Results and Discussion

Running the scheduling program for multiple iterations, the study has reinforced the fact that more resources would inevitably improve the effectiveness of patrolling operations, further emphasizing the importance of optimizing the resources available to a security firm.

When the visit frequencies of the developed system are compared to the output of the scheduler utilizing Uniform Random Strategy (URS), the results show that implementing pure randomization is less effective. The two figures below show a sample situation of protecting four areas. Area A is gien the highest priority, and Area D is the given the lowest. Weighted probability allows the system to deploy more patrols to Area A more frequently. It can be seen that giving all areas equal probability regardless of risk will lead to poorly allocated patrols. This could lead to insufficient patrol presence in higher risk areas.

To test the scheduling program's capability of optimizing resources, another simple scenario was devised. Given 8 patches with varying levels of priority, Patch 1 having the highest priority and Patch 8 having the lowest, the scheduler was tasked to generate schedules with varying number of patrols. Ten schedules were generated where the scheduler is given 1 patrol. Another ten where the program has two patrols. Schedules are also generated for three and four patrols, respectively, and the data collected is displayed on the four graphs below.

The impact of having varying probability weights can be seen in the graphs. When only given one patrol to guard 8 patches, the algorithm was still able to give Patch 1 a relatively good amount of visits compared to the other patches. When there are half as many patrols as there are patches, The higher priority patches are visited over half the time.

The figures below show how the OCSG algorithm handles high and low priority areas, compared again to Uniform Random Strategy. The OCSG algorithm is more efficient in deploying patrols to key areas. When given only three patrols, the algorithm deploys patrols at the same rate as URS with five patrols. The second chart shows that the algorithm visited low priority areas less than URS, but this is caused by the OCSG algorithm's tendency to visit higher priority areas more often. URS only assigns randomly regardless of the actual patch's susceptibility to threats, and from a security standpoint, this is not considered effective.

The key to avoiding predictability lies with the randomization of the visit timings. As enemies watch and observe movement, they are able to approximate how often an area is visited, but they will find it difficult to figure out the timings of the patrols. Simply put, they may be able to estimate how often a patrol is deployed in a patch, but they cannot predict when the patrol will be deployed again there. The uncertainty brought about by constantly changing patterns is often enough to discourage them from attempting attacks.

The scheduler does not provide explicit directions to patrols. Patrols are only instructed to relocate to a patch and patrol the surrounding area. This is done in order to provide patrols freedom in their movements. Micro-management of patrols could be discomforting to personnel, as following explicit sets of instructions could take away their sense of freedom.

There are several limitations that this system is facing. Several of those brought upon by the zoning approach. The user cannot transfer a patrol to a different zone, which if enabled could be an opportunity for more flexible schedules. Patrols within a zone still have a tendency to be concentrated in one area of the zone. Problems also arise if a zone does not have at least one patrol in it, because the program would simply leave the zone unattended and vulnerable.

The scheduling procedure has some limitations as well. The algorithm uses the same weight values across its duration, which means that the user has to create a new schedule if the weights are no longer suitable for the situation. The program is also not designed to create schedules that span multiple shifts. If the user would want to schedule a 24-hour day with three shifts, for example, the scheduler would need to be set up and run three separate times.

The algorithm for assigning patrols to patches is somewhat simplistic and primitive. The assignments take into account only the next interval, not the rest of the schedule. There are still possible schedules in which the patrols have significant differences in mileages. If there is a way for the program to look ahead into the schedule when assigning patrols, there is potential for reducing the average mileages further.

When it comes to determining travel distances, it would be better for the program to be given a patrol's real time location data rather than static distance values based on where the patrol is assigned. This would be useful for adjusting the schedule if a patrol had to respond to an emergency situation. In its current state, the program is not capable of receiving and computing real-time location such as data from Global Positioning System (GPS).