Behavior Control via Social Network Dynamics

I. Background and Motivation

ODOL trucks affect the road infrastructure and road safety [1], [2], [3]. The road infrastructure has a load limit it can support, and by overloading it, the infrastructure wears out quickly. ODOL trucks have a very bad vehicle-dynamics. For example, its center of gravity could be too high or too close to the rear end. These make the truck roll-over easily when cornering or difficult to steer due to relatively small grip on the front tires compared to the rear tires; and these make ODOL trucks prone to road accidents. Other than that, ODOL truck also has a very bad aerodynamics which makes the truck to consume more fuel than it should be. The work [3] shows that operational costs of ODOL trucks are lower, but the road maintenance and road accidents costs increase.

The effectiveness of the government regulation on ODOL trucks rely on the law enforcement. When the law enforcement is not carried out properly, ODOL trucks persist [2]. We are motivated to propose an alternative way to solve ODOL trucks by using social network dynamics [4], [5]. The social network dynamics itself is an independent topic and its approach is general, which means it can also be applied to other social problems.

The advantage of using social network dynamics in solving social problems compared to government regulation is that it does not put additional external constraints to individual member of the society. From this method, public policies are suggested in order to create the social environment needed for the method to work.

The next section explains the approach we use to model human-decision making. Section III explains the leader-follower network. Section IV concludes this paper.

II. Decision Making by Humans

The model that we use to mimic how human make decision is Model Predictive Control (MPC) studied in systems and control theory. In MPC, a system has an objective function it needs to minimize over some constraints. The objective function could be designed to stabilize the system or to track some trajectories. In doing this, at every time step or every time it needs to make a decision, the system predicts its future state based on the mathematical model of the system, and it chooses an optimal control input such that the constraints is being satisfied and also results in the lowest cost on the objective function.

Analogously, human has a certain objective they want to achieve, either a short term objective or long term objective. For example, a person is thirsty and he/she needs water, so obtaining drinking water is his/her objective. The money that he/she has is one of the constraints, where he/she can specify the limit of the highest drinking water price that he/she can afford. If energy consumption to get the water is also included in the constraints, he/she would pick the nearest stores to buy the water.

Based on this model, we can see that the decision that human make depends on at least two things: total constraints and information about the surroundings. Suppose that the person who wants to buy drinking water looks at his/her phone-map about the nearest stores, and the geographically-nearest store is being excluded from the map, then he/she will choose the nearest store based on information obtained from the phone although it is not the real-nearest store. On the other hand, suppose that he/she has a lot of money and has no spending constraints, even if he/she chooses the same store as the person in the above example, he/she could possibly purchase a different drinking water.

The social network dynamics method that we propose in this study exploits the use of information to alter or direct the decision made by humans. As a note, the constraint in the example above is an internal constraint; while examples of external constraint are social norms and public policies.

III. Leader-Follower Network

In a distributed control of multi-robot system, robot share information with each other. Robot A is a neighbor of robot B if B receives information from A. Robots then use information from their neighbors to compute their control input. When a distributed protocol such as consensus protocol is used and the robot network is connected; then asymptotically, the multi-robot system reach consensus. For example, they eventually have similar velocities, heading angle, and so on. A leader in the multi-robot network does not receive information from other robot, and thus the consensus value is determined by the leader’s state, in example the velocities of other robots follow the velocity of the leader robot. See chapter 3 of [4] for the discussion on how to reach consensus in multi-agent systems.

In human society, each individual member has their own decision, which makes the overall system to be naturally distributed, just like the above example. In the context of our study, human share information with each other. For example, person B sees the shirt color of person A (in this case we say person B receives information from person A). Person B might be or might not be using this information in their decision. This is because we cannot force every individual to use any information they receive in making their decision. See [6] for a study of how human use information from other society member in making their decision.

Our approach in this study is to make every member of the society to use certain information in making their decision, and eventually reach a consensus. In the case of ODOL truck, we propose to make several trucks and several companies as leader trucks and leader companies. The idea is to make these leader trucks and companies to follow the law on ODOL trucks and give them additional incentives which make the operational costs of trucks to be roughly similar to trucks practicing ODOL. The costs for these incentives could be taken from the costs that are going to be spent for road maintenance and to mitigate road accidents. Meanwhile, the information about the leader trucks and companies are spread throughout the country. This information, if followed, will make every other trucks and companies to gain more profit.

We assume that earning a living is a primary objective of every member of the society. Thus, information about how other trucks and companies gain more profit will certainly be used by other trucks and companies to make their decision. Eventually, when the social network is connected, all trucks and companies reach consensus which is following the law on ODOL. Meanwhile, government role here is to create the appropriate environment such as to give the incentives for trucks and companies that follow the law on ODOL, and severely punish the “pungutan liar (pungli)”.

IV. Conclusion

We have yet to prove if our proposed method will work in real world. Unknown disturbances such as “pungli” will certainly affect the network dynamics, since “pungli” agent will not gain any profit if trucks are following the law on ODOL. Thus, these factors need to be accounted for in the future study.

As opposed to the method proposed in this study, another study in our future work is to make a natural leader in the society. A natural leader is followed by other member in the society without any external party involvement such as government giving incentives in our proposed method. A condition on how a member can be a natural leader in the society is still being investigated.

References

[1] https://ekonomi.bisnis.com/read/20190904/98/1144532/truk-muatan-berlebih-pemicu-kecelakaan-kemenhub-ini-pr-berat

[2] https://www.thejakartapost.com/news/2013/08/03/emergency-revitalization-northern-java-coastal-road-line.html

[3] Ghisolfi, V., et. al., “Evaluating Impacts of Overweight in Road Freight Transportation: A Case Study in Brazil with System Dynamics”, Sustainability, MDPI, June 2019.

[4] Mesbahi, M. and Egersdtedt, M., “Graph Theoretic Methods in Multiagent Networks”, Princeton Series in Applied Mathematics, 2010.

[5] Jackson, M. O., “Social and Economic Networks”, Princeton University Press, 2008.

[6] Riyanto, Y. E. and Yeo, X. W. J., “Directed Trust and Reciprocity in a Real-LIife Social Network: An Experimental Investigation”, 2014.