(PDF) A hybrid and flexible discovery algorithm for wireless sensor networks with mobile elements

A Hybrid and Flexible Discovery Algorithm for Wireless Sensor Networks with Mobile Elements Koteswararao Kondepu , Francesco Restuccia2,3, Giuseppe Anastasi2, Marco Conti3 1 1 2 3 Dept. of Computer Science & Engineering Dept. of Information Engineering IIT-CNR IMT Institute, Lucca, Italy University of Pisa, Italy National Research Council, Italy

{g.anastasi, f.restuccia}@iet.unipi.it

Abstract—In sparse wireless sensor networks, data collection is have very different mobility patterns, ranging from carried out through specialized mobile nodes that visit sensor deterministic [5] to random mobility [6]. nodes, gather data, and transport them to the sink node. Since In WSNs with MEs (hereafter, WSN-MEs for shortness), visit times are typically unpredictable, one of the main the communication between sensor nodes and MEs is challenges to be faced in this kind of networks is the energy- opportunistic, i.e., they can exchange data only when they efficient discovery of mobile collector nodes by sensor nodes. In are in the communication range of each other. In principle, a this paper, we propose an adaptive discovery algorithm that sensor node could be always in sleep mode and wake up combines a learning-based approach with a hierarchical only for communication. In practice, unless the ME’s motion scheme. Thanks to its hybrid nature, the proposed algorithm is is deterministic, the sensor node cannot know in advance very flexible, as it can adapt to very different mobility patterns of the mobile collector node(s), ranging from deterministic to when the ME will enter its communication range. Hence, a completely random mobility. We have investigated the discovery protocol is used for detecting the presence of the performance of the proposed approach, through simulation, ME [3]. Discovery algorithms commonly used in WSN-MEs and we have compared it with existing adaptive algorithms that are based on periodic listening. In detail, the ME emits only leverage either a learning-based or a hierarchical periodic beacons to announce its presence in the area, while approach. Our results show that the proposed hybrid algorithm sensor nodes wake up periodically (and for a short time) to outperforms the considered adaptive approaches in all the listen for possible beacons. The period between two analyzed scenarios. consecutive activations (wakeup period) of the sensor node should be as long as possible to minimize the energy Keywords: Wireless Sensor Networks, Sparse Sensor consumed during the discovery phase. On the other side, Networks, Mobile Node Discovery, Energy Efficiency. using a too long activation period could compromise the effectiveness of the discovery process, i.e., contacts could be I. INTRODUCTION missed or detected very late (thus, leaving a short time A Wireless Sensor Network (WSN) typically consists of available for data communication). a large number of sensor nodes, densely deployed over a Typically, discovery protocols use fixed parameters (i.e., geographical area. Sensor nodes are tiny devices that can constant wakeup period and/or beacon emission rate [2, acquire data from the surrounding environment, process 4,6]). Better performance, in terms of energy efficiency, can them locally, and/or transfer them to a collection point (sink be achieved through adaptive schemes based on learning node) using multi-hop communication [1]. However, many techniques (for predicting the arrival time of the ME) [7,8], real-life applications do not require a fine-grain sensing that or relying on a hierarchical approach [9,10,11]. Learning- necessitates such a dense deployment. Hence, a sparse based algorithms are very well-suited when the ME’s motion topology can be used where sensor nodes are located at has some regularity that can be learned and exploited to some strategic locations and the distance between predict the next arrival time with a certain accuracy. neighboring nodes is typically much larger than their However, they are unsuitable when the ME moves without transmission range. In a sparse sensor network multi-hop any regular pattern. Hierarchical discovery protocols communication is unfeasible, and data collection is carried typically requires two different radios (namely, a wake-up out through Mobile Elements (MEs), i.e., special mobile radio and a data radio), which are not available in most of nodes that visit sensor nodes regularly, gather data, and the existing sensor platforms. In addition, they are not able transport them to the sink node1 [2, 3]. MEs can also be used to learn and exploit information about the specific mobility in dense sensor networks to allow a more uniform pattern of the ME. distribution of energy consumption among sensor nodes, In this paper, we propose a hybrid discovery protocol thus increasing the network lifetime [4]. Depending on the (hereafter referred to as Hybrid) that combines both a application scenario, MEs can be either part of the external learning-based approach and a hierarchical scheme. The environment (e.g., persons, cars, buses), or part of the proposed protocol is thus very flexible and can adapt to very system infrastructure (e.g., mobile robots). Also, they can different mobility scenarios. Unlike other hierarchical approaches, it does not require two different radios and, hence, it can be implemented in any sensor platform. We 1 have evaluated the proposed protocol, by simulation, and As a special case, the sink node itself can be mobile and play the role of mobile data collector. compared it with other adaptive discovery schemes. The obtained results show that our hybrid approach outperforms the subsequent one. The reward function ρ provides the existing adaptive schemes that only leverage either a immediate reward achieved by executing a task. It is learning-based approach or a hierarchical scheme. positive if a success has been obtained, and negative The rest of the paper is organized as follows. Section II otherwise. Instead, the utility function gives the long-term presents the Hybrid protocol. Section III describes the utility of performing a task. Q is an utility look-up table simulation environment used for our analysis. Section IV whose generic element Q(s, τ) provides the utility of compares the proposed protocol with other adaptive performing task τ in state s. It is defined as the expected protocols. Finally, Section V concludes the paper. value of the sum of the immediate reward ρ and the II. HYBRID DISCOVERY PROTOCOL discounted utility of state s′ resulting from execution of task τ, i.e., As mentioned before, the proposed discovery protocol Q (s,τ )= E [ρ + γ ⋅ e(s ′) s,τ ] (1) combines a learning-based approach with a hierarchical scheme. Specifically, it tries to learn the mobility pattern of where e(s′) = maxτ Q(s′, τ). The expected value in equation the ME and predict the next arrival time, on the basis of the (1) is conditioned to state s and task τ. Since Q-learning is past history, using Q-Learning [12], i.e., a form of done online, equation (1) cannot be applied directly, as the reinforcement learning that does not require a model of the stored utility values may not have converged yet to their environment. The duty cycle of the sensor node is then final values. In practice, Q-learning is used with incremental adjusted according to this prediction. Hence, the sensor node updates as given by the following equation: is in sleep mode for most of the time, and activates only Q (s,τ )= (1 − α ) ⋅Q (s,τ )+α ⋅ [ρ + γ ⋅ e(s ′)] (2) when the ME is about to arrive. Since the prediction may not In equation (2), α is a learning-rate parameter between 0 be accurate, the Hybrid algorithm exploits an additional and 1, that controls the rate at which a sensor node tries to hierarchical scheme to increase its energy efficiency. The learn by giving more (α close to 1) or less (α close to 0) sensor node initially activates with a low duty cycle and weight to the previously learned utility value. Furthermore, γ switches to a higher duty cycle only when the ME is actually is a discount-factor, also between 0 to 1; the higher the nearby. Information about the physical location of the ME is value, the greater the sensor node relies on future reward made available to the sensor node by the ME itself by using rather than on immediate reward. Time is divided into time two different Beacon messages, namely Short Range domains (of fixed duration TD ), and the utility function is Beacons (SRBs) and Long Range Beacons (LRBs). LRBs updated periodically, at the end of each time domain. Then, and SRBs are transmitted in an interleaved way, both with based on the learned utility, the task that maximizes the the same period (i.e., 2 ⋅ TBI , so that the overall beaconing long-term utility is selected for execution in the next future. period is TBI ), but with different transmission power, and As any other learning algorithm, Hybrid includes both an convey different information. SRBs are transmitted with the exploitation and an exploration phase. During the same transmission-power level used during the exploitation phase, the next task is selected according to the communication phase for data exchange. They experience a learned utility (as described above), while in the exploration transmission range r – hereafter referred to as phase it is picked up randomly from the set of available communication range – and are aimed at notifying the tasks. The exploration phase is accessed, at the end of a time sensor node that the ME is within its transmission range and domain, with a probability ε evolving dynamically as data exchange can, thus, take place. Instead, LRBs are sent  (ε max − ε min ) ⋅ (cmax − c )  with a higher transmission power. Therefore, they have a ε = ε min + max0,  (3) transmission range R larger than the communication range r  c max  – throughout R will be referred to as the discovery range – where ε min ( ε max ) is the minimum (maximum) exploration and are used to inform the sensor node that the ME is probability, while c and cmax denote the number of contacts approaching and a contact could potentially occur shortly. detected by the sensor node when equation (3) is evaluated, As mentioned before, the prediction algorithm is based and the maximum number of detected contacts to be on Q-Learning. Specifically, like RADA [8], it follows the Distributed Independent Reinforcement Learning (DIRL) considered for calculating ε. Finally, there is a third phase approach [13], and relies on the following elements: (i) a (namely activation phase), triggered by the reception of a state representation consisting of both system and LRB from the ME, during which the next task to execute is application variables, (ii) a set of tasks (i.e. duty cycles) that chosen deterministically (see below). Although the definition of tasks is strictly related to the specific can be executed by the sensor node, (iii) a reward function application, in Hybrid we defined the following tasks (i.e., ρ associated with each task, and (iv) a utility function Q. duty cycles). The objective of the system is to maximize the long-term • Sleep (SLP). The sensor node keeps the radio in sleep utility that can be achieved by executing the different tasks. mode. This task is selected whenever the ME is not In our system, the state s corresponds to the inter-contact expected to arrive, based on the learned utility. time, as observed by the sensor node, i.e., the time elapsed • Low Duty Cycle (LDC). The sensor node operates with a from the beginning of a certain contact to the beginning of low duty cycle δ L . This task is selected when the ME is expected to arrive in the next time domain, based on the then receives a SRB) or a false activation (if it fails to learned utility. receive a subsequent SRB). To avoid energy wastes due to • High Duty-Cycle (HDC). The sensor node operates with false activations a timer is used. The timeout value Tout is set a high duty cycle δH . Unlike the other tasks, HDC is not according to the worst case, i.e., when the distance between selected on the basis of the learned utilities. It is chosen the sensor node and the ME is zero. Finally, an initial task is whenever a LRB is received from the ME (thus, starting randomly selected from set Λ . the activation phase). TABLE 1. REWARD FUNCTION’S PARAMETERS. Algorithm 1: Hybrid algorithm LRB SRB nc pm ep init s = 0; Q (0, τ) = 0 for all τ; NO NO 0 -1 100 NO YES 1 1 100 Λ = { SLP, LDC, HDC}; YES YES 1 2 100 LRB-rcvd =False; SRB-rcvd=False; YES NO 0 -2 100 Tout = (R + r ) v ; Select an initial task τ from Λ randomly; At each step, the algorithm executes the previously selected end init task, until one of the following events occurs: (i) LRB loop reception; (ii) SRB reception; (iii) timeout expiration; (iv) execute τ; end of the communication phase, and (v) end of current time wait (event); domain. Upon receiving a LRB (case i) the sensor node sets switch (event) { the LRB-rcvd flag and selects the HDC task; finally, the false case (LRB reception): activation timer is started. If the latter timer expires without LRB-rcvd = True; receiving any SRB (case ii), the sensor node selects the LDC τ = HDC; start timer( Tout ); as the next task and resets the LRB-rcvd variable. Instead, if case (timeout): a SRB is received before the timeout expiration (case iii), LRB-rcvd = False; the sensor node sets the SRB-rcvd flag, stops the false τ = LDC; activation timer, and enters the communication phase. At the case (SRB reception): end of the communication phase (case iv), both the LRB- SRB-rcvd=True: stop timer; rcvd and SRB-rcvd variables are reset. Finally, at the end of Start communication phase; case (end of communication): a time domain (case iv), if the communication phase is in SRB-rcvd = False; progress (i.e., a SRB has been received), no action is LRB-rcvd = False; performed. If a LRB has been received (i.e., the sensor node τ = LDC; is inside the activation phase), HDC is maintained as the case (end of time domain): next task. Otherwise, the new resulting state s' (i.e., inter- if SRB-rcvd =False { contact time) is measured. If s′ is similar to a state if (LRB-rcvd =True) τ = HDC; s ′′ previously stored in the Q structure (i.e., the Hamming else { distance between s ′ and s ′′ is less than a pre-defined Calculate new state s ′ ; threshold [13]), s′ is assimilated to s ′′ , otherwise s ′ is added if ∃s ′′ : s ′ ≈ s ′′ then s ′ = s ′′ to the list of known states. Finally, the reward for task τ else add s′ to the list of known states; corresponding to state s is calculated, and Q ( s, τ) is updated Calculate reward for task τ in state s; accordingly. Specifically, the reward for any task is Update Q (s, τ) ; calculated as ρ = (nc ⋅ p m ⋅ e p − 1)⋅ es , where nc , p m and choose a new task τ to execute e p denote the number of contacts detected in the last time // through exploration (with prob. ε ) or exploitation } domain (i.e., 0 or 1), the price multiplier for the executed } // end if task, and the expected price, respectively. The negative part } // end switch of the reward represents the cost for executing the task. This end loop cost depends on to the time during which the sensor node was active during the last time domain (for instance, for the Algorithm 1 shows the actions performed by the sensor LDC task e s = δ L ⋅ T D ⋅ PRX + (1 − δ L ) ⋅ T D ⋅ PSL , where node. Initially, the algorithm initializes the look-up table Q PRX and PSL denote the power consumption in receive mode and the set Λ of tasks that can be selected during the and sleep mode, respectively). The reason behind using a exploration phase (i.e., SLP, LDC, HDC). Boolean variables price multiplier and an expected price is to allow a LRB-rcvd and SRB-rcvd are initialized to False. LRB-rcvd symmetric evaluation of the reward function. Thus, for each (SRB-rcvd) will be set when a LRB (SRB) is received, thus task, the reward is positive if the ME is successfully starting the activation (communication) phase. A node that detected. If the ME is not detected, the reward is negative has received a LRB may experience either a contact (if it (equal to minus es ). The price multiplier p m for task τ is In all the experiments, we performed 15 independent calculated as shown in Table I. replications, each consisting of 1000 visits of the ME to the III. SIMULATION ENVIRONMENT sensor node, and derived confidence intervals with a 90% confidence level. Since we are mainly interested in the To evaluate the performance of the proposed Hybrid discovery process, the channel quality was modeled using protocol, we used the OMNeT++ simulation tool [14]. the disk model, i.e., packet loss is assumed to be 0% when Without losing in generality, we considered a single ME the distance between sensor node and ME is lower than the moving with a constant speed v along a straight line at a fixed distance D from the sensor node. We focused on a communication range r, and 100% otherwise. Unless stated sparse scenario, i.e., we assumed that the distance between otherwise, all the other simulation parameters are as shown neighboring sensor nodes is very large (i.e., larger than the in Table 1. The learning parameters are set as in [8], while discovery range R), so that we can concentrate on a single the power consumption values are derived from the sensor node (the evaluation in a dense scenario can be found datasheet of the Chipcon CC2420 radio transceiver [16]. in [15]). In our analysis we measured the following TABLE 3. SIMULATION PARAMETERS. performance indices. Parameter Value • Discovery Ratio, defined as the ratio between the number LRB/SRB period (2TBI, Hybrid and 2BD ) 200 ms of contacts successfully detected by the sensor node and Beacon period (TBI, RADA and Fixed) 100 ms the total number of potential contacts. Beacon duration (TBD, all) 1 ms • Activity Ratio, defined as the ratio between the total ME Speed (v) 40 Km/h active time (i.e. when the radio is on) and the overall Distance from the sensor node (D) 15 m time spent during the discovery phase. Communication range (r) 50 m • Energy per Contact, defined as the average energy Nominal contact time 8.6 s Discovery range (R, Hybrid and 2BD) 200m consumed by sensor nodes in the discovery phase per Power Consumption in Receive Mode (PRX) 56.4 mW detected contact. Power Consumption in Sleep Mode (PSL) 0.6 µW The Discovery Ratio provides a measure of the Time Domain (TD) 100 s effectiveness of a discovery scheme. Ideally, this index α (Hybrid and RADA) 0.5 should be (close to) 100%. The activity ratio and the energy γ (Hybrid and RADA) 0.5 per contact measure the energy efficiency of the discovery ε max , ε min (Hybrid and RADA) 0.5, 0.05 scheme. Ideally, these indices should be as low as possible. cmax (Hybrid and RADA) 100 To better evaluate the performance of Hybrid, we compared it with the following adaptive solutions that exploit either a IV. SIMULATION RESULTS learning-based or a hierarchical approach. • RADA [8]. This protocol relies on the same prediction A. Impact of the ME mobility pattern algorithm used in Hybrid, however, it does not exploit In our analysis we assumed that the ME visits the sensor the hierarchical mechanism based on LRBs and SRBs. node at regular times, on average every TTOUR (inter-arrival Sensor nodes are typically in sleep mode and activate time). We considered four different ME mobility patters only when the ME is expected to arrive. resulting in a corresponding number of scenarios with • 2BD [11]. This protocol uses a hierarchical approach increasing randomness in the inter-arrival time. based on LRBs and SRBs, but it is not able to predict the • Deterministic: ME arrivals are periodic. The inter-arrival ME’s arrival time. Sensor nodes are always in LDC and time is fixed and equal to 30 min (1800s). switch to HDC upon receiving a LRB. • Gaussian-1: The inter-arrival time is a random variable, For completeness, we also considered a fixed scheme distributed according to a normal distribution, with mean (hereafter referred to as Fixed) where the duty cycle is equal to 30 min and standard deviation equal to 1 min. constant over time and equal to HDC. Table 2 shows the • Gaussian-10: As above, but with standard deviation duty cycle values used by the different protocols. To make equal 10 min. the comparison fair, we considered the same values of HDC • Uniform: The inter-arrival time is uniformly distributed and LDC for the various algorithms. Also, we used the same in [0, 30] min. set of duty cycles for Hybrid and RADA (i.e., HDC, LDC, Figures 1-3 show the impact of the ME's mobility pattern SLP). on the different discovery schemes, in terms of discovery TABLE 2. DUTY CYCLE VALUES ratio, activity ratio, and energy consumed per detected contact, respectively. When the mobility pattern is Algorithm HDC LDC SLP deterministic, all schemes exhibit a discovery ratio close to HYBRID 3% 0.5% 0% RADA 3% 0.5% 0% 100%. However, the activity ratio of Hybrid is significantly 2BD 3% 0.5% - lower than that of the other schemes, resulting in a lower Fixed 3% - - energy spent per detected contact. Figure 1. Impact of the mobility pattern in terms of discovery ratio. Figure 4. Impact of inter-arrival time terms of energy per contact. TABLE 4. ENERGY SAVINGS PROVIDED BY HYBRID. Inter-arrival 2BD RADA Time (s) 300 27.9% 81.5% 600 59.2% 85.2% 1800 79.1% 91.0% 2400 82.6% 92.1% B. Impact of the inter-arrival time In Figure 3 the energy per contact has been calculated assuming an inter-arrival time of 1800s (i.e., 30 minutes). Obviously, the energy consumption is strongly influenced by this value. To investigate the impact of the inter-arrival time Figure 2. Impact of the mobility pattern in terms of activity ratio. on the average energy consumed per detected contact, we considered different values for this parameter (while leaving all the other parameters unchanged). The obtained results, in terms of energy per detected contact, are summarized in Figure 4. As expected, the difference in the energy consumption of the Hybrid scheme, with respect to the other schemes, increases with the inter-arrival time. This is because the higher the inter-arrival time, the longer the time the sensor node spends in the discovery phase. Table 4, which shows the energy savings provided by Hybrid, with respect to 2BD and RADA, better emphasizes the benefits of using the proposed approach. C. Impact of the Discovery Range Figure 3. Impact of the mobility pattern in terms of energy per contact. Both 2BD and Hybrid use a hierarchical mechanism based on LRBs and SRBs; LRBs are transmitted with a When the randomness of the inter-arrival time increases, the transmission range R, larger than the transmission range activity ratio of all the adaptive schemes is approximately used for SRBs. It is thus extremely important to evaluate the the same (i.e., around 0.5%), meaning that the sensor node is impact of the R parameter on their perfomance. To this end, in LDC for most of the time. However, Hybrid outperforms we considered three differerent values for R (i.e., 150m, both 2BD and RADA in terms of percentage of detected 200m, 250m). The obtained results are shown in Figures 5- contacts. Hence, it experiences a lower energy per contact. 7. Fixed and RADA are not influenced by this parameter as Specifically, Hybrid provides a discovery ratio very close to they use a single beacon type. They have been included in that of Fixed in all the considered mobility scenarios with an the plots just for comparison. From our results it clearly activity ratio of about 1/6 (in the worst case), thus achiving a emerges that increasing R increases the probability of huge reduction in energy consumption. detecting a pontential contact. However, after a given values For the sake of space, hereafter we will consider only the (200m in the considered scenario), a further increase in the Gaussian-1 scenario. R value does not produce any significant advantage in terms of discovery ratio, while it increases the energy consumption. This behavior is better emphasized by the based approach with a hierarchical scheme. We have activity ratio, which tends to increase with R. This is because investigated the performance of the proposed algorithm, the sensor node remains in HDC for a time proportional to through simulation, in a sparse scenario. Our results show R. Finally, we can observe that Hybrid slightly outperforms that, it can adapt to very different mobility patterns of the 2BD for all the considered R values, in terms of activity ME and, in comparison with other adaptive discovery ratio. This is because Hybrid can also exploit the prediction algorithms, it allows a very large energy saving, especially algorithm which puts the radio in sleep mode (selecting the when sensor nodes spend a long time in the discovery phase. SLP task) when there is a low probabity to receive a LRB, For the sake of space, we only focused on a sparse scenario. based on the learning utility. The analysis in a dense scenario can be found in [16]. REFERENCES [1] I.F. Akyildiz, W. Su, Y. Sankarasubramaniam and E. Cayirci, “Wireless Sensor Networks: a Survey”, Computer Networks, Vol.38, N. 4, March 2002. [2] S. Jain, R. Shah, W. Brunette, G. Borriello, S. Roy, “Exploiting Mobility for Energy Efficient Data Collection in Wireless Sensor Networks”, Mobile Networks and Applications, Vol. 11, No. 3, 2006. [3] M. Di Francesco, S. Das, G. Anastasi, “Data Collection in Wireless Sensor Networks with Mobile Elements: A Survey”, ACM Transactions on Sensor Networks, Vol. 8, N.1, August 2011. [4] A. Somasundara, A. Kansal, D. Jea, D. Estrin, M. Srivastava, “Controllably Mobile Infrastructure for Low Energy Embedded Networks”, IEEE Trans. on Mobile Computing, Vol. 5, N. 8, 2006. Figure 5. Impact of the Discovery Range on the discovery ratio. [5] A. Chakrabarti, A. Sabharwal, B. Aazhang, “Using Predictable Observer Mobility for Power Efficient Design of Sensor Networks. Proc. International Workshop on Information Processing in Sensor Networks (IPSN 2003). [6] R. Mathew, M. Younis, S. Elsharkawy “Energy-Efficient Bootstrapping Protocol for Wireless Sensor Network”, Innovations in Systems and Software Engineering, Vol. 1, No 2, Sept. 2005. [7] V. Dyo, C. Mascolo, “Efficient Node Discovery in Mobile Wireless Sensor Networks”, Lecture Notes in Computer Science. Proc. DCOSS 2008 (LNCS 5067). Springer, Heidelberg (2008). [8] K. Shah, M. Di Francesco, G. Anastasi, M. Kumar, “A Framework for Resource Aware Data Accumulation in Sparse Wireless Sensor Networks”, Computer Communications, Vol. 34, N. 17, Nov. 2011. [9] J. Brown, J. Finney, C. Efstratiou, B. Green,N. Davies, M. Lowton, G. Kortuem, “Network Interrupts: Supporting Delay Sensitive Applications in Low Power Wireless Control Networks”, Proc. ACM Figure 6. Impact of the Discovery Range on the activity ratio. Workshop on Challenged Networks (CHANTS 2007), Montreal, Canada, 2007. [10] H. Jun, M. Ammar, M. Corner, E. Zegura, “Hierarchical Power Management in Disruption Tolerant Networks Unsing Traffic-aware Optimization”, Computer Communications, Vol. 32, N. 16, 2009. [11] K. Kondepu, G. Anastasi, M. Conti, “Dual-Beacon Mobile-Node Discovery in Sparse Wireless Sensor Networks”, Proc. IEEE Symposium on Computers and Communications (ISCC 2011), Corfu, Greece, June 28 – July 1, 2011. [12] R. Sutton, “Temporal Credit Assignment in Reinforcement Learning”, Dept. of Computer Science, Univ. of Massachusetts, Amherst, USA, COINS Technical Report 84-2 (1984). [13] K. Shah, M. Kumar, “Distributed Independent Reinforcement Learning (DIRL) Approach to Resource Management in Wireless Sensor Networks”, Proc. IEEE Conference on Mobile Ad hoc and Sensor Systems (MASS 2007), Pisa, Italy, 2007. [14] The OMNeT++ Network Simulator. http://www.omnetpp.org. Figure 7. Impact of the Discovery Range on the energy per contact. [15] K. Kondepu, F. Restuccia, G. Anastasi, M. Conti, “A Hybrid Discovery Algorithm for WSNs with Mobile Elements (Extended V. CONCLUSIONS Version), Technical Report DII-2012-4, Univ. of Pisa, April 2012. http:www.iet.unipi.it/~anastasi/papers/tr-2012-4.pdf In this paper, we have proposed an adaptive discovery [16] Chipcon, 2.4 GHz IEEE 802.15.4/ZigBee-Ready RF Transceiver, algorithm for WSNs with MEs, which combines a learning- Chipcon Products from Texas Instruments, 2004, CC2420 Data Sheet.