conditions on the plant dynamics and on the communication protocol, and we provide a methodology to design the network topology to satisfy the above conditions. We apply our results to a case study in building automation, namely to a HVAC wireless control problem. | ||||||||||||||||

6 | TU Delft | Noortje Groot, Mohammad Hajiahmadi, Bart De Schutter, Hans Hellendoorn | Mixed Integer Linear Programming Approaches for Model-based Predictive Traffic Control | WP2 | As alternatives to the rather intensive computations when solving traffic control problems in a model-based predictive control (MPC) context, a transformation of the nonlinear program into a mixed integer linear program (MILP) is proposed. This is done based on two different traffic flow models: METANET and link transmission model (LTM). It should be noted that MILPs are in general computationally complex, however efficient to solve for relatively small problem instances. Therefore, in case studies the trade-off between accuracy and computational speed has been analyzed, where the MILP approach showed a significant speed-up compared to the original nonlinear-nonconvex problem. In the METANET model, piecewise-affine (PWA) approximations of several nonlinear model equations were made. For this purpose, several approaches have been adopted and compared, where we have made use of model properties and physical insight to improve the PWA approximations. The resulting MPC problem based on the PWA METANET prediction model can be written as an MILP. Similar results have been obtained when using this approach for the less complex link transmission model (LTM). Since this model is already PWA, the MPC problem based on LTM can also be transformed into an MILP. An extension of the original LTM to include ramp-metering is proposed, in order to use this model for traffic control purposes. | |||||||||||||||

7 | TU Delft | Zhe Cong, Alfredo Núñez, Bart De Schutter, Robert Babuška | Computational Intelligence Methods for Traffic: Traffic Monitoring using Distributed Interval Fuzzy Models and Ant Colony for Dynamic Traffic Routing | WP2 WP4 | Traffic congestion and the inefficient operation of traffic networks are critical problems due to the important costs produced by travel time delays along with their negative impact on the environment. To tackle the traffic congestion problems, the use of intelligent traffic management and control, seems a good alternative to obtain sustainable mobility of the people, especially in cases that the construction of an alternative road is just not feasible, etc. In this poster we present the results of using two different methods from the Computational Intelligence framework for traffic monitoring and control: Distributed Interval Fuzzy Models and Ant Colony Optimization. Using historical data of the density measured on different sections of the freeway, the idea is to find fuzzy confidence intervals that define the bands that contain almost all the density measurements. The purpose of the proposed approach is twofold. First, to obtain a band as narrow as possible for each of the sections of the freeway. And second, to have a high percentage of the data contained in the bands. The method we propose is completely distributed, and can be used not only to describe any uncertain nonlinear distributed parameter system but also as a key element in a robust controller. An empirical validation of the method is presented by relying on real-life data measured on a portion of the A12 freeway in The Netherlands. The dynamic traffic routing (DTR) refers to the process of (re)directing traffic at junctions in a traffic network corresponding to the evolving traffic conditions as time progresses. We consider the DTR problem for a traffic network defined as a directed graph, to deal with the mathematical aspects of the resulting optimization problem from the viewpoint of network flow theory. Traffic networks may have thousands of links and nodes, resulting in a sizable and computationally complex nonlinear, non-convex DTR optimization problem. To solve this problem Ant Colony Optimization (ACO) is chosen as the optimization method for this problem because of its powerful optimization heuristic for combinatorial optimization problems. However, the standard ACO algorithm is not capable of solving the routing optimization problem aimed at the system optimum, and therefore a new ACO algorithm is developed to achieve the goal of finding the optimal distribution of traffic flows in the network. | |||||||||||||||

8 | TU Delft | Minh Dang Doan, Tamás Keviczky, Bart De Schutter | A distributed model predictive control method using Fenchel's duality for networked systems | Wp2 | We present a distributed version of Han's parallel method for convex optimization that can be used for distributed model predictive control (DMPC) of networked systems described by dynamically coupled linear subsystems. The underlying decomposition technique relies on Fenchel's duality and allows subproblems to be solved using local communications only. The convergence rate of the method is further improved and it is illustrated using a test case of a water system in a multi-reach river setup. | |||||||||||||||

9 | KTH, Sweeden | Martin Jakobsson and Carlo Fischione | A Comparative Analysis of the Fast-Lipschitz Convergence Speed | Fast-Lipschitz optimization is a recently proposed framework useful for a class of distributed optimization problems with important applications over complex, peer-to-peer, and large networks. The properties of Fast- Lipschitz problems allow to compute the solution without having to introduce Lagrange multipliers. This is highly beneficial, since multipliers need to be communicated across the network and thus increase the communication complexity of solution algorithms. Although the convergence speed of Fast-Lipschitz optimization methods often outperforms Lagrangian methods in practice, there is not yet a theoretical analysis. Here, a fundamental step towards such an analysis is provided. The convergence of the Fast-Lipschitz fixed point iterations and a first order Lagrangian method is compared. The challenging part of the analysis consists in deriving bounds on the conditioning number of both the Lagrangian and the Fast-Lipschitz methods. Sufficient conditions for superior convergence of the Fast-Lipschitz method are established. The results are illustrated by simple examples. It is concluded that optimization problems with quadratic cost functions and linear constraints are always better solved by Fast-Lipschitz optimization methods, provided that certain conditions holds on the eigenvalues of the Hessian of the cost function and constraints. | ||||||||||||||||

10 | University of Trento, Italy; ISS-SUPELEC, France | Daniele Fontanelli, Luigi Palopoli, Luca Greco | Deterministic and Stochastic QoS Provision for Real-Time Control Systems | In this work, we propose two adaptive scheduling approaches to support real-time control applications with highly varying computation times. The use of a resource reservation scheduler enables the construction of a dynamicmodel describing the evolution of the computing delays, which can be incorporated in the system closed loop dynamics. The two approaches differ for the assumptions on the sequence of computation time. In the first approach, we have only an aggregate information (best case and worst case computation time) and design an adaptive scheduler that maintains the delay within the maximum bound compatible with the asymptotic stability of the system. In the second case, we assume a deeper knowledge on the distribution of the computation time and design an adaptive scheduler that ensures second moment stability of the system. | ||||||||||||||||

11 | University of Kassel, Germany | D. Gross, M. Jilg, O. Stursberg | Data Distribution in Distributed Model Predictive Control | In this work we investigate a scheme for data distribution in distributed model predictive control (DMPC) of interconnected linear discrete-time systems. In DMPC, information from other subsystems is used to improve local closed-loop performance, what usually results in high communication cost. Thus, a key issue is how and when to communicate information between the controllers. Instead of considering a given and static communication scheme (e.g. communication between all neighbors), we propose an approach to balance communication load and system performance using sensitivity analysis of the DMPC problem. | ||||||||||||||||

12 | IFSTTAR – LTN (Laboratory of New Technologies) Versailles, France | Alexandre De Bernardinis, Gérard Coquery | EV-Electrification: failure modes and related safety aspects | One of the main challenges concerning the electric vehicle (EV) is the all-electrification of the vehicle functions by means of replacing hydraulic actuators by electric ones and by using electric storage devices for power assistance in order to allow the concept of the “full EV”. EV-electrification reinforces the obligation for the electric vehicle to be reliable, available and safe. Failures or malfunctions of some electric devices and electronic systems have to be investigated, in particular regarding the electric traction motor behavior, driving control and safety. Corrective actions have to be taken to ensure the continuity of operation and preserve the passenger safety. Regenerative braking and its impact on the adapted driving assistance functions need also investigation where both electric and automation aspects interact in this case. Also, the electric vehicle interacts with the charging infrastructure within a network, and the electric system composed of the EV with the charging infrastructure should be reliable and safe. Furthermore, one important perspective is the insertion of the EV in the urban “electric road” concept with dedicated electric charging infrastructures supported by automated communication systems. We give in this presentation, first a survey on the mentioned whole set of new concepts that are needed for the EV-Electrification, and then we show the IFSTTAR LTN current work and its technological achievements, including the test benches and projects results. | ||||||||||||||||

13 | University of Pisa, Italy | Stefano Falasca, Massimiliano Gamba, Tommaso Fabbri, Davide Fenucci, Luca Greco, Antoine Chaillet, Antonio Bicchi | An output-feedback dynamic approach for the control over packet-switching networks | We consider the stabilization of a nonlinear system when a digital network is used for communication with the controller. Several network-induced constraints are in place: variable transfer intervals; large, time-varying delays; non-simultaneous access to the network, packetized data transmissions. We adopt a model-based prediction strategy, in which control sequences are sent exploiting the large packet payload. Recent results in this framework have shown that static-feedback stabilization can be achieved if a compromise between model uncertainties, delays and maximum time between communications is met. We here extend this approach to dynamic controllers under the assumption that their internal state is consistently updated. Simulation results illustrate the efficiency of the approach and show that the consistent update of the internal state of the controller is required for a limited impact on the nominal behavior of the controlled system. | ||||||||||||||||

14 | University of Evry, France | Lydie Nouveliere | How to better eco-drive a light vehicle using embedded systems? | The sector of the road transports does not escape the trend consisting of not emitting too much Green House Gas (GHG). It is the second largest CO2 emitter with 6.6 Gt in 2008 [1, 2]. Nowadays, electric and hybrid vehicles make their appearance, but several technical and technological limits induce still a long time before they will replace our conventional thermal / diesel vehicles. But these electric/hybrid vehicles represent a good alternative to reduce the dependence of our displacements on the fossil energies. Thanks to the Advanced Driving Assistance Systems (ADAS), a today light vehicle is now equipped with several embedded automated systems. These ADAS thus permit to the driver to be safe while improving his comfort. Since the automobile sector appears as one of the main Green House Gas (GHG) transmitter, many efforts must be made to answer to the anti-polluting norms that are more and more restrictive.One of the fast and low cost solutions consists of reducing the fuel consumption by acting on the changing of the driver behavior. The main objective is then to help the driver to globally adopt a more economical, ecological as well as safer driving. The system advises the driver by some channels (speed display, alarms,... [3]) via an HMI. No active driving assistance system has been developed and commercialized yet. According to Evans (1979), a driver could reduce his fuel consumption by as much as 14% without increasing trip time, just by reducing acceleration levels and generally driving more ”gently”, combined with a skillful avoidance of stops. An experiment conducted by LIVIC research lab in the Versailles area, France, confirms that a gain between 10% to 20% can be obtained just by changing the driving style. This shows a great potential for fuel saving by future actions that will make the driver be awake of economical and ecological driving.Some prototypes of fuel-efficiency tool support are presented through van der Voort et al. [4]; and Hellstrom et al. [5]. These works are carried out in the context of heavy vehicles. In the field of public transport, the PREDIT-ANR-ANGO project [6] aimed at designing a fuel-efficiency driving assistance system for a city bus. Optimal trajectories have been computed with a known itinerary and on an exclusive right of way. There also are some ADAS on the market like those issued from the GERICO project or from the eco-driving dashboard from HONDA. On the other side, several systems are based on the MDD concept (Modern Drive Devices). The main difference between them comes from the way to give the advice to the driver. The ADAS are able to send the information in real-time, adapted to the journey; the MDD methodology underlines the analysis and devices given to the driver once the journey is finished with some statistical tools in order the driver to improve his next journey. In this presentation, our approach is different from the already existing studies and technical solutions by its strategy used. This one takes into account both infrastructure characteristics and the current vehicle situation. To do that, a digital map containing the road geometry (slopes, curves,…) is used and the legal speed is known [7]. The headway spacing is measured and the vehicle longitudinal characteristics are also known [8, 9]. Starting from these data, the problem of optimization of the fuel consumption is formulated and solved with dynamic programming technique [10]. A real-time strategy is then adopted in order to render the system adaptive to the traffic conditions. A coupling between the safety problem and the low fuel consumption objective is thus reached. After some simulations on MATLAB/SIMULINK, this EDAS (Ecological Driving Assistance System) was tested on a prototype vehicle driving on a national road test track type. An HMI is used to transmit the device to the driver (optimal speed and optimal gear).The experimental results are shown. A first analysis of the economical gain is given.This presentation is mainly linked to the WP2 of HYCON2, by the way of the communication established between the vehicle and its sensors and embedded architecture, between the infrastructure and the vehicle, between the digital map and the optimization algorithm, between the vehicle and the driver thanks to the HMI (Human-Machine Interface).References :[1] IEA (International Energy Agency), “How the energy sector can deliver on a climate agreement in Copenhagen”, Special early except of the Work Energy Outlook 2009 for the UNFCCC meeting, Oct. 2009, Bangkok, Thaïland.[2] IEA (International Energy Agency), ”CO2 emissions from fuel combustion 2010 – highlights, Edition 2010.[3] M. Van Der Voort, M. Dougherty, M. Van Maarseveen, , “A prototype fuel-efficiency support tool”, Transportation Research Part C: Emerging Technologies, 9, 279–296, 2001.[4] J. Barbé, G.A. Boy, M. Sans, “Gerico : A human centered eco-driving system”, In Proc. Int. Conf. 10th IFAC/IFIP/IFORS/IEA Symposium on Analysis, Design, and Evaluation of Human-Machine Systems, Seoul, Korea, 2007.[5] E. Hellstrom, A. Froberg, L. Nielsen, “A real-time fuel optimal cruise controller for heavy trucks using road topography information”, SAE World Congress, number 2006-01-000, 2005, Detroit, USA.[6] L. Nouveliere, M. Braci, L. Menhour, H-T. Luu, “Fuel consumption optimization for a city bus”, UKACC Control Conference, 2008, Manchester, UK.[7] L. Nouveliere, M. Braci, L. Menhour, H. T. Luu, S. Mammar, “Infrastructure based fuel consumption optimization of a vehicle”, 9th international symposium on Advanced VEhicle Control AVEC08, 2008, Kobé, Japan.[8] H-T.Luu, L. Nouvelière, V. Hoarau, S. Mammar, “Vehicle Speed Control For A Safety-Economical-Ecological Compromise : Development Of A Driver Assistance System”, Proceedings of the AVEC10 Conference, August 2010, Loughborough, UK.[9] H-T Luu, L. Nouvelière, S. Mammar, “Towards a Safer Ecological Driver Aassistance System”, Proceedings of the ITS World Congress, Oct. 2010, Busan,Korea.[10] H-T Luu, L. Nouvelière, S. Mammar, “Dynamic programming for fuel consumption optimization on light vehicle”, Proceedings of the IFAC-AAC Conference, July 2010, Munich, Germany. | ||||||||||||||||

15 | IFSTTAR – LTN (Laboratory of New Technologies) Versailles, France | Mariana Netto, Jean-Marc Blosseville | In-vehicle embedded control systems for safety and mobility: use cases in relation with the needed actuator and sensor units and the human-machine cooperation | Making in-vehicle embedded control systems, with a suited choice of the set of units (mainly the actuators and the sensors), and the control algorithm to connect them, and at the same time ensuring a constructive human-machine interaction, is not a simple task. The first step should be the definition of the scope of action of the assistance system, that is based on the macroscopic level of the traffic situation for which the system is expected to increase safety. This will help in the following step to begin to set the choice of the mode on which the system will act: lateral, or longitudinal. If we use a top-down reasoning, three groups of assistance systems, based on the time-to-the- accident, the associated sensors and the connection with the infrastructure, can be defined (see [2] for a description of these groups). If accident precaution more than prevention is concerned, communicating warning based systems (see eg [1], [7]) are to be used. In the case of the continuous monitoring of the surround of the vehicle for frontal, rear and lateral collision avoidance, warning and active systems are required [4]. Finally, if the accident is imminent, control is necessarily required. The needed sensors and communicating systems, depend on which group the assistance system belongs. The time-to-the-accident is also called temporal frame of the system. Going further, the assistance system temporal frame is in turn closely related to the level of cooperation with the human-being: as an example, if the system is constructed to be the less intrusive as possible with the human action, it should be then activated in the last moment, when it would not be possible at all to the human-being to recover the vehicle stability [6]. With respect to this case, in the case of lateral control, we could imagine that more actuators are needed, requiring multivariable control, and in general more complex and performing algorithms, since the vehicle is already in a very instable state. Different cooperation levels between the driver and the system are then related to the group on which the assistance system belongs ([2] and [5] provide research on HMI in embedded vehicle control systems). This work analyses a set of different scenarios and use cases for which assistance systems are built to improve safety or mobility, some recent control algorithms proposed by us as well as others proposed in the literature, discussing the advantages/disadvantages of each one, the sensors, the actuators, and then the related cost, the need for communication systems or not and the chosen temporal frames, including HMI information, as well as robustness issues linked to the inputs to the systems (lane detection by camera for example). | ||||||||||||||||

16 | University of Pisa, Italy | S. Martini, A. Fagiolini, and A. Bicchi | Behavior Classification, Security, and Consensus in Societies of Robots | We consider how very large numbers of heterogenous robots, differing in their bodies, sensing and intelligence, may be made to coexist, communicate, and compete fairly towards achieving their individual goals, i.e. to build a ``society of robots''. We introduce a formalism that allows a large variety of possible cooperative systems to be uniformly modeled and we discuss some characteristics that the rules defining acceptable social behaviors should possess. We consider threats that may be posed to such a society by the misbehaviors of some of its members, due either to faults or malice, and the possibility to detect and isolate them through cooperation of peers. We study the problem of classifying a set of robotic agents, based on their dynamics or the interaction protocols they obeys, as belonging to different ``species". We finally study distributed algorithms that members of the society can use to reach a consensus on the environment and on the integrity of the peers, so as to improve the overall security of the society of robots. | ||||||||||||||||

17 | University of Padova, University of Turin, Italy | D. Borra, E. Lovisari, R. Carli, F. Fagnani, S. Zampieri | Autonomous Calibration Algorithms for Networks of Cameras. | We deal with the important applicative problem of distributed cameras calibration. We model a network of N cameras as an undirected graph in which communicating cameras can measure their relative orientation in a noisy way. These measures can be used in order to minimize a suitable cost function. The shape of this cost function depends on a vector of integers K. We propose two algorithms which in a distributive way estimate such K, comparing advantages and disadvantages of both. Simulations are run on a grid network to prove effectiveness of the algorithms. | ||||||||||||||||

18 | University of L'Aquila, Italy | Alessandro Borri, Giordano Pola, Maria D. Di Benedetto | A Symbolic Approach to the Design of Nonlinear Networked Control Systems | Networked control systems (NCS) are spatially distributed systems where communication among plants, sensors, actuators and controllers occurs in a shared communication network. NCS have been studied for the last ten years and important research results have been obtained. These results are in the area of stability and stabilizability. However, while important, these results must be complemented in different areas to be able to design effective NCS. We approach the control design of NCS using symbolic (finite) models. Symbolic models are abstract descriptions of continuous systems where one symbol corresponds to an ”aggregate” of continuous states. We consider a fairly general multiple-loop network architecture where plants communicate with digital controllers through a shared, non-ideal, communication network characterized by variable sampling and transmission intervals, variable communication delays, quantization errors, packet losses and limited bandwidth. We first derive a procedure to obtain symbolic models that are proven to approximate NCS in the sense of alternating approximate bisimulation. We then use these symbolic models to design symbolic controllers that realize specifications expressed in terms of automata on infinite strings. Efficient algorithms for the synthesis of the proposed controllers are also derived which are inspired by on-the-fly techniques studied in computer science. | ||||||||||||||||

19 | University of L'Aquila, University of Pavia, Italy | Antonella Ferrara, Domenico Bianchi, Giancarlo Ferrari Trecate, Maria Domenica Di Benedetto | Networked control for traffic systems. | Road mobility is indispensable in modern day life and traffic congestion is becoming more and more a high-priority problem in most countries all over the world. For this reason researchers are currently investigating solutions to control traffic in such a way that the traffic state is kept far from congestion and the secondary problems related to congested traffic, such as pollution, are alleviated. Traffic control centers monitor the traffic situation based on video images and measurements from, e.g., loop detectors. Information about traffic density, average velocities, incidents, and so on, used with suitable traffic models enable to predict future traffic states and to establish the control actions to apply. Often, in realistic scenarios, the control actions are computed in a control center located far from the traffic system and then transmitted, through a wireless communication channel, to the actuators placed along the road, i.e. on ramp traffic lights in the considered case of ramp metering control. More specifically, we assume that the communication channel is affected by delays and packet loss. We propose to adopt a control strategy of model predictive control (MPC) nature based on the use of a buffer to compensate the time delays, so that the designed traffic networked control scheme provides performances significantly close to those of the ideal case (i.e. no delay). Moreover, in order to limit the computational load, a modified version of the control strategy is proposed, in which the length of the control horizon of the predictive control algorithm, and, consequently, the buffer length, is updated according to a delay estimation. A simulation analysis of the different networked control proposals is provided, along with the comparison with a conventional MPC equipped with an hold mechanism. The superiority in terms of performance of the proposed networked MPC with adaptive buffer length is apparent. | ||||||||||||||||

20 | UNIPD | Saverio Bolognani, Sandro Zampieri | Distributed reactive power compensation in a smart microgrid | We consider the problem of optimal reactive power compensation for the minimization of power distribution losses in a smart microgrid. We first propose an approximate model for the power distribution network, which allows us to cast the problem into the class of convex quadratic, linearly constrained, optimization problems. We then consider the specific problem of commanding the microgenerators connected to the microgrid, in order to achieve the optimal injection of reactive power. For this task, we design a randomized, gossip-like optimization algorithm. We show how a distributed approach is possible, where microgenerators need to have only a partial knowledge of the problem parameters and of the state, and can perform only local measurements. For the proposed algorithm, we provide conditions for convergence together with an analytic characterization of the convergence speed. The analysis shows that, in radial networks, the best performance can be achieved when we command cooperation among units that are neighbors in the electric topology. Numerical simulations are included to validate the proposed model and to confirm the analytic results about the performance of the proposed algorithm. | ||||||||||||||||

21 | ||||||||||||||||||||

22 | ||||||||||||||||||||

23 | ||||||||||||||||||||

24 | ||||||||||||||||||||

25 | ||||||||||||||||||||

26 | ||||||||||||||||||||

27 | ||||||||||||||||||||

28 | ||||||||||||||||||||

29 | ||||||||||||||||||||

30 | ||||||||||||||||||||

31 | ||||||||||||||||||||

32 | ||||||||||||||||||||

33 | ||||||||||||||||||||

34 | ||||||||||||||||||||

35 | ||||||||||||||||||||

36 | ||||||||||||||||||||

37 | ||||||||||||||||||||

38 | ||||||||||||||||||||

39 | ||||||||||||||||||||

40 | ||||||||||||||||||||

41 | ||||||||||||||||||||

42 | ||||||||||||||||||||

43 | ||||||||||||||||||||

44 | ||||||||||||||||||||

45 | ||||||||||||||||||||

46 | ||||||||||||||||||||

47 | ||||||||||||||||||||

48 | ||||||||||||||||||||

49 | ||||||||||||||||||||

50 | ||||||||||||||||||||

51 | ||||||||||||||||||||

52 | ||||||||||||||||||||

53 | ||||||||||||||||||||

54 | ||||||||||||||||||||

55 | ||||||||||||||||||||

56 | ||||||||||||||||||||

57 | ||||||||||||||||||||

58 | ||||||||||||||||||||

59 | ||||||||||||||||||||

60 | ||||||||||||||||||||

61 | ||||||||||||||||||||

62 | ||||||||||||||||||||

63 | ||||||||||||||||||||

64 | ||||||||||||||||||||

65 | ||||||||||||||||||||

66 | ||||||||||||||||||||

67 | ||||||||||||||||||||

68 | ||||||||||||||||||||

69 | ||||||||||||||||||||

70 | ||||||||||||||||||||

71 | ||||||||||||||||||||

72 | ||||||||||||||||||||

73 | ||||||||||||||||||||

74 | ||||||||||||||||||||

75 | ||||||||||||||||||||

76 | ||||||||||||||||||||

77 | ||||||||||||||||||||

78 | ||||||||||||||||||||

79 | ||||||||||||||||||||

80 | ||||||||||||||||||||

81 | ||||||||||||||||||||

82 | ||||||||||||||||||||

83 | ||||||||||||||||||||

84 | ||||||||||||||||||||

85 | ||||||||||||||||||||

86 | ||||||||||||||||||||

87 | ||||||||||||||||||||

88 | ||||||||||||||||||||

89 | ||||||||||||||||||||

90 | ||||||||||||||||||||

91 | ||||||||||||||||||||

92 | ||||||||||||||||||||

93 | ||||||||||||||||||||

94 | ||||||||||||||||||||

95 | ||||||||||||||||||||

96 | ||||||||||||||||||||

97 | ||||||||||||||||||||

98 | ||||||||||||||||||||

99 | ||||||||||||||||||||

100 |

Loading...