A New Class of Mobility Models for Ad Hoc Wireless Networks

1. A New Class of Mobility Models for Ad Hoc Wireless Networks Rahul Amin Advisor: Dr. Carl Baum Clemson University SURE 2006

2. Brief Overview • Background on Random Waypoint Model • Description of New Model • Observations and Conclusions • Future Work

3. Random Waypoint Model • Choose a random point (waypoint) distributed uniformly over some area • Choose a random velocity and move from current waypoint to the next using this velocity 0 2 v1 v2 v3 1 3

4. Motivation • Random Waypoint Model is too idealistic • Nodes can move freely without restrictions • Model a more real-world scenario • Obstructions in mobility • Obstructions in communication

5. New Model Description • Outer circle radius fixed at 1000 m • Inner circle represents obstruction and its radius can be varied • Obstruction can affect mobility as well as communication • Constant velocity model used (10 m/sec) • Distribution sampled every 1 sec

6. New Model Description (contd.) • Waypoint is described by Radius (R) and Angle (Θ) • R and Θ are independent • Generate a Uniform Random Variable in (0,1) interval using Mersenne Twister algorithm

7. New Model – Boundary Prevention • Node smartly predicts if it is going to collide with the obstruction • To prevent collision, the waypoint is discarded and a new waypoint is generated 0 3 2 1 3 2

8. Collision Prediction Calculations

9. Generated Waypoint Efficiency • Efficiency decreases as the radius of obstruction is increased • Acceptable efficiency – not going to slow simulation drastically

10. Steady State Density • Peak value shifts right as obstruction radius is increased • Close to being spatially uniform

11. Network Partitions • Partition • The inability of any one node to be able to connect to any other node for a given distribution • Spanning Tree • Tree that spans every node in the distribution without forming loops • Kruskal’s Minimum Spanning Tree Algorithm used to study network partitions

12. Network Partitions - Mobility Blocking, No Communication Blocking • The maximum hop distance used was ½R = 500 m • In this range, lowest Probability of Partition when obstruction radius = 400 m

13. Network Partitions – Mobility & Communication Blocking • The maximum hop distance used was ½R = 500 m • Pretty similar characteristics to just mobility blocking

14. Average Required Power Per Node • Maximum hop Distance: 2R  No Partitions • Assumes perfect knowledge of required power values

15. Effects of Imperfect Knowledge on Required Power Values • Nodes = 30 • Update Period: Time before nodes figure out that the best path to minimize power has changed • As the update period increases, required power increases

16. Conclusions • Effects of obstructions on Random Waypoint Model were studied • A more customizable model presented

17. Future Work • Use Markov velocity model • Create multiple obstructions with different radii • Change the path metrics for choosing the routes required for minimum power

18. Acknowledgements • Dr. Carl Baum • Clemson University SURE Program • National Science Foundation • ECE faculty and Graduate Students

19. Questions