«To cite this version: Marc-Olivier Killijian, Matthieu Roy, Gaetan Severac. The ARUM Experimentation Platform : an Open Tool to evaluate Mobile ...»
The ARUM Experimentation Platform : an Open Tool
to evaluate Mobile Systems Applications
Marc-Olivier Killijian, Matthieu Roy, Gaetan Severac
To cite this version:
Marc-Olivier Killijian, Matthieu Roy, Gaetan Severac. The ARUM Experimentation Platform
: an Open Tool to evaluate Mobile Systems Applications. AMiRE 2011: 6th International
Symposium on Autonomous Minirobots for Research and Edutainment, May 2011, Bielefeld,
Germany. pp.1, 2011. hal-00667832
HAL Id: hal-00667832 https://hal.archives-ouvertes.fr/hal-00667832 Submitted on 10 Feb 2012 HAL is a multi-disciplinary open access L’archive ouverte pluridisciplinaire HAL, est archive for the deposit and dissemination of sci- destin´e au d´pˆt et ` la diﬀusion de documents e eo a entiﬁc research documents, whether they are pub- scientiﬁques de niveau recherche, publi´s ou non, e lished or not. The documents may come from ´manant des ´tablissements d’enseignement et de e e teaching and research institutions in France or recherche fran¸ais ou ´trangers, des laboratoires c e abroad, or from public or private research centers. publics ou priv´s.
e The ARUM Experimentation Platform : an Open Tool to evaluate Mobile Systems Applications Marc-Olivier Killijian, Matthieu Roy, and Gaetan Severac
This paper present the ARUM robotic platform. Inspired by the needs of realism in mobile networks simulation, this platform is composed of small mobiles robots using real, but attenuated, Wi-Fi communication interfaces. To reproduce at a laboratory scale mobile systems, robots are moving in an 100 square meters area, tracked by a precise positioning system. In this document we present the rational of such simulation solution, provide its complete description, and show how it can be used for evaluation by brieﬂy explaining how to implement speciﬁc algorithms on the computers embedded by the robots. This work is an application of multi-robotics to research, presenting solutions to important problems of multi-robotics.
1 Objectives In this paper, we present the ARUM robotic platform1 targeted at evaluating performance, resilience and robustness of mobile systems. To obtain an efﬁcient evaluation platform, three speciﬁc criteria were considered: Control conditions (real time monitoring, repeatability, ﬂexibility, scalability), Effective implementation (easiness of conﬁguration, devices autonomy, portability, low cost, miniaturization), and Realistic environment (network scale, trafﬁc load, node mobility, positioning, radio broadcast behaviour). To our knowledge this platform is the only one to date to integrate all these features in a single environment.
Marc-Olivier Killijian, Matthieu Roy LAAS–CNRS, 7 avenue du colonel Roche, 31077 Toulouse, France Universite de Toulouse e-mail: firstname.lastname@example.org Gaetan Severac ONERA (DCSD), The French Aerospace Lab, 31055 Toulouse, France, Universit´ de Toulouse, e-mail: email@example.com e 1 ARUM stands for “an Approach for the Resilience of Ubiquitous Mobile Systems”
Fig. 1 A picture of the the ARUM Platform Indeed, current evaluation strategies for distributed and mobile systems can be
split in ﬁve categories:
• Simulators. Simulators are cheap and fast to set up, with almost no limitation in the number of nodes. Due to their scalability and simplicity, they are well suited for initial testing. Furthermore, they may speed up development of theoretical researches because since they allow a perfect monitoring and repeatability [21, 22]. Nevertheless, simulation is based on models of the running environment, and thus cannot reﬂect the real complexity of natural environments, particularly for radio communication and mobility pattern[7, 5, 3].
• Emulators. Emulators are built to physically reproduce connections events using real wired network hardware[16, 19]. They provide features interesting for protocol implementation but they still use simulation to reproduce wireless communication behaviour and mobility.
• Testbeds. The ARUM platform we present in this paper can be classiﬁed in this category. Testbeds are closer to reality thanks to the use of real hardware. They exist since years now, from the historical MIT RoofNet, to the more recent MoteLab2 service. Ideal to ﬁnalize and validate applications before real-life experimentations, testbeds provide much more realistic results than emulators or simulators. But they are also expensive, time consuming and limited by the physical resources/hardware used. Because of those limitations, only a few of them implement real mobility. To the best of our knowledge, two platforms using mobile robots have been recently developed: MINT and Mobile Emulab.
Original solutions that emulate mobility can also be found in the literature, like MOBNET  which varies the transmission power levels of ﬁxed access points.
It is interesting to notice that most of these platforms have to deal with large variations in the communication noise level because of environment perturbations.
2 Harvard Sensor Network Testbed - http://motelab.eecs.harvard.eduThe ARUM platform 3
Such difﬁculties can be problematic during applications development phases, but they are representative of conditions encountered in the real life.
• Hybrid simulators use both simulated networks and real devices, taking advantages and disadvantages of each[18, 23]. They are particularly suited to study, at a low cost, the interconnection of some real devices to a huge network, the latter being simulated.
• Real live experiments. This is, obviously, the more realistic kind of experimentation, but they present inherent technical problems which can bring more technical difﬁculties than scientiﬁc beneﬁts[11, 15]. They are absolutely necessary for commercial applications, because it is impossible to truly simulate real environment yet. Yet, they are very expensive, error-prone, and they do not provide repeatability of experiments, due to the wide variability of real environments. As such, such platforms are not used in the context of research and education.
Fig. 2 Accuracy of evaluation solutions for mobile systems depending on their respective costs Among all these technologies, there is no good or wrong solutions, the best choice depends on a speciﬁc needs and available resources, as shown in Figure 2.
In our case, both for scientiﬁc and for demonstration reasons, we decided to implement a testbed, the ARUM platform. Indeed our primary goal was to complement simulation and allow realistic evaluation of mobile systems, at a laboratory scale. It ﬁnally appeared to be a good platform for demonstration and education, since the platform can be used pedagogically to present various aspects and problems raised by mobile systems. The work describe here is an application of mini-robots to research, in a ﬁeld different from robotics. However the solutions tested and implemented here can be applied to several important problems in multi-robotic ﬁeld (e.g.
positioning, mobility, communication...).
4 Marc-Olivier Killijian, Matthieu Roy, and Gaetan Severac
To complete the goals presented in previous section, we designed an experimental evaluation platform composed of mobile devices. We dispose of a room of approximately 100m2 to emulate systems of different sizes, hence we decided to scale every parameter of the system to ﬁt within our physical constrains. Technically speaking, each mobile device is built with : a programmable mobile hardware able to carry the device itself, a lightweight processing unit equipped with one or several wireless network interfaces and a positioning device. Hardware modelling required a reduction or increase of scale to be able to conduct experiments within the laboratory.
To obtain a realistic environment, all services have been modiﬁed according to the same scale factor.
Table 1 Scale needs In our case, we considered vehicular ad-hoc network experiments . A typical GPS embedded in a moving car is accurate to within 5-20m. So, for our 100m2 indoor environment to be a scaled down representation of a 250000m2 outdoor environment (a scale reduction factor of 50), the indoor positioning accuracy needs to be 10 − 40cm. Table 1 summarizes the required change in scale for all peripherals of a node.
We understand here that to meet those requirements some parts of the development were much more important. The focus was put on the reduced Wi-Fi interfaces, the precise positioning and the node mobility.The different parts of the platform will be detailed in the following section.
3 Technical solutions
3.1 Mobility To reproduce mobile systems conditions, the devices used in the platform must be mobile. But when conducting experiments, a human operator cannot be behind each device, so mobility has to be automated. This is why we considered the use of simple small robots in order to carry around the platform devices. The task of these robots is to implement the mobility of the nodes following a movement scenario.
The ARUM platform 5
Fig. 3 A Mobile Node Picture, without the embedded computer
A node, represented in ﬁgure 3, is implemented in the system using a laptop computer that is carried by a simple robotic platform, that includes all hardware devices, the software under testing and the software in charge of controlling robots movements. Notice that software under testing and control software are totally independent, there are running on the same computer for practical reasons only.
For the mobile platform we use Lynxmotion 4WD rover. We selected it instead of other smaller robot (e.g. Lego Mindstorm) because this rover is able to carry a payload of 2 Kg during a few hours, running at a maximum speed of 1m/s. It is also relatively cheap (cf. table 2) and easy to build. We equipped it with infra-red proximity sensors to avoid collision, a top deck to support the laptop, a positioning system and a modiﬁed Wi-Fi interface.
The motion control software, running on the carried laptop, communicates speeds orders (linear speed and angular speed) to the robot. The mobility patterns are drawn by an operator for each mobile robot, using a dedicated software, that sends it to the mobile nodes control software. This enables ﬂexibility – each node has its own mobility pattern – and repeatability – a pattern can be saved and replayed.
Positioning is a critical point of the platform. Firstly, we need to reproduce the kind of information produced by actual market solutions such as GPS, pondered by our scale factor. Secondly, we need a precise and real-time position of the mobile node to allow an accurate motion control of the robot. Our speciﬁcations required a 6 Marc-Olivier Killijian, Matthieu Roy, and Gaetan Severac precision within the centimetre and a minimum refresh of 2 Hz. Several technologies are currently available for indoor location , mostly based either on scene analysis (e.g. using motion capture systems) or on triangulation (of RF and ultrasound  or wireless communication interfaces ). During the building of the platform, we tried four different solutions.
We ﬁrst tested the Cricket system , developed by MIT. Cricket is based on simultaneous ultrasound/RF messages and triangulation. Beacons ﬁxed on the celling send periodically RF message with their ID and position, and, in the time, they send an equivalent message by ultrasounds. The ﬂight time of the RF message is insigniﬁcant compare to that of the ultrasound. So the receiver, embedded on the mobile robot, can estimate the ﬂight time of the ultrasound messages and and calculate, with at least 3 different messages received, its position. This position is then send to the mobile node via a serial connection. In theory, this system is very efﬁcient, but in practice we were confronted to important limitations due to ultrasound disturbances. The ultrasound speed in the air can change depending of the temperature, so the results obtained can vary in the same way, and the ultrasound are also very sensitive to noise and perturbations. Neon lights was perturbing the systems and robots vibration, when they were moving, generated a lot of disturbances in the results. Finally we had to abandon this technology.