
SILUMOD: A Simulation Language for User Mobility Models Definition in Multihop Networks Yacine Bencha¨ıb Claude Chaudet Institut Tel´ ecom´ / Tel´ ecom´ ParisTech / LTCI Institut Tel´ ecom´ / Tel´ ecom´ ParisTech / LTCI CNRS UMR 5141 CNRS UMR 5141 [email protected] [email protected] Abstract mobility and propose a set of built-in mobility models. A mo- In this paper, we present SILUMOD, an open source (LGPL) bility model is the definition of how nodes move and interact internal domain-specific language that we developed to fa- with their environment (obstacles, other mobiles, etc.). cilitate the study and deployment of users mobility mod- Choosing the right mobility model to evaluate a protocol els for multihop wireless networks. Based on Scala1, SILU- or an algorithm is not trivial. The mobility model should be MOD uses the same syntax and the same libraries as this representative and generic. It should not bias performance language, while adding a series of keywords and operators in a direction or the other. The coherence between model- reflecting the common characteristics of the movements of ing, simulation and experimental results should also be guar- mobile nodes. SILUMOD combines ease of use and effi- anteed and the same mobility model should be available to ciency by ensuring rapid and easy implementation of most all the corresponding tools, which usually use different for- mobility models. malisms to define nodes movements. Simulators, on the one hand, often use a sequence of positions in a coordinates sys- Categories and Subject Descriptors I.6.2 [Simulation and tem, possibly mimicking a real trace. Robots, on the other Modeling]: Simulation Languages hand, are often unable to localize themselves precisely and General Terms Languages, Experimentation do not share the same clock. Their movements are often de- fined by specifying sequences of rotation angles and dis- Keywords Domain-specific language, Wireless multihop tances, or by controlling each wheel’s individual speed. As networks, Mobility models translation is difficult, simple random mobility models such as random waypoint, proposed in [5], or random direction 1. Introduction proposed in [14] are often favored, even though they seldom Research in the field of wireless multihop networks (ad- reflect reality, as shown in [18]. hoc, sensors, vehicular, ...) has extensively studied static net- In this paper, we introduce SILUMOD, a domain specific works and it is today shifting towards mobility. Among the language based on Scala to describe nodes mobility in an set of tools available to the community, several experimen- easy and representative way. The objectives of SILUMOD tal platforms (e.g. IBBT Wireless Lab2, SensLab3, FIT4, are threefold: ...) now include mobile nodes, embarked for example on • As it is a high-level language, researchers can express wheeled robots or on model trains. All packet-level simu- mobility without thinking about implementation. Spe- lators such as ns-2, ns-3, Omnet++ or Opnet handle nodes cific tools and libraries shall perform the conversion to 1 http://www.scala-lang.org/ simulators traces or to particular robots model. 2 http://www.ibbt.be/en/develop-test/ilab-t/wireless-lab • Using similar translation techniques, mobility models 3 http://www.senslab.info can be automatically characterized and checked through 4 http://fit-equipex.fr/english formal verification tools. • Defining a common language allows feeding a mobility models online repository. Models in this library can con- Permission to make digital or hard copies of all or part of this work for personal or stitute a relevant reference for protocols comparison. classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and the full citation In this article, we first detail the goals behind SILUMOD on the first page. To copy otherwise, to republish, to post on servers or to redistribute in section 2. We then present related works in section 3. to lists, requires prior specific permission and/or a fee. APPLC ’12 June 14, 2012, Beijing Section 4 extracts from the literature the characteristics of Copyright c 2012 ACM [to be supplied]. $10.00 a realistic mobility model and lists a set of requirements for SILUMOD. The language is then exposed in section 5 $ns_ at 10.0"$node_(3) setdest 50.0 50.0 10.0" and instantiated through two simple examples in section 6. $ns_ at 20.0"$node_(3) setdest 70.0 150.0 3.0" Section 7 discusses SILUMOD and its evolutions before concluding the paper. Listing 1: Trajectory definition example in NS-2 simulator 2. From mobility specification to the implementation points, but robots that are not equipped with localization de- vices move by defining linear and angular speed, or by con- Three categories of tools are generally used to evaluate pro- trolling the rotation speed of each individual wheel. Trans- tocols and algorithm for wireless multihop networks and lating a high-level mobility specification into such an atomic their representation of nodes mobility differ considerably. language may require more or less complex compilation, but Mathematical modeling often uses statistical properties this translation can be provided for each robot type once and of the mobility models such as the dynamics of the connec- for all. tivity graph, a graph in which edges between two vertices The objective of our work is therefore to provide a high- reflect the situation in which the two nodes are in communi- level description language for mobility that is easy enough cation range. Nodes mobility makes this graph evolve over to hide implementation capabilities and expressive enough to time. Individual mobility may be characterized by the inter- allow complex patterns specification. This language should contact time distribution or by the contact length distribu- allow, through a set of tools, to characterize and verify for- tion. Global properties are rendered by the evolution struc- mally properties of the models and to instantiate a run of the tural parameters such as the graph degree distribution or the model as a list of positions for a simulator or a robot of self- number of connected components, for example. Such statis- localization, or as a list of directives that define the behavior tics, as well as other mobility models properties may be ob- of a robot. tained on the basis of the model description by analysis. Lin et al. [11], for example, use renewal theory to characterize the properties of a few random mobility models. If such anal- 3. Related Work ysis is not possible or too complex, statistical properties may The literature is filled with numerous mobility models that be obtained by extensive simulation. correspond to a scenario or another. However, few articles Simulation tools such as Opnet Modeler5, Omnet++6, aimed at providing a generic way to specify nodes mobility. NS-27 or GlomoSim / Qualnet8 expect nodes movements to Stepanov et al. [16] propose a generic model that decom- be represented as a series of (time, position) couples. List- poses users movement definition in three components. First, ing 1 represents an example of definition of a node mo- the environment, composed of obstacles, points of interest, bility in the NS-2 simulator. The node with ID 3 starts at etc., defines the scene in which mobiles evolve and imposes date t = 10s to move linearly towards the position (x = constraints on the movements. Environment can be extracted 50:0; y = 50:0) at a speed of 10 m=s, then turns at date from a geographic database. Second, the users behavior is t = 20 s towards a second position. These waypoints can be defined by their activities. Users move when they are on defined randomly by internal code or by an external tool, or a trip, going from one place to another. Third, the move- can be extracted from a real trace. Between two waypoints, ment dynamics (speed, acceleration, ...) define how the users which can be arbitrarily close, the node is usually supposed move. The authors propose an XML format to allow users to to move linearly at a given speed. Most simulators imple- specify these three components. A java-based tool (Canu- ment natively the random waypoint and random direction MobiSim) then translates this XML specification to the NS- models. However, several contributions such as [1] point that 2, Glomosim and QualNet formats. Authors provide a few random models do not correspond to reality and bias perfor- random, vehicular and physics-based models as examples. mance evaluation results. The generation of a list of timed Robsys ([17]) focuses on mobile robots programming and positions is immediate from a high-level specification of a proposes a hierarchical language to specify the movement mobility model. of a robots fleet. The high-level, object-oriented language Experimental platforms that use mobile robots usually is dedicated to the specification of multiple robots (agents) rely on languages like which allow specifying a series of behavior, providing ways to express synchronization, com- movements in terms of positions, direction changes, speed munication, etc. while the lower-level language is closer to etc. depending on the physical robot capacities. A robot ca- assembly language and dedicated at controlling one single pable of self-localization will be able to move between way- robot. Mobility is expressed as a set of instructions for a robot (moving to a position, advancing of a few meters, turn- 5 http://www.opnet.com/solutions/network_rd/modeler.html ing, etc.). 6 http://www.omnetpp.org/ Finally, the Condition Probability Event (CPE) model [12] 7 http://www.isi.edu/nsnam/ns/ proposes to represent pedestrians mobility as a set of rules 8 http://pcl.cs.ucla.edu/projects/glomosim/ that mix probabilities and interactions with the environment.
Details
-
File Typepdf
-
Upload Time-
-
Content LanguagesEnglish
-
Upload UserAnonymous/Not logged-in
-
File Pages8 Page
-
File Size-