The significance of addressing reliability issue associated with Cooperative-approach is critical as it affects greatly the decision-making process. To this end, without actual publication, we had to account the missing entries in cache for those vehicles which refrained from publishing. In the following section we discuss the design of adaptive caching to overcome this issue.
5.2.1 Ensuring Reliability with Virtual Cache Entries
In order to account in cache those vehicles that were prevented from publishing to avoid redundant publications, we propose adding virtual cache entries. This idea is based on Redundant Aggregation, commonly used in information fusion for wireless sensor networks to avoid network costs. Accordingly, as mentioned in the work of [MLN+17], instead of transmitting similar information, aggregation of those similar information can be done to avoid network costs while maintaining accuracy and reliability. To this end, each of the aggregated information have same values for attributes as that of the latest information that was transmitted .
In our work, the virtual cache entries are added based on vehicular rate. Vehicular Rate is the number of vehicles moving through a point in some duration, expressed in time units. As discussed in Subsec- tion 4.4.1, a publication contains the publisher identifier with the published time, publishing location, published value and associated TTL, if any. Apart from these basic publication attributes, we propose to also include the observed vehicular rate by the publisher during the time of publishing.
In VANETs with 802.11p and beacon messages, a vehicle can opportunistically measure the number of vehicles around it. Based on the number of beacon messages that a vehicle processes from other vehicles around it, a vehicular rate per time can be obtained. With this rate now known to publisher, it is also
included in the publication as an attribute to support Adaptive Cache Management implementing virtual entries.
Figure 5.3:Adaptive Cache Management using virtual cache entries.
Adaptive Cache Management: The process of adding virtual entry is as shown in Figure 5.3. If a
notification is explicitly received via cellular communication, it is simply added to the cache as shown in Block-A in figure. Therefore, without any virtual entries added, the cache would simply contain all the notifications received in the form of subscription.
The flow for adding virtual entry starts provided at least one entry reflecting the original notification exists in cache. In other words, virtual entries are added only if the vehicle is subscribed and has received atleast one notification that is still valid in cache. The reason for this assumption is because the required
vehicular rate is sent only with original notification, and to add virtual entries, this rate can only be obtained provided the original notification still exists in cache. As any expired cache entries are removed automatically, it is assured that no virtual entries with expired TTL are added.
Virtual entries are calculated for each original notification that are valid in cache. Hence, during the iteration through each of the received notifications that are valid, virtual entries would be added with a published time starting immediately after the published time of the current original notification (marked as Ti).
Based on vehicular rate obtained at (i) , marked as [VR(i) ] (i.e., vehicular rate observed when this original notification was published by publisher), the published time for each of the virtual entry added can be uniformly distributed. The published time of a virtual entry (marked as TVE) is a summation of the published time of current original notification and the vehicular rate obtained at (i) .
It is given as: TVE= Ti+ VR(i) .
Subsequently, for the next virtual entry to be added, its published time would be incremented by vehic- ular rate. That is: TVE= TVE+ [VR(i)] . This way all the virtual entries added for a notification would have different published times distributed uniformly based on vehicular rate.
Furthermore, these virtual entries are replicated with the attributes and their corresponding values from current original notification (Ti) as followed in Redundant Aggregation. Hence they all have the same published location and published value as that of the original notification except for different published time. To this end, virtual entries for current notification are added until a time (marked as Tmax), which is either the published time of the next-received-notification entry in cache (marked as Ti+1) or until current-time (marked as Tcurr). In the former case, the loop for adding virtual entries will continue with the same procedure with the following next-received-notification entry, whereas in later case, it will terminate. This process is shown in Block-B.
Block-B signifies the number of virtual entries added for each original notification which is dependent on Tmax and vehicular rate. Current-time signifies the present system time at which a vehicle at AOI wants to make rerouting decision based on its cached contents. Hence, any virtual entries, if added, should be only until this time.
5.2.2 Challenges of Varying Vehicular Rate
In real-world, using 802.11p and beacon messages in VANETs, vehicles can measure vehicular rate around its surrounding. But the communication range of these technologies is limited to couple of hundred of meters. Due to this, rate observed by a publishing vehicle accounts only for vehicles coming behind it that were detected by its limited communication range during its publishing time.
As virtual entries are being added based on this rate until next actual publication (which can be at times only after long time), change of vehicular rate until next publication, if witnessed any, is neglected. Consideration this practically probable scenario is again crucial for ensuring reliability during rerouting decisions.
In Figure 5.2, only Vehicle-1 publishes the road status with a Traffic-Jam value. Lets assume that it also sends observed vehicular rate accounting other three vehicles coming behind it in the publication. Subsequently, using Adaptive Cache Management, this rate is considered for adding virtual entries until next publication. However, now suppose, this vehicular rate decreases significantly almost to zero, and for the worst case the next publication is only after long time, then accordingly, this approach is still adding more virtual entries than it is supposed to be. Ideally, in this case, the number of virtual entries added should have been lessened. However, since outdated rate is still considered, though no vehicles are passing through POI, vehicles at AOI, accounted falsely by adding more virtual entries. Given the possibility of varying vehicular rate between successive publications and subsequently to incorporate it, in the next section we discuss how we measure this rate more accurately.