การประเมินเปรียบเทียบแบบจำลองบน ns-3 และระบบทดสอบจริง : กรณีศึกษาของระบบแจ้งเตือนอุบัติการณ์ที่ใช้ VANET / เพียงพูน จักรแก้ว = Comparative evaluation of NS-3 simulator and testbed platform : case study of incident warning system using VANET / Piangpoon Jakkaew
This thesis is proposed with the experimental simulation and testbed of incident warning system for vehicles with a vehicular ad-hoc network (VANET) capability. Based on the ns-3 platform, a new code has been developed to simulate and emulate the mes- sage forwarding of incident warning system application. A warning message dissemination mechanism is a simple flooding protocol. The developed code is located on the layer of incident warning application upon the protocol stacks of UDP and IP in the Linux operating system. First experiments have been reported on an ad hoc network testbed with four nodes each with an IEEE 802.11b/g wireless interface card. The testbed consists of 2-hop ad hoc scenario. Message delivery ratio and message receiving delay have been reported in the scenarios with and without interference effects. Despite of small testbed size, the obtained results suggest interesting findings that can only be confirmed by the testbed, but not from simulations because using simple parameter settings in the simulation cannot reflect the real environment. Second experiments have been reported on VANET with three vehicle nodes with a constant velocity. The testbed consists of 2-hop ad hoc scenario with interference effects within the university area used in the test. Values of transmission power and prop-agation loss model in the latter simulation have been found in this experiment to reflect the specific environment of testbed. In particular, these values have later been used in the large network-size simulation of 250 nodes with a constan tvelocity. In this final simulation, which cannot inevitably be replicated by the actual testbed measurements, the focus is on comparing results between the simple setting using log distance propagation loss model and specific environment setting. Based on two settings, trend of the results are same in almost scenarios. But the values of the resultant delay in reaching the nodes nearby of specific environment setting are higher than simple setting. The results also suggest that if we send incident warning messages more frequently, then the message receiving delay of each node will be higher and the delivery success ratio will be lower. This finding therefore strongly reflects the necessary trade-off that, in practice, must be carefully calibrated by choosing an optimal rate of incident warning message generation. Future research into that optimization is expected to warrant a worthy fruitful implication.