Enhancement of the AODV routing protocol for VANET based on intelligent link quality estimation

Imagine a world where your car doesn't just drive itself but also talks to every other car on the road, whispering secrets about traffic jams, icy patches, or sudden accidents. This isn't science fiction; it's the reality of Vehicular Ad Hoc Networks, or VANETs. Think of these cars as a giant, moving school of fish that constantly needs to share information to stay safe. But here's the catch: unlike fish in a calm pond, cars zip around at high speeds, turning corners, and getting blocked by buildings. This makes their "conversations" incredibly shaky. The signals they use to talk can fade, get interrupted, or get lost in the noise of a crowded street.

To keep these conversations going, the cars need a "map" or a set of rules to figure out the best path to send a message. In the digital world, this is called a routing protocol. The most common rulebook used today is called AODV. It's like a traveler who only looks at a map and asks, "Which path has the fewest stops?" (hop count). While this works okay in a quiet neighborhood, it's a disaster in a chaotic city. If the traveler picks a path with the fewest stops but those stops are on a crumbling bridge, the message falls into the river. The big question researchers are trying to solve is: How do we make these cars smarter so they don't just count stops, but actually check if the road ahead is safe and stable before they commit to the journey?

This is exactly where the researchers Hamida Ikhlef and Ali Melit step in with their new idea. They took the standard AODV rulebook and gave it a brain upgrade, creating a new system they call AODV-LSENN. Instead of just counting stops, their new protocol uses a "smart assistant" (a neural network) to peek at the physical conditions of the road. It checks things like how strong the radio signal is, how far apart the cars are, how fast they are moving relative to each other, and even if the digital message is getting garbled during transmission.

The researchers trained this smart assistant, which they named LSENN (Link State Estimation Neural Network), to act like a weather forecaster for the road. Instead of just saying "it's raining," it predicts if a connection is Connected (sunny and clear), Transient (cloudy and about to get stormy), or Disconnected (a total blackout). If the assistant predicts a storm is coming on a specific road, the car doesn't wait for the message to crash; it immediately switches to a different, safer route. It's the difference between a driver who waits until they hit a pothole to swerve, and one who sees the pothole from a mile away and steers around it.

To test if this "smart driver" actually works, the authors didn't put real cars on real roads. Instead, they built a massive, detailed video game simulation using a tool called NS-3. They created a virtual city with a grid of streets (like Manhattan) and filled it with anywhere from 10 to 80 virtual cars zooming around at speeds between 5 and 20 meters per second. They ran the simulation for 150 seconds, letting the cars try to send messages using both the old AODV method and their new AODV-LSENN method.

The results were quite telling, especially depending on how crowded the virtual city was. In a low-density scenario with just 10 cars, the new system was a clear winner. It managed to deliver 100% of the messages successfully, while the old system only got 95.03% through. The new system also cut the time it took for a message to travel from start to finish (end-to-end delay) down to just 12.6 milliseconds, compared to a sluggish 38.98 milliseconds for the old method.

As the city got busier, with 40 cars (a medium density), the gap widened even more. The new AODV-LSENN protocol delivered 83.87% of packets, while the old one struggled at 65.46%. The delay for the new system was 86.98 milliseconds, whereas the old system took a whopping 251.11 milliseconds. This happened because the smart assistant was successfully spotting unstable links and rerouting traffic before the messages got lost.

However, the story changes a bit when the city gets packed with 80 cars. In this high-density chaos, both systems started to struggle. The new system's success rate dropped to 52.61%, and the old one fell to 49.48%. The delay for both skyrocketed, reaching over 300 milliseconds. The researchers explain that this isn't because their smart assistant failed, but because the "air" itself became too crowded. With so many cars talking at once, the radio signals started crashing into each other (collisions) and the channel got saturated. It's like trying to have a conversation in a room where 80 people are all shouting at once; even the smartest person can't be heard clearly.

Despite the traffic jam, the new system still held its ground better than the old one. It consistently showed lower "jitter," which is the measure of how much the delivery time varies. In the 40-car scenario, the new system had a jitter of just 11.88 milliseconds, while the old one wobbled wildly at 203.68 milliseconds. This means the new protocol provides a much smoother, more predictable ride for data, which is crucial for safety messages that need to arrive on time, every time.

In short, the paper suggests that by adding a layer of intelligent prediction to the routing rules, we can make car-to-car communication significantly more reliable, especially when the roads aren't completely gridlocked. The authors found that while they can't magically fix the physics of radio waves in a super-crowded city, they can definitely help the cars navigate the chaos much better than before. The simulation proves that knowing when a road is about to break is just as important as knowing which road is the shortest.