Dynamic Leadership Protocol for S-nets Gregory J. Barlow, Thomas C. Henderson, Andrew L. Nelson, and Edward Grant North Carolina State University University.

Dynamic LeadershipProtocol for S-nets

Gregory J. Barlow, Thomas C. Henderson, Andrew L. Nelson, and Edward Grant

North Carolina State University

University of Utah

Introduction

• Distributed sensing is an alternative to using large amounts of on-board sensors on mobile robots

• Smart sensor networks can be used for distributed sensing, communication, and computation

• This work presents a leadership protocol that forms clusters in a smart sensor network for distributed sensing

S-nets

• S-element: a stationary agent capable of computation, communication, and sensing. S-elements have a limited communication range.

• S-net: a network of spatially distributed S-elements.

• S-cluster: a group of S-elements with one agent as the leader.

Dynamic S-net Leadership Algorithm

• The DSNL algorithm is a distributed algorithm run by each S-element

• Each S-element must have a unique identification number

• Our goal is to form S-clusters with one leader for each cluster

• As S-elements are added to and removed from the S-net, clusters should update dynamically

S-element State

id_num unique ID number

leader Boolean, whether node is a leader

resolved Boolean, whether node is resolved

nodelist list of all nodes in communication range

remaining list of unresolved nodes

cluster list of resolved nodes in the cluster

lastcluster list of nodes in the cluster during the

previous generation

noncluster list of resolved nodes not in the cluster

DSNL Algorithm

Update lists of S-elements

Resolve the node’s leadership status

Resolve nodes in remaining

Execute task code once resolved

Objectives

1. The node that has the lowest ID number of all unresolved nodes in communication range should resolve as a leader

2. Any node that is in communication range of a leader should resolve as a follower

3. Every node should be a leader or a follower

4. When a follower is removed, its leader should remove it from cluster

5. When a node’s leader is removed, that node should re-resolve

S-nets implementation in simulation

S-nets implementation in simulation

Time

Node

1 2 3 4

0

1 L L

4 L L F F

10 L L

12 L L FL = leader, F = follower

S-net implementation using a robot colony

S-nets implementation using 20 S-elements

Conclusions

• We developed a leadership protocol for S-nets that allows dynamic updating of clusters

• We also developed an implementation of algorithm for embedded systems

• We successfully tested the leadership protocol in simulation and on a colony of mobile robots