Abstract
Police officer presence at an intersection discourages a potential traffic violator from violating the law. It also alerts the motorists' consciousness to take precaution and follow the rules. However, due to the abundant intersections and shortage of human resources, it is not possible to assign a police officer to every intersection. In this paper, we propose an intelligent and optimal policing strategy for traffic violation prevention. Our model consists of a specific number of targeted intersections and two police officers with no prior knowledge on the number of the traffic violations in the designated intersections. At each time interval, the proposed strategy assigns the two police officers to different intersections such that at the end of the time horizon, maximum traffic violation prevention is achieved. Our proposed methodology adapts the PROLA (Play and Random Observe Learning Algorithm) algorithm to achieve an optimal traffic violation prevention strategy. Finally, we conduct a case study to evaluate and demonstrate the performance of the proposed method.
Introduction
Traffic enforcement faces a fundamental "feedback paradox." While the presence of a marked police car (MC) is highly effective at preventing violations, it simultaneously obscures the underlying risk data. Because drivers follow the rules when they see an officer, that officer cannot determine if they are at a high-risk intersection or a naturally safe one. Traditional resource allocation often relies on historical data or intuition, which may be outdated or biased. This research addresses the challenge of optimal resource allocation under uncertainty: how can police departments identify the "Best Intersection" (BI) for prevention when the act of enforcement hides the very data needed to find it?
The PROLA Policing Strategy
The proposed solution models policing as an online learning task using two distinct agents: a Police Officer with a Marked Car (POMC/MC) and a Police Officer with an Unmarked Car (POUC/UC).
The MC acts as the "Actor," visible to drivers and actively preventing violations. Its goal is to maximize "unobservable rewards" (prevented crimes). Conversely, the UC acts as the "Observer," invisible to drivers and assigned to intersections at random to gather ground-truth data on violation rates.
The core of this strategy is the PROLA (Play and Random Observe Learning Algorithm). The algorithm maintains weights for each intersection, which are used to determine the probability of assigning the MC to a specific location. The UC's observations are used to update these weights. If the UC detects high violation rates at a particular spot, the algorithm increases the probability that the MC will be deployed there in subsequent rounds. This approach effectively decouples action from observation, allowing the system to learn from "untainted" data.
Experimental Evaluation
The strategy was evaluated through a simulated case study involving a geographical region with $K$ intersections (ranging from 10 to 50). Violation rates were generated using a Bernoulli distribution. For instance, in a 10-intersection scenario, "Intersection 6" was designated as the "Best Intersection" with a 60% violation probability, while others ranged from 4% to 20%. The simulation ran for up to 2,000 time intervals to observe long-term convergence and performance.
Results and Discussion
The results demonstrate that the PROLA-based strategy successfully identifies and prioritizes high-risk areas over time.
As shown in Figure 5(a), the probability of the MC being assigned to the "Best Intersection" (Intersection 6) climbs steadily, reaching nearly 90% as the algorithm learns the environment. Furthermore, Figure 5(b) illustrates that the "Total Regret"—the difference between the algorithm's performance and a theoretical perfect strategy—grows at a decreasing rate. This confirms that the algorithm becomes increasingly efficient. While regret increases as the number of intersections ($K$) grows, the system maintains its ability to identify optimal locations even at higher scales.
Conclusion
This research presents a "no-regret" online learning approach to traffic enforcement that overcomes the feedback paradox. By utilizing a tag-team of marked and unmarked vehicles, police departments can adaptively allocate resources to maximize public safety without prior knowledge of local traffic patterns. Future work may explore more complex reward structures beyond binary violations, account for human factors like officer fatigue, and test the algorithm in highly dynamic environments where traffic "hot spots" shift rapidly.
Summary updated