e-Parking Meter Management System





































Failure Recovery:


Loss of Head Meter:


If the current head meter stops working or loses connectivity with the group, the other meters must be able to recover from this. The main challenge in this is that the current head meter is responsible for choosing the next head meter. It was implemented in this way to reduce the possibility of there being multiple head meters at the same time resulting in conflicts. If the group loses connectivity to the head meter, the link qualities of all meters connected to the head meter will begin to decrease multiplicatively and reaching zero in at most 5 update cycles. Once the link qualities start decreasing, the meters will form transient loops as they try to route to the head meter through each other. During this phase, the Hop Count of each of the meters in the loop would be increasing and update packets are sent whenever Hop Count changes, so there will be a burst in update transmissions at this point ensuring that the routing loop is transient. Once the Hop Count reaches a number greater than the number of meters in the group (which is pre-defined), they would have established that the head meter is no longer connected to the group. They would then enter the countdown phase as in initialization and whichever finishes the countdown first will be the new head meter and will announce this by sending a burst of updates.


Loss of Non-Head Meter:


If a meter loses connectivity with the group (because it stopped working or for some other reason), the link qualities associated with this meter will decrease rapidly with each cycle. In cases where a meter has multiple good links, data will usually not be lost. However, in cases where a meter has few alternative routes and the dead meter was in the best route, that route might still be the best route for the first few cycles. There is the possibility of data loss due to this if the meter does not have reliable links with any of its neighbors. After at most five cycles, the link quality will be 0 and the entry for the “dead” meter will be removed from the routing table, so data packets will not be sent to it.




This occurs when a group of meters get cut of from the group which contains the current head meter. This means no meter will have a route to the head meter resulting in a transient loop which will lead to an initialization countdown. Once a new head meter has been found, this fragmented group will function as a full-fledged group. It is not possible for a single meter to successfully form a fragment. A meter can only start the initialization countdown if there is a loop which results in a “count to infinity”. A single meter cannot count to infinity and therefore cannot form a fragment. This is advantageous as it might be better for an isolated meter to not send data at all rather than send to the central station every cycle, as this will result in rapid exhaustion of its battery.  At the central station, authorities will be able to see that they are not receiving data from that meter, allowing them to resolve whatever problem or obstruction is keeping the meter in isolation.




This will occur when a group that has previously formed two or more fragments recovers global connectivity. There will be two or more current head meters in this case. The recovery from this is innate. Once update packets traverse from one group to the other, one of the old head meter will be over ridden depending on the propagation of the packets. This will result in there being only a single head meter.