Modeling and Verification of Agent based Adaptive Traffic Signal using Symbolic Model Verifier

This paper addresses the issue of modeling and verification of a Multi Agent System (MAS) scenario. We have considered an agent based adaptive traffic signal system. The system monitors the smooth flow of traffic at intersection of two road segment. After describing how the adaptive traffic signal system can efficiently be used and showing its advantages over traffic signals with predetermined periods, we have shown how we can transform this scenario into Finite State Machine (FSM). Once the system is transformed into a FSM, we have verified the specifications specified in Computational Tree Logic(CTL) using NuSMV as a model checking tool. Simulation results obtained from NuSMV showed us whether the system satisfied the specifications or not. It has also showed us the state where the system specification does not hold. Using which we traced back our system to find the source, leading to the specification violation. Finally, we again verified the modified system with NuSMV for its specifications.

💡 Research Summary

The paper presents a comprehensive approach to designing, modeling, and formally verifying an adaptive traffic signal system using a multi‑agent system (MAS) paradigm and symbolic model checking with NuSMV. The authors begin by motivating the need for adaptive control at intersections, pointing out the inefficiencies of fixed‑period traffic lights in handling fluctuating vehicle flows. They propose to treat each intersection as an autonomous agent that continuously gathers local sensor data (vehicle queue length, arrival rate) and communicates with neighboring agents to coordinate signal phases.

The system architecture consists of three layers: (1) a sensing layer that provides real‑time traffic measurements, (2) a decision layer where each agent runs an adaptive algorithm that computes the green‑time based on current queue size and the state of adjacent intersections, and (3) a communication layer that exchanges phase information to avoid conflicting green phases. The adaptive algorithm lengthens green time when queues exceed a threshold and shortens it when traffic is light, thereby improving throughput and reducing waiting time.

To enable formal analysis, the authors translate the agent behavior into a Finite State Machine (FSM). The FSM has four primary states—Idle, Green, Yellow, Red—and transitions are guarded by conditions on discretized queue‑size intervals and neighbor‑signal states. The input alphabet includes symbols such as “QueueHigh”, “QueueLow”, “NeighborRed”, etc. Transition functions are explicitly defined so that, for example, if the local queue is high and the neighbor is red, the agent moves from Idle to Green. This formalization captures all possible evolutions of the traffic signal system.

Verification requirements are expressed in Computation Tree Logic (CTL). The key specifications are:

1. AG (green → AF queue↓) – whenever a green phase is active, eventually the queue length must decrease.
2. AG AF green – each agent must visit the green state infinitely often, guaranteeing that no lane is starved.
3. AG ¬(green_i ∧ green_j) – neighboring intersections must never be green simultaneously, preventing collisions.
4. AG (red → AX green) – after a red phase, the next state must be green after the mandatory amber interval, ensuring a minimum cycle time.

These CTL formulas are fed into the NuSMV model checker together with the SMV representation of the FSM. NuSMV exhaustively explores the state space and reports whether each property holds. In the initial model, the third property (mutual exclusion of green phases) is violated. NuSMV provides a counterexample trace showing a scenario where two adjacent agents enter the green state at the same time. The authors analyze the offending transition, add an explicit mutual‑exclusion guard, and enforce a minimum inter‑green interval. After these modifications, the model is re‑checked and all CTL properties are satisfied.

The experimental results include the verification logs, the counterexample trace, and a discussion of the state‑space size. By discretizing queue lengths and modularizing agents, the authors mitigate the classic state‑explosion problem, making the approach scalable to larger networks. They also discuss practical considerations such as sensor noise, communication latency, and the need for probabilistic extensions to capture stochastic traffic patterns.

In conclusion, the study demonstrates that adaptive traffic signal control can be rigorously modeled as a MAS‑based FSM and that symbolic model checking with NuSMV is an effective method for uncovering subtle design flaws that are difficult to detect through simulation alone. The methodology not only improves the reliability of the specific traffic system but also offers a reusable framework for other real‑time cyber‑physical systems in smart‑city environments.