Skip to content

Lab 3 — VANET: buildings, corners and reconvergence

Goal. See why cars in a city lose routes, and why fixed infrastructure helps. Then read the measured gap between AntHocNet and a protocol that knows the topology. Time. 60 min. Site. Campaign levels 3.1 and 3.4 (#campaign?open=all).

← Classroom pack · Answer key

Background (5 min)

In a street grid, buildings block the radio at every corner. Two cars a few metres apart lose their link the moment one turns into a side street. A route that worked a second ago is gone, and the protocol must find another. This is reconvergence, and it is VANET's largest measured loss (vanet.md).

1. Predict (5 min)

  1. Level 3.1 has four cars and one call. What ends the call: distance or something else?
  2. Level 3.4 has sixteen cars, and ten of them park and switch off at 60 s. Sketch the delivery you expect over 160 s, without any roadside units.
  3. Now sketch it with five roadside units: four near the corners and one at the centre.

2. Run (25 min)

  1. Open the classroom link and play 3.1 Around the corner. Follow car 1 and watch its links as it reaches a corner. Answer the quiz.
  2. Play 3.4 Traffic jam without placing anything: press Run. Every 10 s, note the delivery percentage shown for each call.
  3. Retry 3.4. Place roadside units at (200, 200), (600, 200), (200, 600), (600, 600) and (400, 400), then run it again and note the same numbers.

3. Explain (15 min)

  1. In 3.1, why does the call drop although the cars stay well inside radio range?
  2. In 3.4, compare delivery before and after 60 s, with and without the units. What do fixed units give the ants that cars cannot?
  3. Even with units, delivery after 60 s is lower than before. Why?

4. Compare with ns-3 (10 min)

On the ns-3 VANET main cell (20 seeds, 900 s), AntHocNet delivers 42.1 %. An oracle that always knows the shortest current path delivers 81.7 % on the same field (vanet.md). 1. The oracle shows the network is connected most of the time. Where does the other ~40 % go? 2. Link-lifetime prediction (#574, the Look-ahead upgrade) discounts a next hop whose link is about to break. Why could a prediction based only on distance and speed miss a break at a corner?