Process initiation based on Random genes significantly reduces convergence time. Alternative methods exist To Shorten the process. These methods are essential to explore. They can be used to follow a different pattern. Despite this, they may not necessarily be optimal [13], [14]. This warrants a comparison. Population size plays a massive role in the GA. It guarantees More diversity. It ensures a vast section of the parameter space is probed. However, there's A trade-off. A significant population easily allows for wide Genetic diversity. It allows for a broad parameter space to be explored randomly. However, there are negatives to consider. Achieving fitness for all individuals can take a substantial amount of time. The focal problem significantly influences population size. To cover a broad parameter range, we may need to increase the population size exponentially. A key suggestion is to have twice as many individuals as parameters under study. Typically, population sizes range from 20 to 100 people [11], [15], [16].

Every single member of the population gets reviewed. What do we find? In its parameter settings, we find the fitness level. Imagine we are making a robot. This robot should walk. Here is what the fitness of each member means. It's the distance it can walk. Or in problems such as the knapsack problem, the concept is similar. In the knapsack problem [17]. The benefit obtained by objects inside the knapsack that minimize weight is the adjustment. Upon considering our case, the scenario is slightly different. The concept is similar. However, it will be the average wait time for each vehicle. This wait time is at every intersection. For all vehicles used, this figure is needed. This will aid in assessing the congestion levels. This would not be possible if individuals were not evaluated. The evaluation gives an insight into the fitness level. Individuals are judged based on their ability to carry out specific tasks. These are serious considerations when designing a transportation system. There must be a smooth, efficient flow of traffic. Evaluation of every individual in the population is a crucial aspect of this process. Ultimately, we have a metric to assess the performance of the system [12]. This is the average waiting time at each intersection. The waiting time at every intersection is a critical consideration. It's an essential piece of information to determine the efficiency of traffic flow. Vehicles must not come to a standstill. Waiting times must be kept to a minimum. Our evaluation of the population provides a clear picture of this scenario. By utilizing these parameters, we can design an improved transportation system that effectively addresses congestion.

After all the individuals constituting the population have had their abilities examined, the next step is to select the group that exhibits the best adaptation to its surroundings. The Methods to pick the crème de la crème range widely. Roulette wheel selection is one such mechanism; Boltzmann selection is another. Tournament selection and rank selection are more. We also find steady-state selection and elitist selection. There are several others as well. One of the simplest and most effective Strategies is as follows [18–20]. We select a certain percentage of the individuals. They possess the highest level of fitness. With this mechanism, it is possible to gather the best-adapted individuals. They fit the environment of the current generation. Parents are what we call these Selected individuals. They will be the ones generating the children. The percentage of selective individuals plays a crucial role in the correct functioning of the GA. A very high value will hinder rapid convergence. However, a very low value will also lead to low genetic diversity. A high probability of converging towards a local minimum exists [11], [15], [16]. Typical values for this percentage occur in a Range of 5 to 10. These values vary with population size. The remaining individuals are known as children. These parents will produce them. The total of the population will be obtained this way. For this, we employ crossover and mutation mechanisms. Although sophisticated methods exist for the Genetic Algorithm's selection process, rank selection is preferable in initial generations. This is to get quick preliminary results. More sophisticated techniques are found elsewhere. [20].

After all the individuals constituting the population have had their abilities examined, the next step is to select the group that exhibits the best adaptation to its surroundings. The Methods to pick the crème de la crème range widely. Roulette wheel selection is one such mechanism; Boltzmann selection is another. Tournament selection and rank selection are more. We also find steady-state selection and elitist selection. There are several others as well. One of the simplest and most effective Strategies is as follows. We select a certain percentage of the individuals. They possess the highest level of fitness. With this mechanism, it is possible to gather the best-adapted individuals. They fit the environment of the current generation. Parents are what we call these Selected individuals. They will be the ones generating the children [21–23]. The percentage selectivity of individuals plays a crucial role in the correct functioning of GA. GA: A very high value will hinder rapid convergence. However, a very low value will also lead to low genetic diversity. A high probability of converging towards a local minimum exists [24]. Typical values for this percentage occur in a Range of 5 to 10. These values vary with population size. The remaining individuals are known as children. These parents will produce them. The total population will be obtained this way. For this, we employ crossover and mutation mechanisms. Although sophisticated methods exist for the Genetic Algorithm's selection process, rank selection is preferable in initial generations. This is to get quick preliminary results. More sophisticated techniques are found elsewhere [11], [15], [16].

To achieve high genetic diversity, apply mutation mechanisms to a minority of genes. This is in the population. With this mutation, the value of every gene may fluctuate randomly with low probability. It prevents the GA from steadily moving towards a local best outcome. Then, it opens up other gene combinations to be randomly examined. Not too frequently, however, as it would turn into a random hunt. The mutation rate indicates how often a gene is altered. When it's 0% the genes of offspring are stable. If the rate is 100%, all Offspring genes Are Modified within A particular range. This range is less than 20% of the permissible range for each gene. The rate of mutation must be between 0.1% and 1%. It allows exploration. This exploration leads to a wider parameter space. This Way, genetic diversity is achieved, and at the same time, random [48], [49]. Parameter space exploration is avoided. The mutation rate must decrease. The margin of change in parameters must also decrease as the generations are formed. The goal is to achieve a swift convergence. This process is shown in Fig. 2. It depicts the Procedure followed by the GA up to the Convergence point. For iREDVD, we settle on a population size of 256 individuals [11], [15], [16]. We have 16 selected parents. The Crossover rate is at 80%. The Mutation Rate is 0.1%. Generation count can go up to 20.

Waiting time for the entire population. The X-axis represents the generations.

Fig. 3 sketches a picture. It shows the value of the normalized Waiting time. This is for vehicles per intersection. The figures are for the optimization process. The GA executes it in the training phase. As can be seen, after around five Generations, the GA hits promising values. This is for almost all simulated individuals. The simulation could have been concluded at that point. But it was left running for more. This was done to find a strong set of parameters. These are abilities that, with Time, can prove their superiority over others. The time needed for each simulation was recorded. This was part of the optimization Process for the test scenario. It was roughly 1 min. This indicates simulating 256 individuals for 21 generations on a Server with 16 processing cores [5]. This required about 48 hours. Each simulation was repeated 10 times. After that, the average value and standard deviation of each variable were determined. Post that, the best parameter configuration was implemented. Execution of the Traffic light control algorithm is nearly instant. It no longer requires an artificial intelligence algorithm. It's based on A light method. A fast method. A quick method! Low-power IoT devices can swiftly execute this method.

For further information, one can look at [5]. Now, the values of the optimized parameters after training are displayed in Table 4. This reveals some interesting numbers during the optimization process. The parameter limdec denotes The Number of times the green time of Traffic lights is reduced. The purpose is to decrease the total cycle time at the intersection. If limdec is 1, even a minor decrease in traffic prompts a reduction in total cycle time. This ultimately also reduces the green time. The opposite holds for an Increase in Total cycle time. The parameter liminc controls the increase.

The total cycle time is that of the traffic lights. If the traffic increases, this cycle Time also increases. A high value indicates a cautious approach [5]. It points to the fact that Iredvd must be certain before an increase in total cycle time. So, when traffic is levied, the total cycle time of the traffic lights is only to be incremented when the certainty of iREDVD is at a high level. Data from the training scenario is tabulated. Refer to it as

Let us compare it to the traditional algorithms, specifically Fixed 30, Fixed 45, and Fixed 60. Additionally, GreenWave 30, GreenWave 45, and GreenWave 60 are considered for comparison. Our improvement over the original REDV algorithm proposed in [5] remains visible. It underlines the need for proper Functionality of the algorithm to adjust parameters. Compared to the original REDVD variant found in [5], there is a 27% improvement. An improvement means a reduction of over 8 seconds in average normalized waiting time. Regarding other Key performance metrics, such as average travel time and speed, we can see further improvement. Average CO emissions have decreased by 13% to 46%.

Average fuel consumption decreases 7% to 18%. All of these KPI show improvements with iREDVD. It ranges from 7% to 45%. Notably, speed increases by 16 to 26%. There are decreases in other metrics. The average CO emissions and average fuel consumption have decreased by 7% to 18%. You see these compared to traditional algorithms [5].

Upon completion of the training, we applied stress to the algorithm. This happened during testing scenarios. We did use the optimized values.

We used optimized values for the configuration parameters. Results in Tables 6 and 7 indicate improvements. Improvements are very close to those obtained during the training phase. This confirms something. We confirmed the iREDVD's ability. iREDVD can adjust to unfamiliar conditions. These are conditions it was not trained for.

It maintains its robustness. It can be used in real deployments. Consider a case in point. Consider the Software's performance on two tests. In Test 1, iREDVD achieves a reduction in waiting time. They achieve 34% to 49% reduction compared to other control techniques. In Test 2, the results are more impressive. The iREDVD achieves a reduction of 78% to 82%. How massive is this decrease? Think about a few things. Think about the original waiting time. Now imagine a 6 to 100-second drop in that waiting time at every intersection. That's the equivalent. The improvements are not Limited to reductions. Other metrics reflect improvement. Consider the reduction of carbon monoxide emissions. They go down by around 25%. Then there is the reduction in average fuel. It goes down between 6 to 17%. The iREDVD software not only offers time savings but also makes a significant contribution to overall environmental health. It does so by improving vehicular efficiency. Its impact is seen not just in time saved, but also in Reduced emissions.

That impact is felt in lowered fuel consumption. Implications of the iREDVD software are wide-ranging. They're not limited to a single sphere. The algorithm can be applied to many scenarios. This robustness is a quality that sets it apart. It's a quality that makes it a candidate for use in real-world deployments

The exceptional performance of fine-tuned iREDVD is due to its optimized design. The ability to adjust to road Conditions is behind it. We can find it in Fig. 10.a And Also in Fig. 10.b. There, you Can See how iREDVD proactively adjusts the cycle time in Fig. 10.a during training. There's also an adaptation during unknown Circumstances in testing Fig. 10.b. The cycle lengthens when more vehicles pass through the intersection, and conversely, it shortens.

Further insight is provided in Fig. 10.a. And there's also more in Fig. 10.b. The time for each branch is separate. Traffic doesn't matter. It's independent. iREDVD adjusts to traffic conditions. It accounts for asymmetric traffic. See Fig. 10.b at Hour 4. At that Point, iREDVD acquires a different time for each branch. This is key. It self-adjusts to traffic conditions. Independence drives iREDVD to a more efficient time distribution. This also leads to a reduction in waiting time