Artificial intelligence (AI) based smart traffic management

1 . An Artificial Intelligence (AI) based traffic control apparatus comprising:

at least one processor,

a memory,

wherein the at least one processor and memory are installed on at a traffic signal bearing multiple traffic lights, and

the memory stores machine-readable instructions that when executed by the processor, cause the processor to:

identify traffic entities comprising vehicles, inanimate moveable objects and living beings in image data streams obtained from a plurality of cameras,

wherein at least two of the plurality of cameras are installed on at least two traffic lights pointed in at least two different directions at an intersection to image at least two traffic streams in the at least two different directions,

wherein the vehicles include different types of vehicles;

generate representations of different types of vehicles in the at least two traffic streams as multiples of the single entity based on a number of wheels on each of the different types of vehicles;

obtain two aggregates of the representations corresponding to the at least two traffic streams as the corresponding relative densities;

obtain corresponding relative densities of the at least two traffic streams from the two aggregates;

determine, based at least on the relative densities of the at least two traffic streams, times for crossing the intersection of the at least two traffic streams; and

manipulate the traffic lights at the intersection cyclically, wherein in a given cycle each traffic light of the at least two traffic signals receives a green signal and an order of the traffic lights to receive the green signal in the given cycle is determined based on the times of the at least two traffic streams to cross the intersection.

2 . The AI-based traffic control apparatus of claim 1 , wherein the at least one processor and the memory are installed on a fixture including one or more of the plurality of cameras.

3 . The AI-based traffic control apparatus of claim 2 , wherein the machine-readable instructions further cause the at least one processor to:

flush a subset of the image data streams corresponding to the given cycle from the memory at an end of the given cycle.

4 . The AI-based traffic control apparatus of claim 1 , wherein the machine-readable instructions to identify the individual vehicles cause the at least one processor to:

append at least two of the image data streams received from at least two of the plurality of cameras pointing in a same direction of the at least two different directions.

5 . The AI-based traffic control apparatus of claim 1 , wherein the machine-readable instructions to identify the individual vehicles cause the at least one processor to:

provide the image data streams to a plurality of object detection models; and

identify, by the plurality of object detection models a type of the individual vehicles from the image data streams.

6 . The AI-based traffic control apparatus of claim 5 , wherein the machine-readable instructions further cause the at least one processor to:

train the plurality of object detection models in identifying a corresponding type of vehicle.

7 . The AI-based traffic control apparatus of claim 5 , wherein the plurality of object detection models include You Only Look Once (YOLO) models.

8 . The AI-based traffic control apparatus of claim 1 , wherein the machine-readable instructions to manipulate the traffic signals at the intersection cyclically further cause the at least one processor to:

transmit a signal to turn green to the traffic light pointed in one of the at least two directions having a maximum value for the time to cross the intersection for a corresponding subset of the individual vehicles.

9 . The AI-based traffic control apparatus of claim 8 , wherein the machine-readable instructions further cause the at least one processor to:

transmit a signal to change the traffic light when the time elapses.

10 . The AI-based traffic control apparatus of claim 1 , wherein the machine-readable instructions further cause the at least one processor to:

identify the vehicles of the different types from the image data streams in parallel with the manipulation of the traffic lights.

11 . A processor-executable method of optimizing traffic signals comprising:

receiving, by at least one processor installed at a traffic signal and executing instructions stored in a memory installed at the traffic signal, image data streams including images of traffic entity streams captured by a plurality of image sensors installed in different directions at the traffic signal in an intersection wherein the traffic signal has a plurality of sets of traffic lights facing the different directions;

analyzing, by the at least one processor using a plurality of machine learning (ML) models stored in the memory, the image data streams for vehicular identification;

identifying based on the analysis, by the at least one processor, traffic entities representing vehicles of different types from the image data streams;

generating, by the at least one processor, representations of different types of vehicles in the at least two traffic streams as multiples of the single entity based on a number of wheels on each of the different types of vehicles;

obtaining, by the at least one processor, two aggregates of the representations corresponding to the at least two traffic streams;

obtaining, by the at least one processor, corresponding relative densities of the at least two traffic streams from the two aggregates;

calculating, by the at least one processor, a time for crossing the intersection by the traffic entity streams in the different directions based at least on the corresponding relative densities of the traffic entity streams;

determining, by the at least one processor, one of the traffic entity streams having a maximum value for the time for crossing the intersection;

transmitting, by the at least one processor, a first signal to turn green, to a set of traffic lights of the plurality of sets of traffic lights facing one of the traffic streams with the maximum value for the time for crossing the intersection;

sensing, by the at least one processor, an end of a period of the maximum value of the time for crossing the intersection; and

transmitting by the at least one processor, a second signal to change color to the set of traffic lights at the end of the period.

12 . The method of claim 11 , further comprising:

repeating cyclically, by the at least one processor, the steps of analyzing, identifying, calculating, determining, transmitting a signal to turn green, sensing, and transmitting a second signal to change color, until each set of traffic lights in the plurality of traffic lights facing the different directions receive the first signal at least once.

13 . The method of claim 12 , further comprising:

flushing, by the at least one processor, the image data streams accumulated during a cycle of traffic signal changes.

14 . The method of claim 11 , wherein the traffic entity streams includes living beings and moveable objects and the traffic signal is a pedestrian crossing signal.

15 . The method of claim 11 , wherein the traffic entity streams includes vehicles differentiated based on a number of wheels on a particular vehicle type and the traffic signal is a signal that allows passage of the vehicles.

16 . A non-transitory storage medium comprising machine-readable instructions that cause at least one processor to:

receive image data streams including images of traffic streams captured by a plurality of image sensors installed in different directions at a traffic signal in an intersection wherein the traffic signal has a plurality of sets of traffic lights facing the different directions;

analyze by a plurality of machine learning (ML) models, the image data streams for vehicular identification;

identify, by the at least one processor, vehicles of different types from the image data streams;

generate representations of different types of vehicles in the at least two traffic streams as multiples of the single entity based on a number of wheels on each of the different types of vehicles;

obtain corresponding aggregates of the representations for each of the traffic streams;

obtain respective relative densities of the traffic streams from the corresponding aggregates;

calculate, a time for crossing the intersection by the traffic streams in the different directions from the respective relative densities;

determine one of the traffic streams having a maximum value for the time for crossing the intersection;

transmit, a first signal to turn green, to a set of traffic lights of the plurality of sets of traffic lights facing one of the traffic streams with the maximum value for the time for crossing the intersection;

sense an end of a period of the maximum value for the time for crossing the intersection; and

transmit a second signal to change color to the set of traffic lights at the end of the period.

17 . The non-transitory storage medium of claim 16 , comprising further instructions that cause at least one processor to:

repeat cyclically, by the at least one processor, the steps of analyzing, identifying, calculating, determining, transmitting a signal to turn green, sensing, and transmitting a second signal to change color, until each set of traffic lights in the plurality of traffic lights facing the different directions receive the first signal at least once.

18 . The non-transitory storage medium of claim 16 , comprising further instructions that cause at least one processor to:

flush at an end of a cycle of traffic signal changes wherein each of the plurality of sets of traffic lights has received a green signal, the image data streams accumulated during the cycle.

19 . The non-transitory storage medium of claim 16 , comprising further instructions that cause at least one processor to:

obtain a number of individual vehicles in the image data streams.