27.03.2014 Views

SEKE 2012 Proceedings - Knowledge Systems Institute

SEKE 2012 Proceedings - Knowledge Systems Institute

SEKE 2012 Proceedings - Knowledge Systems Institute

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

The rest of this paper is organized as follows: In Section<br />

II, a case study to compare this work and previous works is<br />

presented. In Sections III and IV, behavioral modeling and<br />

emergent behavior detection are discus sed and ou r new<br />

algorithm is presented. Finally, in Section V, co nclusions<br />

and future work are given.<br />

II.<br />

CASE STUDY<br />

To obtain a f air comparison between the proposed<br />

technique and th e technique presented in [13], the same<br />

prototype of mine sweeping robot is used. The robot's<br />

objective is to pass 10km maze-based routes while<br />

identifying and flagging the existing mines. The robot is not<br />

given any map and just takes advantage of the information<br />

received from its sensors. The robot has four sensors: a GPS<br />

sensor that operates to co ntrol the certain distance (about<br />

10km) robot is allowed to pass , an infrared (IR) sensor to<br />

identify mines, a battery sensor that checks if the battery can<br />

provide the power and an ultra-sonic (ULT) sensor which<br />

helps the robot to n avigate the routes by finding the<br />

obstacles. The messages from these sensors are sent to t he<br />

client control to be processed and sent to the server control<br />

in order to take necessary actions.<br />

Battery<br />

GPS IR sensor ULT sensor Client control Server control<br />

Send signal (Battery has power )<br />

Send signal (obstacle detected)<br />

Motors move forward<br />

Rotate<br />

Stop motors<br />

Motors move forward<br />

Battery<br />

GPS<br />

IR sensor<br />

ULT sensor<br />

Send signal (Battery has power)<br />

III.<br />

Send signal (no obstacle detected)<br />

Client control<br />

Send signal (no mine detected)<br />

Figure 2. MSC m2<br />

Server control<br />

Motors move forward<br />

Rotate<br />

Rotate<br />

BEHAVIORAL MODELING<br />

Motors move forward<br />

Motors move forward<br />

Rotate<br />

Motors move<br />

Motors move<br />

Motors move<br />

In this paper, the process of converting the scenarios into<br />

the corresponding state machines based on the definitions<br />

given in [13] is perf ormed. These definitions are t hen<br />

applied to the case study to validate their effectiveness.<br />

As proposed in [13], in a message sequence chart MSC,<br />

Finite State Machines (FSMs) are built for any component i.<br />

In Figure 3, the FSMs for the client control component in<br />

MSC m1 ( Figure 1) are shown. <br />

are the initial<br />

state, the i th state and the final state of the component client<br />

control in MSC m1, respectively.<br />

s m1 0<br />

Send signal (battery has power)<br />

s m1 1<br />

motors move forward<br />

s m1 2<br />

Rotate<br />

s m1 3<br />

Send signal( obstacle detected)<br />

Send signal (mine detected)<br />

Stop<br />

Motors stop<br />

s m1 7<br />

stop motors<br />

Send signal(mine detected)<br />

s m1 6<br />

stop<br />

s m1 5<br />

stop motors<br />

s m1 4<br />

Stop motors<br />

Stop<br />

Motors stop<br />

s m1 8<br />

Stop<br />

s m1 f<br />

Figure 1. MSC m1<br />

Flag mine location<br />

Two scenarios that represent the behavior of the robot are<br />

presented. In Figure 1, MSC m1 shows where robot moves<br />

since the battery has power. Later, based on the message<br />

sent from IR sensor, the motor stops due to detection of an<br />

obstacle. Then, ULT sensor sends a message indicating<br />

detection of a mine but the robot has already stopped and so<br />

the mine is just flagged.<br />

In Figure 2, th e robot m oves since the ULT sensor,<br />

infrared and battery sensors do not cause the robot to s top.<br />

As in [13], messages are generalized to represent objective<br />

instead of implementation. For example, there are t hree<br />

messages in Figure 1, which have the same purpose of "send<br />

signal" where the contents of the messages are different.<br />

Figure 3. FSM for component client control of MSC m1<br />

Then, a certain value is given to each state of the FSMs<br />

of each component. This technique, is based o n the<br />

definitions initially presented in [7].<br />

Definition 1: If component i needs the result of message<br />

for doing , then message is a semantical<br />

cause for message and is shown by ,<br />

where represents the j th message interacted by the<br />

component i of MSC m.<br />

The semantic cause is an invariant feature of the system.<br />

As an example, in Figure 3, the message "send signal (mine<br />

detected)" is a semantic cause for performing message<br />

"stop". Each state of a component is defined by considering<br />

the messages that are se mantic causes for messages that<br />

come after it. The semantic cause is resulted from the<br />

domain knowledge and defined as:<br />

71

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!