The era of Internet of things, which will include smart, connected vehicles, is rapidly approaching. It will open the door for new technologies that will have an important role in providing wireless connectivity between a wide variety of devices in the future.
This doctoral thesis is focused on organisation and optimisation of wireless vehicular ad-hoc networks in emerging intelligent transportation systems. The nature of highly dynamical vehicle movements present a major challenge in designing a reliable, stable and useful communication network that provides the required quality of service.
To become acquainted with the field of inter-vehicular communications, we researched this scientific area and reviewed the related standards. We concluded that the standardisation is not yet finished and is moving in the direction of a unified global standard. Due to different research and development cycles, different regions have standardised more or less the same services in a different manner, which is only now becoming apparent and presents a weakness for a quick and economical deployment. USA, Japan and Europe are now trying to unify the standards before mass deployment begins, and thus avoid unnecessary segmentation. It is expected that these standards will also be adopted by other countries that are not actively participating in the standardisation process.
Since the standardization does not deal with network optimization, but merely prescribes the limits for quality of services that must be provided, we focused on the field of vehicle clustering. Clustering is the process of grouping together similar elements, in our case vehicles, to form a cluster. They effectively lower the high vehicle dynamics and stabilize the network. Our research revealed more than 30 different clustering algorithms for vehicle clustering. Their advantages and disadvantages were then taken into account during the development of a new clustering protocol.
The vast majority of clustering protocols use location services, especially the satellite positioning system, to identify vehicles with similar movement patterns. This poses a risk for the clustering protocol reliability, because the location services are not available everywhere and their accuracy can vary significantly, e.g. in parking structures. To avoid these problems, we consciously decided against the use of location services and designed a vehicle interconnection metric that solely relies on the communication capabilities between vehicles.
The vehicle interconnection metric that we designed is based solely on the periodic beacon frames transmitted by all the vehicles. Vehicles continuously monitor the reception of these beacon frames from other nearby vehicles and identify those with a similar movement pattern. It is evident that only vehicles with similar movement patterns (position, direction and speed) can keep in touch continuously for a longer period of time. In case of erroneous beacon frame reception, the algorithm reacts quickly and strictly, therein handling the communication degradation between vehicles.
We designed a new vehicle clustering protocol that pursues the aim of improving the connectivity between vehicles. Each vehicle tends to associate with two cluster heads instead of just one, which is the common case with other clustering protocols. This way, longer connectivity losses are avoided during the reclustering phase, because the vehicle is quite probably connected to the other cluster head. The probability of full connectivity loss is lower if a vehicle is using two cluster heads instead of only one. The speciality of our clustering protocol is also the inverted working principle in which the default role of each vehicle is a cluster head, not a cluster member. The switch from one role to another can be done instantly because the vehicle interconnection metric always reflects the current state of the network.
A new vehicle movement simulation scenario between Ljubljana and Medvode was developed, which mimics the vehicle movement on a typical work day on that stretch of the road. The new protocol was tested using this new scenario in the ns-3 network simulator and SUMO vehicle movement simulator tandem. The simulations were performed both in normal as well as in increased vehicle densities, thus providing the insight into the behaviour of the protocol in different traffic scenarios.
The evaluation results of the new protocol, and indirectly also of the new vehicle interconnection metric, can be seen as very positive. In comparison with the clustering protocol MOBIC, connectivity between vehicles was increased and the formed clusters are more stable. The new protocol also has quite a low protocol overhead which makes it suitable for use in urban environments.
In this thesis, we have shown that unusual, non-conventional clustering approaches are both possible and effective. It turned out that very simple location service independent metrics can be used in all usage scenarios. We also used two cluster heads instead of only one, without compromising the network stability, and an inverted clustering principle with cluster head in the default role. The designed vehicle interconnection metric and clustering protocol are opening new opportunities and challenges for novel scientific research in this field.