Academia.eduAcademia.edu

Figure 3 – uploaded by Tuong Manh Vu

See full PDFdownloadDownload figure
EXPERIMENT 2 RESULTS USING OUR SYSTEM aaa eT EXPERIMENT 2 RESULTS USING ANYLOGIC As previously, the experiment was then altered to run in virtual mode in AnyLogic, which simply executes transitions as fast as possible, and without any GUI in our system. The results are presented in Table III for our system and in Table IV for AnyLogic.

Table 3 EXPERIMENT 2 RESULTS USING OUR SYSTEM aaa eT EXPERIMENT 2 RESULTS USING ANYLOGIC As previously, the experiment was then altered to run in virtual mode in AnyLogic, which simply executes transitions as fast as possible, and without any GUI in our system. The results are presented in Table III for our system and in Table IV for AnyLogic.

Related Figures (13)

A Finite-State Machine (FSM) is conceived as an abstract machine that can be in one of a finite number of states. It can change from one state to another when initiated by a triggering event or condition; this is called a transition [18]. There are different types of FSMs that can be used in a variety of different situations. We distinguish in this paper specifically between deterministic and stochastic FSMs. While deterministic FSMs are based on mathematical formulas and can be formally proven, stochastic FSMs use stochastic rules for deciding about transitions and therefore the exact outcome cannot be predicted; however it may be estimated using models and theories. A turnstile is a good example of a simple deterministic FSM. It can either be locked or unlocked, and there are predefined conditions that determine which state the turnstile is in, and the transitions it can take from each state. Fig. | demonstrates this. Fig. 1 Asimple deterministic FSM (Source: [19])
A Finite-State Machine (FSM) is conceived as an abstract machine that can be in one of a finite number of states. It can change from one state to another when initiated by a triggering event or condition; this is called a transition [18]. There are different types of FSMs that can be used in a variety of different situations. We distinguish in this paper specifically between deterministic and stochastic FSMs. While deterministic FSMs are based on mathematical formulas and can be formally proven, stochastic FSMs use stochastic rules for deciding about transitions and therefore the exact outcome cannot be predicted; however it may be estimated using models and theories. A turnstile is a good example of a simple deterministic FSM. It can either be locked or unlocked, and there are predefined conditions that determine which state the turnstile is in, and the transitions it can take from each state. Fig. | demonstrates this. Fig. 1 Asimple deterministic FSM (Source: [19])
Fig. 3 Hierarchical attack state from NPC guard The HSM provides a structural concept for representing behaviour, which can otherwise be complex and resulting in somewhat of "Spaghetti" code when trying to split into logical if statements, in order to structure the functionality into logical categories of behaviour. Fig. 3 shows the hierarchical state "attack" from Fig. 2.
Fig. 3 Hierarchical attack state from NPC guard The HSM provides a structural concept for representing behaviour, which can otherwise be complex and resulting in somewhat of "Spaghetti" code when trying to split into logical if statements, in order to structure the functionality into logical categories of behaviour. Fig. 3 shows the hierarchical state "attack" from Fig. 2.
Fig. 2 Typical FSM for an NPC guard in a game shows the first level of a HSM of a typical NPC enemy in a typical first person shooter game.
Fig. 2 Typical FSM for an NPC guard in a game shows the first level of a HSM of a typical NPC enemy in a typical first person shooter game.
Fig. 4 UML diagram for an office worker (Source: [13])
Fig. 4 UML diagram for an office worker (Source: [13])
Fig. 5 UML class diagram for our ABS environment
Fig. 5 UML class diagram for our ABS environment
Fig. 6 UML version of a more complex NPC Guard What we can see in Fig. 6 is that within each composite state there is an initial state pointer. This indicates that a composite state effectively contains a state machine, and so exhibits properties of both, a state machine and a normal state. Therefore it makes sense to reuse both, the state and state machine object, through use of multiple inheritance. Inheriting properties from multiple objects can be a complex problem in object-oriented programming [22]. However, in our case it seems an appropriate solution to the CompositeState problem.
Fig. 6 UML version of a more complex NPC Guard What we can see in Fig. 6 is that within each composite state there is an initial state pointer. This indicates that a composite state effectively contains a state machine, and so exhibits properties of both, a state machine and a normal state. Therefore it makes sense to reuse both, the state and state machine object, through use of multiple inheritance. Inheriting properties from multiple objects can be a complex problem in object-oriented programming [22]. However, in our case it seems an appropriate solution to the CompositeState problem.
EXPERIMENT | RESULTS USING OUR SYSTEM
EXPERIMENT | RESULTS USING OUR SYSTEM
EXPERIMENT | RESULTS USING ANYLOGIC As can be seen in Table I, the number of time steps executed per second is linearly affected by the number of
EXPERIMENT | RESULTS USING ANYLOGIC As can be seen in Table I, the number of time steps executed per second is linearly affected by the number of
Fig. 8: Our system model after 5 time steps
Fig. 8: Our system model after 5 time steps
Fig. 7: UML state-chart for a simple state machine agent
Fig. 7: UML state-chart for a simple state machine agent
Table III shows that the performance for our system has remained almost the same as in Experiment | in terms of memory usage and time steps executed per second. The same is true for AnyLogic in relation to the time steps executed per second, as can be seen in Table IV. However, memory usage has gone up quite significantly. This demonstrates the advantages of our state machine design. One can save potentially a lot of memory when constructing models that feature more complex statecharts.
Table III shows that the performance for our system has remained almost the same as in Experiment | in terms of memory usage and time steps executed per second. The same is true for AnyLogic in relation to the time steps executed per second, as can be seen in Table IV. However, memory usage has gone up quite significantly. This demonstrates the advantages of our state machine design. One can save potentially a lot of memory when constructing models that feature more complex statecharts.
This experiment implements the idea of composite state machines. In addition to the simple agent behaviour experiment, the stateRed state will now contain a clone of the previous state machine. The agent should move into the stateRed state and move between the states within this composite state. The UML statechart for the more complex agent is provided in Fig. 10. Fig. 10: UML state-chart with a composite state machine agent As with the previous experiment, a counter was used in both AnyLogic and our tool to count the number of agents in each state. Fig. 11 and 12 illustrate the number of agents in each state after 5 time steps in both, our system and in AnyLogic. In Fig. 11, the numbers 0 and 1 represent stateBlue and stateRed, respectively, and the numbers 2 and 3 represent compStateBlue and compStateRed, respectively. As can be seen in Figs. 11 & 12, the number of agents in each state varies between our tool and AnyLogic due to random number being used to determine whether a transition should be made, however the agent behaves as expected with a composite state.
This experiment implements the idea of composite state machines. In addition to the simple agent behaviour experiment, the stateRed state will now contain a clone of the previous state machine. The agent should move into the stateRed state and move between the states within this composite state. The UML statechart for the more complex agent is provided in Fig. 10. Fig. 10: UML state-chart with a composite state machine agent As with the previous experiment, a counter was used in both AnyLogic and our tool to count the number of agents in each state. Fig. 11 and 12 illustrate the number of agents in each state after 5 time steps in both, our system and in AnyLogic. In Fig. 11, the numbers 0 and 1 represent stateBlue and stateRed, respectively, and the numbers 2 and 3 represent compStateBlue and compStateRed, respectively. As can be seen in Figs. 11 & 12, the number of agents in each state varies between our tool and AnyLogic due to random number being used to determine whether a transition should be made, however the agent behaves as expected with a composite state.
Fig. 11: Our system model after 5 time steps
Fig. 11: Our system model after 5 time steps
Related topics:
Agent Based Simulation
580 California St., Suite 400
San Francisco, CA, 94104
© 2025 Academia. All rights reserved