WO2018141403A1

WO2018141403A1 - System, device and method for managing traffic in a geographical location

Info

Publication number: WO2018141403A1
Application number: PCT/EP2017/052398
Authority: WO
Inventors: Sigurd Spieckermann; Vinay Sudhakaran; Leny Thangiah; Varsha RAVEENDRAN
Original assignee: Siemens Aktiengesellschaft
Priority date: 2017-02-03
Filing date: 2017-02-03
Publication date: 2018-08-09
Also published as: CN110494902A

Abstract

A method of managing traffic flow in a geographical location is provided. The method includes determining a traffic density in real time based on a traffic environment for at least one intersection (130), wherein the traffic environment comprises vehicular traffic and pedestrian traffic. The method includes predicting a forecast traffic density based on the traffic environment and a historical traffic environment associated with the at least one intersection (130). Further, the method includes selecting a traffic signal profile for the at least one intersection (130) based on the traffic density and the forecast traffic density. The traffic flow in the geographical location is managed based on the traffic signal profile of the at least one intersection (130).

Description

System, Device and Method for managing traffic in a geographical location

With increase in the number of vehicles, managing traffic flow on roads has become a daunting task. Generally, the traffic flow is managed at intersections having traffic signposts with a static signal time to control the traffic flow. The increase in vehicles or traffic density leads to significant traffic congestion, as the static signal time does not take into account variation in the traffic density. In addition, several factors such as rash or erratic driving and poor road conditions contribute to increased traffic congestion and lead to a chaotic traffic environment.

To manage the traffic flow in situations where the traffic density is varied, a human agents i.e. policemen are used to manage the intersections either in-person or remotely by using cameras. This method is limited to controlling only one intersection and is difficult to co-ordinate with other intersections .

To address the issue of scalability a centralised traffic control maybe used to process large volume of traffic data. The traffic data is collected from several sensors such as induction loops, wireless ground sensors, passive infrared detectors, high resolution camera system and radars. These sensors are not suitable for the chaotic traffic environment where no lane discipline of the vehicles is maintained. Additionally, the centralized traffic control is expensive to implement in view of the large volume of traffic data that requires to be processed.

A decentralized traffic control can be used to manage traffic to reduce the complexity of the centralized traffic control. Such a decentralized traffic control system and method is disclosed in US 20130176146. The method describes a system of agents in a traffic environment, where each agent represents a traffic light controller at an intersection to control the flow of traffic. Each agent collects data local to the junction using one or more sensors and applies distributed W- Learning model to adapt to the varying traffic conditions in a decentralized distributed manner. However, the method makes an assumption from sensor data that is used to collect traffic information. This assumption of the traffic flow may not be suitable in situations where there is a chaotic traffic environment .

In light of the above there exists a need to accurately man- age traffic flow in a geographical location having a chaotic traffic environment. Therefore, it is an object of the pre- sent invention to provide a method, a device and a system for effectively manage traffic flow in a chaotic traffic environ- ment .

The method, device and system according to the present invention achieve the aforementioned object by detecting traffic density at an intersection based on a traffic environment at the intersection, and predicting a forecast traffic density. The present invention also teaches selecting a traffic signal profile for the intersection based on the traffic density and the forecast traffic density and managing the traffic flow in the geographical location based on the traffic signal profile of the intersection.

As used herein, "traffic environment" indicates vehicle traf- fic and pedestrian traffic for a geographical location in which the traffic flow is to be managed. Accordingly, the traffic environment includes co-ordinates of the geographical location and data associated with weather, air quality, time, day, scheduled events and unscheduled events for the geo- graphical location. The term "geographical location" includes multiple intersections and therefore, the traffic environment also includes information associated with each intersection, such as lane closure or road maintenance. Further, "intersec- tion" means a traffic intersection point that has either manned or unmanned traffic signal-posts that is used to manage traffic flow at the intersection. Each intersection also includes one or more lanes, whose traffic flow is managed by the traffic sign-posts.

According to the present invention, a computer implemented method of managing traffic flow for the traffic environment comprising vehicular traffic and pedestrian traffic is provided. The method includes determining a traffic density in real time based on the traffic environment for one or more intersections. As used herein, "traffic density" includes number of vehicles of each vehicle type, number of private vehicles, number of public vehicles and passenger occupancy. Since the traffic environment includes information related to vehicular traffic and pedestrian traffic, the traffic environment is used to determine the traffic density. For example, the traffic density is determined in real-time by recognizing traffic objects such as car, scooter, bike, trams etc. in the traffic environment.

The method also includes predicting a forecast traffic density based on the traffic environment and a historical traffic environment associated with the intersection. As used herein, "historical traffic environment" includes information relating to traffic at the intersection for a time instant in the past. For example, to predict the forecast traffic density at time "t" the historical traffic environment includes information regarding vehicle traffic, pedestrian traffic and data associated with weather, air quality, time, day, scheduled events and unscheduled events for the geographical location, captured at time "^xt-x". The historical environment also includes the traffic density at time t-x. The forecast traffic density includes the number of vehicles of each vehicle type, number of private vehicles, number of public vehicles and passenger occupancy. Further, the method includes selecting a traffic signal profile for the intersection based on the traffic density in real-time and the forecast traffic density. As used herein, the "traffic signal profile" includes one of red, yellow and green colour profiles to indicate whether traffic in a lane should stop, proceed with caution, and proceed, respectively. The traffic signal profile also indicates a signal cycle time for the next green signal. The traffic signal profile is selected by comparing the forecast traffic density and the traffic density. If the traffic density differs significantly from the forecast traffic density the traffic signal profile is recomputed. In other words, the signal cycle time for the next green time is either increased or reduced for a lane in the intersection. In an embodiment, when there the traffic density cannot be determined in real-time, the forecast traffic density is used to generate the traffic signal profiles. For example, when the weather conditions lead to poor visibility of the traffic at the intersection, the forecast traffic density is used to generate the traffic signal profiles. In another example, a capture device used to capture the traffic density at the intersection may be faulty, in such a situation the forecast traffic density is used to generate the traffic signal profiles. Furthermore, the method includes managing the traffic flow for the traffic environment based on the traffic signal profile of the intersection. The traffic signal profile selected for the intersection is communicated in real time to neighbouring intersections in the geographical location. In an em- bodiment, the traffic signal profile for the intersection is communicated by means of a wireless communication network and a transmitter associated with the intersection. Traffic signal profiles for the neighbouring intersections are determined by performing the above steps for each of the intersec- tions based on the traffic signal profile of the intersection. This is further communicated to intersections that are adjoining to the neighbouring intersections and repeated till traffic signal profiles for all intersections in the geo- graphical location are computed and a co-ordinated green is achieved. The term "co-ordinated green" indicates synchronized green signal profile across intersections in the geographical location so that the traffic flow is continuous with minimum wait time. The co-ordinated green is an optimization problem to achieve as many coordinated green signals across intersections as possible to ensure smooth traffic flow based on various parameters such as time of the day and forecasted traffic density. Additionally, the co-ordinated green indicates maximizing the distance travelled before having to wait at an intersection. The method effectively achieves the co-ordinated green in a de-centralized manner by determining the traffic signal profiles at each of the intersections based on the traffic signal profile of the neighbouring intersections.

According to an embodiment of the present invention, the method includes capturing the traffic environment in realtime for the intersection and determining lane occupancy for two or more lanes associated with the intersection based on the traffic environment. The traffic environment can be captured using multiple capture devices in real-time. In an embodiment, the traffic environment captured by means of a remote server is relayed in real-time. In another embodiment, the captured traffic environment is normalized to ensure that the lane occupancy is accurately determined notwithstanding weather or light conditions. The captured traffic environment is analyzed to determine the lane occupancy. This analysis can be performed by using neural networks on images captured. In an exemplary embodiment, the traffic environment is analyzed using convolutional neural network. The convolutional neural network is advantageous as it is able to accurately determine lane occupancy in a chaotic traffic environment.

The method also includes determining traffic objects in the traffic environment based on a predetermined identification model. As used herein "traffic objects" include vehicle type, vehicle occupancy and pedestrian type. For example, the traf- fic objects includes and is not limited to a car, bike, scooter, three-wheeler, bus, public transport vehicle, ambulance, vehicle for physically challenged. Further, the traffic objects include type of pedestrian such as elder pedes- trian, physically-handicapped pedestrian, visually impaired pedestrian, child pedestrian and animal. The predetermined identification model includes training data sets, in which valid vectors to be positively recognized and counter examples are rejected. At the end of training an appropriate num- ber of neurons with appropriate influence fields are trained to recognize the objects, patterns, and decisions required to identify the traffic objects. Accordingly, using the predetermined identification model results in back propagation in which weighted inputs to neuron are determined by an itera- tive process. The predetermined identification model results in a decision, defined by a prototype example of the correct recognition of the traffic object.

The method further includes determining traffic anomaly asso- ciable with the intersection. As used herein, "traffic anomaly" means accident, vehicle break-down, traffic violations, etc. In an embodiment, the traffic anomaly can be at a location that could cause irregular traffic flow at the intersection. For example, an accident at a neighbouring intersection could cause re-routing of traffic on one lane. Accordingly the traffic signal profile for the lane must be altered based on the change in traffic on the lane. Therefore, the method is advantageous as it is able to quickly adapt to the changes in the traffic environment and accordingly selects the traf- fic signal profile for the intersection.

According to another embodiment, the traffic environment is captured in real-time using one or more capture devices placed at a plurality of capture points. The capture devices can be placed on street lights, drones, traffic sign-post at the intersection, street hoarding. By distributing the capture devices along lanes of the intersection, the traffic environment is captured evenly across the lanes of the inter- section. This enables precise determination of the traffic density at the intersection. As used herein, the capture devices can be manned device or unmanned device provided with an image sensor, a motion sensor, a Global Positioning System (GPS) device. For example, the capture devices can be cameras located on traffic-sign posts, street lights, drones, etc. In another example, the capture devices can be mobile computing devices with GPS sensors that are able to relay details of wait-time time at a position due to heavy traffic congestion.

According to yet another embodiment, the method includes capturing media data comprising social media corresponding to the intersection. For example, the media and social media sites provide real-time information of traffic at various ge- ographical locations. A server parses through the media data to detect if there is the traffic anomaly associable to the intersection. This information is communicated from the server to a computing device located at the intersection, which selects the traffic signal profile for the intersection.

According to an embodiment, the method includes generating a traffic image of the traffic environment. The traffic image is a representation of the traffic environment. The traffic image is processed to determine the traffic density, lane oc- cupancy and traffic objects. The method also includes recognizing the traffic objects independent of vehicle position, relative vehicle proximity and pedestrian position based on the predetermined identification model. The predetermined identification model is trained considering vehicle position and relative vehicle proximity in chaotic traffic environments. For example, in chaotic traffic environment, induction loops provided on the road will not be able to accurately determine the traffic density. The predetermined identification model maps the traffic objects by using neural networks. This enables accurate recognition of the traffic objects and thereby ensures that the traffic density is estimated correctly .

In an exemplary embodiment, the predetermined identification model is trained by inputting a set of parameters to one or more nodes of an input layer of a neural network, such as convolutional neural network. The set of parameters includes identification parameters to identify vehicles and pedestrians such as shape of the vehicle, shape of the pedestrian, position of the vehicle, position of the pedestrian, vehicle height, vehicle width, vehicle depth and pedestrian height. The traffic objects are recognized by computing product of the set of parameters and associated weighted coefficient at each of the nodes of input layer, and computing sum of the products of the set of parameters and the weighted coefficient at respective nodes of a first intermediate layer of the neural network.

Further, the traffic objects are recognized by computing product of the sum of products and associated weighted coefficient at each of the nodes of the intermediate layer. With the increase in the identification parameters, additional intermediate layers are added to the neural network to accurately recognize the traffic objects. Thereafter, the traffic objects are recognized by co-relating the sum of the products using a look up table.

In another embodiment, the predetermined identification model is an un-trained model that is generated using Generative Ad- versarial Networks (GANs) or Deep Convolutional Generative Adversarial Networks (DCGANs) . The GAN/DCGAN tries to learn joint probability of the traffic image and the traffic objects simultaneously. The GAN/DCGAN model is advantageous as the underlying structure of the traffic image is mapped even when there are no identification parameters. This is desirable when a new vehicle type is captured at the intersection.

According to yet another embodiment, the method includes passing the traffic image through a plurality of layers of a neural network, wherein the plurality of layer comprises convolution layers and pooling layers. In an embodiment, the traffic image is represented as a 3-dimensional array of pixels having intensities for each parameter such as colour, height, width. The 3-dimensional array is transformed through convolution layers and pooling layers using a max-function . The max-function is used to aggregate a maximum value for a spatial region across the convolution layers. According to an embodiment, the method includes comparing the traffic density with the forecast traffic density to generate a traffic difference. The traffic difference is compared with a threshold to determine if the traffic difference has exceeded the threshold. The threshold is determined based on the traffic environment for the intersection. The traffic signal profile for the intersection is selected based on the forecasted traffic density if the traffic difference is within the threshold. For example, if the traffic at the intersection is heavily congested at 1800 hrs on every Monday of first three weeks, the forecast traffic density is determined based on the historical traffic environment that suggests the traffic will be heavily congested at 1800 hrs. The traffic signal profile for 1800 hrs is selected based on the forecast traffic density. At 1800 hrs, the traffic density is deter- mined in real-time and compared with the forecast traffic density. If the traffic density is comparable with the forecasted traffic density, the traffic signal profile selected based on the forecast traffic density is used. If the traffic density varies from the forecast traffic density, then the traffic signal profile is recomputed.

According to another embodiment, the method includes deter- mining an optimal traffic signal profile for the intersection, if the traffic difference exceeds the threshold. The optimal traffic signal profile is based on the traffic difference. In an embodiment, the optimal traffic signal profile and the traffic difference are arranged using a look up ta- ble. The look up table is generated based on the historical traffic environment for the intersection.

The optimal traffic signal profile is tested in real-time by identifying if a traffic anomaly occurs at the intersection. For example, if the optimal traffic signal profile has a signal cycle time of 60 seconds before the next green and a traffic anomaly is identified at the intersection. The traffic density is then checked to see if there is traffic congestion. Accordingly, it can be determined that the traffic anomaly at the intersection can be associated with the optimal traffic signal profile. If the traffic anomaly is identified, then the optimal traffic signal profile is re-computed. If there is no traffic anomaly identified then the optimal traffic signal profile is selected as the traffic signal pro- file.

According to an embodiment, the method includes communicating the traffic signal profile of the intersection to neighbouring intersections. Thereafter, traffic signal profiles for the neighbouring intersections are generated based on the communicated traffic signal profile of the intersection. This step is repeated across all the intersections in the geographical location. Accordingly, the method includes itera- tively generating traffic signal profiles for a plurality of intersections based on the traffic signal profiles for the neighbouring intersections. Further, the method includes managing the traffic flow for the plurality of intersections based on the traffic signal profiles for the neighbouring intersections by generating the co-ordinated green for the geographical location. The method effectively achieves the coordinated green in a de-centralized manner by determining the traffic signal profiles at each of the intersections based on the traffic signal profile of the neighbouring intersections.

According to the present invention also disclosed are a computing device and a system for managing traffic flow in a geographical location. The computing device includes one or more processors, a memory coupled to the processors, a communication unit and a database. The memory is configured to store computer program instructions defined by modules, for example, traffic density estimator, traffic prediction module, profile determination module, etc. Each processor is configured to execute the instructions in the memory. The memory may be volatile memory such as random access memory (RAM) , static RAM (SRAM) , dynamic RAM (DRAM) or any other memory accessible by the processor. The memory may operate as a routing table, register, virtual memory and/or swap space for the processor. Further, the storage module may represent any known or later developed non-volatile storage device such as, but not limited to, a hard drive, a solid-state drive (SSD) , flash ROM (read-only memory) or an optical drive.

In an embodiment, the computing device includes an edge device and a capturing means communicatively coupled to the edge device. The edge device is a compact computing device that has resource constraints in terms of computing power. The computing device includes a traffic density estimator configured to determine a traffic density in real time based on the traffic environment for an intersection and a traffic prediction module configured to predicting a forecast traffic density based on the traffic environment and a historical traffic environment associated with the intersection. Further, the computing device includes a profile determination module configured to selecting a traffic signal profile for the intersection based on the traffic density and the forecast traffic density.

The communication unit of the computing device is capable of communicating the traffic signal profile of the intersection to a traffic sign-post associated with the intersection for managing the traffic flow for the traffic environment. Accordingly, the computing device is advantageous as it is able to effectively manage traffic flow in the traffic environment .

According to an embodiment, the computing device includes a traffic analyzer configured to determine lane occupancy for two or more lanes associated with the intersection based on the traffic environment. The computing device also includes a traffic image generator configured to generate a traffic image of the traffic environment and a traffic object determination module configured to determine traffic objects in the traffic environment based on a predetermined identification model .

According to another embodiment, the traffic object determination module includes a neural network module configured to pass the traffic image through a plurality of layers of a neural network, wherein the plurality of layer comprises convolution layer and max pooling layer. According to an embodiment, the traffic analyzer includes a comparator configured to compare the traffic density with the forecast traffic density, to generate a traffic difference. Also, the traffic analyzer includes a threshold module to determine whether the traffic difference has exceeded a threshold, wherein the threshold is determined based on the traffic environment for the intersection. According to another embodiment, the traffic analyzer includes an optimal profile module configured to determine an optimal traffic signal profile for the at least one intersection, if the traffic difference exceeds the threshold. The traffic analyzer also includes an anomaly identifier config- ured to identifying a traffic anomaly at the at least one intersection associable to the optimal traffic signal profile determined by the anomaly identifier. The optimal traffic signal profile re-determined if the traffic anomaly is identified and wherein the optimal traffic signal profile is se- lected as the traffic signal profile if the traffic anomaly is not identified.

According to an embodiment, the communication unit includes a transmitter communicatively coupled to the processor to transmit the traffic signal profile of the intersection to neighbouring intersections. The communication unit also includes a receiver communicatively coupled to the process, to receive traffic signal profiles of the neighbouring intersections, wherein the traffic signal profile of the at least one intersection is based on the traffic signal profiles of the neighbouring intersections.

The present invention also provides a system for managing traffic flow in a geographical location. The system includes a server, a network communicatively coupled to the server and a computing device communicatively coupled to the server via the network. The computing device associated with an intersection in the traffic environment. The computing device is capable of managing traffic flow at the intersection based on the traffic environment at the intersection.

According to an embodiment, the system includes a plurality of capture devices communicatively coupled to the server via the network, the plurality of capture devices are one of manned devices and unmanned devices selected comprising image sensors, a motion sensors, a GPS devices and communication devices. The plurality of capture devices are capable of capturing media data comprising social media corresponding to the at least one intersection.

According to another embodiment, the server includes a database that stores data associated with the traffic environment, wherein the traffic environment comprises data associated with weather, air quality, time, day, scheduled events and unscheduled events for at least one geographical location comprising a plurality of intersections.

According to yet another embodiment, the server includes a traffic model generator to generate a predetermined identification model based on the traffic environment. The predetermined identification model includes training data sets, in which valid vectors to be positively recognized and counter examples are rejected, at the end of which an appropriate number of neurons with appropriate influence fields are trained to recognize the objects, patterns, and decisions required to identify the traffic objects. Accordingly, using the predetermined identification model results is back propagated in which weighted inputs to neuron are determined by an iterative process. The predetermined identification model results in a decision, de-fined by a prototype example of the correct recognition of the traffic object.

The server further includes a forecast model generator to generate a forecast model based on a historical traffic environment. The historical traffic environment includes information relating traffic at the intersection for a time instant in the past. In an embodiment, the forecast model generator generates the forecast model using neural networks, such as recurrent neural network (RNN) . The advantage of recurrent neural network is that RNN maintains an internal state that is utilized for forecasting sequence values of the forecast traffic density. In RNN the present value of the forecast traffic density depends on a previous state. For example, if there is traffic congestion at an intersection due to an accident or an unplanned gathering/event in the vicinity, then the forecast traffic density in the next time instant would be highly dependent on the current time instant. Therefore, it can be observed that the forecast traffic density at the intersection at time t+1 has a strong correlation with the density at time t and hence, RNN is used to predict the forecast traffic density accurately.

The above-mentioned and other features of the invention wil now be addressed with reference to the accompanying drawing of the present invention. The illustrated embodiments are in tended to illustrate, but not limit the invention.

The present invention is further described hereinafter with reference to illustrated embodiments shown in the accompanying drawings, in which: FIG 1 is a block diagram of a system for managing traffic flow for a geographical location;

FIG 2 is a block diagram of an edge device for managing traffic at the intersection;

FIG 3 illustrates determination of lane occupancy at an intersection; FIG 4 illustrates recognition of traffic objects by the edge device of FIG. 2;

FIG 5 illustrates generation of a forecast model by the system of FIG 1;

FIG 6 is a block diagram of a system for managing traffic flow for a plurality of intersections;

FIG 7 illustrates a method of managing traffic flow for the plurality of intersection using the system according to FIG 6; and

FIG 8 illustrates a method of managing traffic flow in a traffic environment.

Various embodiments are described with reference to the drawings, wherein like reference numerals are used to refer to like elements throughout. Further, numerous specific details are set forth in order to provide thorough understanding of one or more embodiments of the present invention. These examples must not be considered to limit the application of the invention to configurations disclosed in the figures. It may be evident that such embodiments may be practiced without these specific details. FIG 1 is a block diagram of a system 100 for managing traffic flow for a geographical location including an intersection 130. As shown in FIG. 1, traffic at the intersection 130 includes vehicular traffic and pedestrian traffic. The intersection 130 is connected to lanes 132, 134, 136 and 138. The intersection 130 includes a traffic sign-post 140, to indicate whether the traffic at the lanes 132, 134, 136 and 138 should stop (red), proceed cautiously (yellow) or proceed (green) . Further, the intersection 130 is provided with a camera 120, to capture traffic environment associated with the intersection 130.

The system 100 includes a server 102, a network 108, a database 110, the camera 120 and a computing device 150. The system 100 also includes computing devices 152 and 154 that are associated with neighbouring intersections.

As shown in FIG 1, the server 102 is communicatively coupled to the database 110. The database 110 is, for example, a structured query language (SQL) data store or a not only SQL (NoSQL) data store. In an embodiment, the database 110 may be a location on a file system directly accessible by the server 102. In another embodiment, the database 110 may be configured as cloud based database implemented in a cloud computing environment, where computing resources are delivered as a service over the network 108. As used herein, "cloud computing environment" refers to a processing environment comprising configurable computing physical and logical resources, for example, networks, servers, storage, applications, services, etc., and data distributed over the network 108, for example, the internet. The cloud computing environment provides on-demand network access to a shared pool of the configurable computing physical and logical resources. The net- work 108 is, for example, a wired network, a wireless network, a communication network, or a network formed from any combination of these networks.

The server 102 includes a controller 104 and a memory 106. The server 102 is communicatively coupled to the network 108. The memory 106 is configured to store computer program instructions defined by modules, for example, traffic model generator 112 and forecast model generator 114. In the present embodiment, modules 112 and 114 are implemented on the cloud computing environment as indicated in FIG 1.

Further, FIG 1 shows that the computing device 150 includes a traffic object determination module 162, a traffic density estimator 164 and a profile determination module 166. The traffic object determination module 162 receives a predetermined identification model from the traffic object generator 112 as indicated by arrow 144. The captured traffic environment from the camera 120 is used by the traffic object generator module 162 to determine traffic objects, as indicated by arrow 121. The traffic density estimator 164 receives information from capture devices 122 and 124 located at the intersection, as indicated by arrow 142. In the present example the capture device 122 is a car Global Positioning System (GPS) device 122. The capture device 124 is a mobile computing device 124 that can be used to communicate real-time traffic data to the computing device 150, directly or via the server 110. The profile determination module 166 receives real-time traffic density from the traffic density estimator 164. The profile determination module 166 also receives realtime data associated with the traffic environment at the intersection 130, as indicated by arrow 146. For example, the profile determination module 166 receives real-time data related to weather and pollution level at the intersection 130, from the server 110. The profile determination module 166 compares the traffic density in real-time versus a forecasted traffic density based on a forecast model received from the forecast model generator 114, indicated by arrow 148. The profile determination module 166 selects a traffic signal profile for the intersection 130 based on the traffic density and the forecast traffic density. The selected traffic signal profile is reflected on the traffic sign-post 140, as indicated by the arrow 160. The computing device 150 includes other components such as processor, memory and communication unit, as described in detail in FIG 2.

FIG 2 is a block diagram of an edge device 200 for managing traffic at the intersection 130. The edge device 200 can be used in place of the computing device 150. The edge device 200 has a small form factor that is capable of connecting with the network 108. The edge device 200 includes a processor 202, a memory 204 and a communication unit 208. The memory 204 may include 2 Giga byte Random Access Memory (RAM) Package on Package (PoP) stacked and Flash Storage. The communication unit 208 includes a transmitter 248, a receiver 258b and Gigabit Ethernet port. The edge device 200 also includes a High-Definition Multimedia Interface (HDMI) display 206 and a cooling fan (not shown in the figure) .

The memory 204 of the edge device 200 is provided with modules stored in the form of computer readable instructions, for example, 210, 212, 214, 216, 218 and 220, . The processor 202 is configured to execute the defined computer program instructions in the modules. Further, the processor 202 is configured to execute the instructions in the memory 204 simultaneously. The operation of the edge device 200 in the system 100 is explained as under. The camera 120 relays real-time data of the traffic environment at the intersection 130. The traffic environment is processed by the traffic image generator 210 to generate a traffic image of the traffic environment for each time instant. The traffic analyzer 212 analyzes the traffic image to determine lane occupancy for the lanes 132, 134, 136 and 138. The determination of the lane occupancy is further explained in FIG 3. In an embodiment, the traffic analyzer 212 analyzes the traffic environment captured by the camera 120 to determine the lane occupancy. The traffic analyzer 212 includes a comparator 222, a threshold module 224, an optimal profile module 226 and an anomaly identifier 228. Thereafter, the traffic object determination module 214 determines traffic objects in the traffic environment based on a predetermined identification model. The traffic objects include vehicle type, vehicle occupancy and pedestrian type. The predetermined identification model is a training model generated by the traffic model generator 112 using a neural network module 215. The process of determining the traffic objects by the traffic object determination module 214 is explained further in FIG . After the traffic objects are recognized by the traffic object determination module 214, the traffic density estimator 216 determines the traffic density in real time. The number of vehicles and pedestrians at the intersection is computed. Also, the number of vehicles in each lane 132, 134, 136 and 138 is computed by the traffic density estimator 216. Further, the vehicles are classified based on their type, such as cars, bikes, bus, public transport, private transport, emergency service vehicle, etc. The number of vehicles in each of the classifications is also computed. When the edge device 200 is determining the traffic density, the traffic prediction module 218 predicts a forecast traffic density based on the traffic environment and the historical traffic environment associated with the intersection 130. Particularly, the traffic prediction module 218 generates the forecast traffic density based on the forecast model from the forecast model generator 114. The generation of the forecast traffic density is further explained in FIG 5.

The traffic density and the forecast density are then compared by the profile determination module 220. Based on the comparison, the traffic signal profile for the intersection 130 is selected. In an embodiment, a traffic difference be- tween the traffic density and the forecast traffic density is determined. If the traffic difference is within a threshold generated by the threshold module 224 of the traffic analyzer 212, the profile determination module 220 selects the traffic signal profile based on the forecasted traffic density. If the traffic difference is above the threshold generated, the profile determination module 220 computes an optimal traffic signal profile using optimal profile module 226. The optimal traffic signal profile is then tested by determining whether any traffic anomaly is present at the intersection 130 due to the optimal traffic signal profile. The traffic anomaly is identified by the anomaly identifier 228. The optimal traffic signal profile re-determined if the traffic anomaly is identified. If the traffic anomaly is not identified the optimal traffic signal profile is selected as the traffic signal pro- file.

FIG 3 illustrates determination of lane occupancy at an intersection. As shown in the figure, the intersection includes four lanes 302, 304, 306 and 308. Each lane 302, 304, 306 and 308 is associated with traffic signal profiles 312, 314, 316 and 318. To determine the lane occupancy, traffic environment including the vehicular traffic is captured by a capture means, such as the camera 120 and other devices 122 and 124.

As seen in FIG 3, each lane 302, 304, 306 and 308 is divided in four segments and a traffic density in each segment is determined. For example, in lane 302, the traffic density in a first segment 302a is 67.9%, a second segment 302b is 23.9%, third segment 302c is 5.3% and fourth segment 302d is 0%. Similarly, the traffic density is determined for segments 304a, 304b, 304c and 304d as 35.6%, 95.4%, 82.1% and 2%. This determination is done for lanes 306 and 308, in which segments 306a, 306b, 306c and 306d have traffic density of 82.3%, 84.8%, 23.0% and 0%, respectively. Segments 308a, 308b, 308c and 308d have traffic density of 11.1%, 20.8%, 3.2% and 0%.

Generally, the traffic environment is captured only close to the intersection and the traffic flow is managed based on the traffic density closest to the intersection. If such a method is employed, the traffic density at segment 306a is the highest and therefore, lane 306 will be given a green traffic signal profile. However, as seen the traffic density in lane 304 is higher than that of 306. According to prior art, the lane 304 is given the green traffic signal profile instead of lane 306. Therefore, the present invention is advantageous as it accurately determines the lane occupancy as a whole and also determines lane occupancy in each segment of a lane and comparing the same across the respective segments in other lanes. Also, the use of other capture devices such as GPS devices, mobile phone or cameras located on street lights, the traffic environment at each lane is captured more accurately and results in better management of the traffic flow. FIG 4 illustrates recognition of the traffic objects by the edge device 200. The edge device 200 uses neural networks to recognize the traffic objects. In the present embodiment, convolutional neural networks are used to recognize the traffic objects. As shown in the figure, a traffic image 450 of the traffic environment at an intersection is taken. The convolutional neural network extracts relevant information from pixels of the traffic image 450 and inputs the same into a fully-connected neural network with a softmax output layer 414 yielding the traffic density of the considered vehicle types, e.g. four-wheelers, two-wheelers, and a special type "other" for other vehicle types or areas without vehicles. The convolutional neural network is trained on a predetermined identification model including a dataset of images mapped to the traffic objects.

In particular, the traffic image 450 is represented as a 3- dimensional array of pixels intensities for 3 array dimensions for feature maps including colour, height, and width. The traffic image 450 is transformed through convolutional feature extraction layers according to the following equation : where I denotes the layer index, k denotes the feature map index, ₀ corresponds to the image pixel array,

are the filters and biases, which correspond to the Z-th layer and /c-th feature map, learned from training examples, and f is an element wise function such as tanh(x) or max(0,x) .

As show FIG 4, pooling layers 404 and 408 are used subsequent to convolutional layers 402, 406 and 410. The pooling layers 404 and 408 aggregate spatially local regions using a max- function i.e. the maximum value of the spatial local region is selected. For example, spatially local regions of size 2x2 may be aggregated using the max-function, i.e. the maximum value of the 2x2 region is selected. Common aggregation functions are the maximum or average function, but other functions are possible.

After the convolutional layers 402, 406 and 410 with the pooling layers 404 and 408, the resulting 3-dimensional array is either flattened to a vector of length number of feature maps including colour, height and width or a global pooled layer 412. In the present embodiment, the global pooled layer 412 yields a vector of length number of the feature maps. In another embodiment, the final feature map is flattened to a vector, the resulting vector may be transformed further by means of one or multiple subsequent fully-connected layers according to the following equation:

t = W_t + bd

where / denotes the layer index, ₀ corresponds to said vector, Wi and fe; are the weight matrix and biases vector in the Z-th layer learned from the predetermined identification data, and is an element wise function such as tanh(x) or max(0, x) . In the softmax output layer 414, is chosen to be the s oftmax-function :

[softmax(x)]i =

The softmax-function yields a distribution of the traffic density of the vehicle types based on the predetermined identification model.

In an embodiment, the predetermined identification model includes mapping of the traffic objects for all weather and light conditions. In another embodiment, the traffic image 450 is normalized for brightness prior to passing them into the convolutional neural network. This normalization step is also performed while generating the predetermined identification model .

FIG 5 illustrates generation of a forecast model by the system 100. In the present embodiment, a recurrent neural network 500 is used to generate the forecast model for the system 100. The recurrent neural network with time lag is used to model nonlinear relations between forecast traffic density with historical traffic density and other external inputs.

According to the recurrent neural network, the following inputs are considered:

Weather input 502(xl(t)) including weather data, such as temperature, precipitation, and humidity are taken as the weather input 502. The weather input 502 includes weather data for any period in the past.

Pollution input 504 (x2 (t) ) including air quality data, such as amount of S02, CO, N02 in the atmosphere.

Time input 506 (x3(t)) including data of time of the day. Traffic density has a strong correlation with the time of day, for example at peak hours (8 - 11 AM and 6 - 8 PM) . Accordingly, the forecast traffic density includes likelihood of congestion based on the time of the day.

Day of week 508 (x4 (t) ) is considered as an input. During the weekends it is observed that the traffic density is relatively lower. Hence, the day of week 508 is considered while generating the forecast model.

Holiday input 512 (x5 (t) ) includes information relating to scheduled holidays. The holiday input 510 is used to train the forecast model with the intelligence that during the holiday season it is observed that traffic on highways leading out of town is higher. Event input 514 (x6 (t) ) including planned and unplanned events that could affect traffic flow is considered. For example, if information on planned road repairs near the intersection is available then the traffic congestion can be taken into account while generating the forecast model and thereby the forecast traffic density.

Neighbouring traffic density input 516 (x7 (t) ) . By consider^¬ ing the neighbouring traffic densities at neighbouring intersections, a co-ordinated green traffic signal profile can be achieved. The system 100 includes several computing devices 150, 152 and 154 at different intersections to manage the traffic flow. Since this is a decentralized method, it is advantageous to consider the neighbouring traffic density because if there is traffic congestion at a neighbouring inter- section there is a high probability that it may percolate to the other intersections in the geographical location.

Pedestrian density 518 (x8 (t) ) including the type of pedestrians and pedestrian occupancy at the intersection is considered. The pedestrian density 518 is generated by employing a neural network on a traffic image of the traffic environ^¬ ment at the intersection.

Historic traffic density 522 (x9 (t) ) is also considered as an input for the recurrent neural network to generate the forecast model. The historic traffic density 522 at the intersec- tion is used to predict the forecast traffic density.

As shown in FIG 5, hidden neurons 1 to n are indicated as 510-510n. Further, weight of the hidden neurons 1 to n and recurrent weight are denoted as ar_n 530 and a 540, respec- tively. Predicted outputs of hidden layers are fed back as inputs at context neurons y(t) 525 for each time step. The recurrent neural network 500 determines weighted sum of hidden neurons and output of hidden neuron indicated by 550-550n. The weighted sum of hidden neuron n is u(t + 1) = a_tl x_t(t) + + ar_n y_n(t) . The output of hidden neuron n is y(t + 1) =

/(i<(t + l)), where / is a non-linear transfer function. It can be observed from the above equation that the output at t+1 is dependent on the past predicted values along with the current input values. For the output neuron, the forecast traffic density 580 is z(t + 1) = /(∑(ø y(t + 1)). The forecast model is deployed on a computing device, such as edge device 200 in a runtime container commonly referred to as Analytical Model Container Framework (AMCF) . In an embodiment, during operation the forecast model receives historical data for ever hour that is used to generate the forecast traffic density for the next hour.

FIG 6 is a block diagram of a system 600 for managing traffic flow for a plurality of intersections 602, 604 and 606. Each intersection 602, 604 and 606 includes multiple lanes 610a- 610n, 620a-620n and 630a-630n. Also, the intersections 602, 604 and 606 include traffic sign-posts 618, 628 and 638 to manage traffic flow in the respective lanes 610a-610n, 620a- 620n and 630a-630n.

Further, each intersection 602, 604 and 606 includes respec- tive traffic agent 615, 625 and 635. The traffic agents 615, 625 and 635 include a computing device, such as edge device 200 and capturing device, such as camera 120. The traffic agents 615, 625 and 635 are connected to signal controllers 616, 626 and 636, respectively. Functionality of the traffic agents 615, 625 and 635 is indicated by functional blocks, which are used to determine traffic signal profiles for each of the intersections 602, 604, 606. The traffic agent 615 includes a traffic density estimator 612, a traffic prediction module 614 and a profile determination module 613. Similarly, the traffic agent 625 includes a traffic density estimator 622, a traffic prediction module 624 and a profile determination module 623. The traffic agent 635 includes a traffic density estimator 632, a traffic prediction module 634 and a profile determination module 633.

As shown in FIG 6, the traffic agent 615 selects a traffic signal profile for each lane 610a-610n by determining a traf- fic density in real time based on the traffic environment for the intersection 602 using the traffic density estimator 612. The traffic agent 615 predicts a forecast traffic density based on the traffic environment and a historical traffic environment associated with the intersection using the traffic prediction module 614. Further, the traffic agent 615 selects a traffic signal profile for the intersection based on the traffic density and the forecast traffic density using the profile determination module 613. The above steps are done simultaneously for the lanes 610a-610n in the intersection 602. The traffic agents 625 and 635 select traffic signal profiles for intersections 604 and 606 as described for the traffic agent 625. The traffic agents 615, 625 and 635 communicate the selected traffic signal profiles to the signal controllers 616, 626, and 636. Additionally, the traffic agents 615, 625 and 635 communicate the selected traffic signal profiles to each other, so that a co-ordinated green can be achieved across the intersections 602, 604 and 606. The co-ordinated green is an optimization problem to achieve as many coordinated green signals across the intersections 602, 604 and 606. The method of communicating the selected traffic signal profiles to achieve the co-ordinated green is explained in FIG 7. FIG 7 illustrates a method 700 of managing traffic flow for a plurality of intersections 702, 704 and 706 using the system 600. As shown in the figure, the plurality of intersections 702, 704 and 706 are adjacent to each other in a geographical location 710. Each intersection 702, 704 and 706 includes a traffic agent 715, 725 and 735.

The traffic agent 715 observes traffic environment for the intersection 702 and determines the traffic density at time^¾t' using convolutional neural network, as indicated by the arrow 711. Further, the traffic agent 715 also receives forecast model and determines the forecast traffic density for time't', as indicated by the arrow 718. A traffic policy 713 is selected at based on the traffic density and the fore- cast traffic density. Based on the traffic policy 713, traffic signal profile is selected at 717. The traffic signal profile is reflected at a traffic sign-post for the intersection 702, indicated by the arrow 720. Also, the traffic signal profile is communicated to the traffic agent 725 at the adjacent intersection 704, as indicated by the arrow 719.

Similarly, the traffic agent 725 observes traffic environment for the intersection 704 and determines the traffic density at time't+l' using convolutional neural network, as indicated by the arrow 721. Further, the traffic agent 725 also receives forecast model and determines the forecast traffic density for time't' , as indicated by the arrow 728. A traffic policy 723 is selected at based on the traffic density and the forecast traffic density. Based on the traffic policy 723, traffic signal profile is selected at 727. The traffic signal profile is reflected at a traffic sign-post for the intersection 704, indicated by the arrow 730. Also, the traffic signal profile is communicated to the traffic agent 735 at the adjacent intersection 706, as indicated by the arrow 729.

The system 600 employs a multi-agent deep reinforcement learning algorithm is used to learn optimal behaviour of the plurality of intersections 702, 704 and 706, such that the co-ordinated green is achieved. The algorithm trains a global policy that includes the traffic policy (i.e. sub-policy) 713 and 723 per traffic agent 715 and 725. The traffic policy for traffic agent 735 is not shown in FIG 7, however, it is generated similar to that of 713 and 723. The traffic policy 713 and 723 are the traffic signal profiles for the intersection 702 and 704 that are determined by the traffic agent 715 and 725.

To generate the global policy, the system 600 represents the traffic environment that is observed or captured as Of. Fur^¬ ther, the forecast model and the forecast traffic density are represented as mf . The traffic policies 713 and 723 are rep- resented as p and the traffic signal profile selected by each traffic agent 715, 725 and 735 is represented as . In the above notations, t indicates time and a indicates a lane in the intersection 702, 704 and 706. Further, each traffic agent 715, 725 and 735 observes the current state (i.e. traffic environment) s_t e S at time t' and chooses the traffic signal profile u_t e U according to the traffic policy p . The traffic agent 715, 725 and 735 also observes a reward signal r_t and transitions to a new traffic environment s_t+1. The reward signal r_t is determined by identifying whether a traffic anomaly has occurred in the geographical location 710 due to the traffic signal profile u_t . The system 600 tries to achieve the co-ordinated green by maximizing number of continuous green traffic signal profiles over a discounted return. The system 600 performs the optimization using the equation R_t = r_t + ?r_t + ?r_t+1 + βτ?₊₂ + where r_t is the reward received at time t and /?e[0, l]is a discount factor. The discount factor β is dependent on the traffic environment of each intersection. For example, the discount factor β may be more if the intersection is on a highway and if it is weekend where there are more vehicles moving between cities.

The system 600 also uses Q-function to determine the coordinated green. The Q-function of the traffic policy p is Q (S, U) , i.e. a set including the traffic environment and the traffic signal profiles. The Q-function includes Q values for the traffic environment, which is represented as Q_u . Further, the Q-function also includes Q values for traffic signal profile of the neighbouring intersection (indicated by arrows 719 and 729), which is represented by Q_m . The system 600 then selectively picks uf and mf from Q values for the traffic environment Q_u and Q values for traffic signal profile of the neighbouring intersection Q_m using Epsilon greedy policy. The Epsilon greedy policy is a way of selecting a random action i.e. a traffic signal profile with uniform distribu- tion from a set of available actions i.e. traffic signal profiles .

FIG 8 illustrates a method 800 of managing traffic flow in a geographical location. The method begins, at step 802, with capturing a traffic environment in real-time for an intersection in the geographical location. The traffic environment is captured using a camera placed on street lights, traffic sign-post at the intersection, street hoarding. The traffic environment can also be determined from a camera provided on an unmanned device associated with the intersection.

At step 804, lane occupancy for two or more lanes associated with the intersection is determined based on the traffic environment. Further, at step 805, traffic objects in the traffic environment are determined based on a predetermined identification model. The traffic objects include vehicle type, vehicle occupancy and pedestrian type. The traffic objects are identified by generating a traffic image of the traffic environment and recognizing the traffic objects independent of vehicle position, relative vehicle proximity and pedestrian position based on the predetermined identification model. In an example embodiment, the traffic objects are identified by passing the traffic image through a plurality of layers of a neural network. The plurality of layers comprises convolution layers and pooling layers .

At step 806, information regarding the traffic environment is captured from using devices such as GPS enabled device, mobile computing devices. The information is also captured from media including social media. This information is used to determine if there is any traffic anomaly associable with the intersection at step 808. The traffic anomaly includes acci- dent, vehicle break-down, traffic violations. At step 810, a traffic density is determined in real time based on the traffic environment for the intersection.

The method further includes, predicting a forecast traffic density based on the traffic environment and a historical traffic environment associated with the intersection at step 820. The forecast traffic density is computed based on several parameters associated with the traffic environment and the historical traffic environment as indicated in step 815. The forecast traffic density is computed using a recurrent neural network as indicated at step 825.

Furthermore, the method includes at step 860, selecting a traffic signal profile for the intersection based on the traffic density and the forecast traffic density. To select the traffic signal profile, the traffic density is compared with the forecast traffic density, to generate a traffic difference. Further, at step 830 it is determined whether the traffic difference has exceeded a threshold. The threshold is determined based on the historical traffic environment for the intersection. If the traffic difference is less than the threshold, the traffic signal profile is selected based on the forecast traffic density. The traffic signal profile is output via a traffic sign-post for the intersection at step 850.

If the traffic difference is more than the threshold, then at step 840, the method determines an optimal traffic signal profile for the intersection, the optimal traffic signal profile is based on the traffic difference. If the traffic difference is significantly greater than the threshold, the optimal traffic signal profile will vary. The optimal traffic signal profile is tested in real-time by identifying if any traffic anomaly is present at the intersection. If the traffic anomaly is present, then the optimal traffic signal profile is recomputed till no traffic anomaly is identified. If there is no traffic anomaly, the optimal traffic signal is output to the traffic sign-post at step 850.

The above disclosed method, device and system may be achieved via implementations with differing or entirely different components, beyond the specific components and/or circuitry set forth above. With regard to such other components (e.g., cir- cuitry, computing/processing components, etc.) and/or computer-readable media associated with or embodying the present invention, for example, aspects of the invention herein may be implemented consistent with numerous general purpose or special purpose computing systems or configurations. Various exemplary computing systems, environments, and/or configurations that may be suitable for use with the disclosed subject matter may include, but are not limited to, various clock- related circuitry, such as that within personal computers, servers or server computing devices such as routing/connectivity components, hand-held or laptop devices, multiprocessor systems, microprocessor-based systems, set top boxes, smart phones, consumer electronic devices, network PCs, other existing computer platforms, distributed computing environments that include one or more of the above systems or devices, etc.

In some instances, aspects of the invention herein may be achieved via logic and/or logic instructions including program modules, executed in association with the circuitry, for example. In general, program modules may include routines, programs, objects, components, data structures, etc. that perform particular tasks or implement particular control, delay or instructions. The inventions may also be practiced in the context of distributed circuit settings where circuitry is connected via communication buses, circuitry or links. In distributed settings, control/instructions may occur from both local and remote computer storage media including memory storage devices.

The system and computing device along with their components herein may also include and/or utilize one or more type of computer readable media. Computer readable media can be any available media that is resident on, associable with, or can be accessed by such circuits and/or computing components. By way of example, and not limitation, computer readable media may comprise computer storage media and communication media. Computer storage media includes volatile and non-volatile, removable and non-removable media implemented in any method or technology for storage of information such as computer readable instructions, data structures, program modules or other data. Computer storage media includes, but is not limited to, RAM, ROM, EEPROM, flash memory or other memory technology, CD-ROM, digital versatile disks (DVD) or other optical storage, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other medium which can be used to store the desired information and can accessed by computing component. Communication media may comprise computer readable instructions, data structures, program modules or other data embodying the functionality herein. Further, communication media may include wired media such as a wired network or direct-wired connection, and wireless media such as acoustic, RF, infrared and other wireless media. Combinations of the any of the above are also included within the scope of computer readable media.

In the present description, the terms component, module, device, etc. may refer to any type of logical or functional circuits, blocks and/or processes that may be implemented in a variety of ways. For example, the functions of various circuits and/or blocks can be combined with one another into any other number of modules . Each module may even be implemented as a software program stored on a tangible memory (e.g., random access memory, read only memory, CD-ROM memory, hard disk drive) to be read by a central processing unit to implement the functions of the invention herein. Or, the modules can comprise programming instructions transmitted to a general purpose computer or to processing/graphics hardware via a transmission carrier wave. Also, the modules can be implemented as hardware logic circuitry implementing the functions encompassed by the invention herein. Finally, the modules can be implemented using special purpose instructions (SIMD instructions), field programmable logic arrays or any mix thereof which provides the desired level performance and cost .

As disclosed herein, implementations and features consistent with the present inventions may be implemented through computer-hardware, software and/or firmware. For example, the systems and methods disclosed herein may be embodied in various forms including, for example, a data processor, such as a computer that also includes a database, digital electronic circuitry, firmware, software, or in combinations of them. Further, while some of the disclosed implementations describe components such as software, systems and methods consistent with the invention herein may be implemented with any combination of hardware, software and/or firmware. Moreover, the above-noted features and other aspects and principles of the invention herein may be implemented in various environments. Such environments and related applications may be specially constructed for performing the various processes and operations according to the present invention or they may include a general-purpose computer or computing platform selectively activated or reconfigured by code to provide the necessary functionality. The processes disclosed herein are not inherently related to any particular computer, network, architecture, environment, or other apparatus, and may be implemented by a suitable combination of hardware, software, and/or firmware. For example, various general-purpose machines may be used with programs written in accordance with teachings of the invention herein, or it may be more convenient to con- struct a specialized apparatus or system to perform the required methods and techniques .

Aspects of the method and system described herein, such as the logic, may be implemented as functionality programmed into any of a variety of circuitry, including programmable logic devices ("PLDs"), such as field programmable gate arrays ("FPGAs"), programmable array logic ("PAL") devices, electrically programmable logic and memory devices and standard cell-based devices, as well as application specific integrated circuits. Some other possibilities for implementing aspects include: memory devices, microcontrollers with memory (such as EEPROM) , embedded microprocessors, firmware, software, etc. Furthermore, aspects may be embodied in microprocessors having software-based circuit emulation, discrete logic (sequential and combinatorial), custom devices, fuzzy (neural) logic, quantum devices, and hybrids of any of the above device types. The underlying device technologies may be provided in a variety of component types, e.g., metal-oxide semiconductor field-effect transistor ("MOSFET") technologies like complementary metal-oxide semiconductor ("CMOS"), bipolar technologies like emitter-coupled logic ("ECL"), polymer technologies (e.g., silicon-conjugated polymer and metal- conjugated polymer-metal structures), mixed analog and digital, and so on.

It should also be noted that the various logic and/or functions disclosed herein may be enabled using any number of combinations of hardware, firmware, and/or as data and/or instructions embodied in various machine-readable or computer- readable media, in terms of their behavioural, register transfer, logic component, and/or other characteristics. Computer-readable media in which such formatted data and/or instructions may be embodied include, but are not limited to, non-volatile storage media in various forms (e.g., optical, magnetic or semiconductor storage media) and carrier waves that may be used to transfer such formatted data and/or instructions through wireless, optical, or wired signalling media or any combination thereof. Examples of transfers of such formatted data and/or instructions by carrier waves include, but are not limited to, transfers (uploads, downloads, e- mail, etc.) over the Internet and/or other computer networks via one or more data transfer protocols (e.g., Hypertext Transfer Protocol (HTTP), File Transfer Protocol (FTP), Simple Mail Transfer Protocol (SMTP) , and so on) .

Unless the context clearly requires otherwise, throughout the description and the claims, the words "comprise," "comprising," and the like are to be construed in an inclusive sense as opposed to an exclusive or exhaustive sense; that is to say, in a sense of "including, but not limited to." Words using the singular or plural number also include the plural or singular number respectively. Additionally, the words "herein," "hereunder," "above," "below," and words of similar import refer to this application as a whole and not to any particular portions of this application.

Although certain presently preferred implementations of the present invention have been specifically described herein, it will be apparent to those skilled in the art to which the inventions pertain that variations and modifications of the various implementations shown and described herein may be made without departing from the scope of the inventions herein. Accordingly, it is intended that the inventions be limited only to the extent required by the appended claims and the applicable rules of law.

Claims

1. A method of managing traffic flow in a geographical location, the method comprising:

determining a traffic density in real time based on a traffic environment for at least one intersection (130) in the geographical location, wherein the traffic environment comprises vehicular traffic and pedestrian traffic;

predicting a forecast traffic density based on the traf- fic environment and a historical traffic environment associated with the at least one intersection (130);

selecting a traffic signal profile for the at least one intersection (130) based on the traffic density and the forecast traffic density; and

managing the traffic flow for in the geographical location based on the traffic signal profile of the at least one intersection (130).

2. The method according to claim 1, further comprising: capturing the traffic environment in real-time for the at least one intersection (130);

determining lane occupancy for two or more lanes associated with the at least one intersection (130) based on the traffic environment;

determining traffic objects in the traffic environment based on a predetermined identification model, wherein the traffic objects include vehicle type, vehicle occupancy and pedestrian type; and

determining traffic anomaly associable with the at least one intersection {130), wherein the traffic anomaly comprises accident, vehicle break-down, and traffic violations.

3. The method according to claim 2 , wherein determining traffic objects in the traffic environment based on a prede- termined identification model, further comprises:

generating a traffic image of the traffic environment; and recognizing the traffic objects independent of vehicle position, relative vehicle proximity and pedestrian position based on the predetermined identification model. 4. The method according to claim 3, wherein recognizing traffic objects independent of vehicle position, relative vehicle proximity and pedestrian position, comprises:

passing the traffic image through a plurality of layers of a neural network, wherein the plurality of layer comprises convolution layers and pooling layers.

5. The method according to claim 1, wherein selecting a traffic signal profile for the at least one intersection (130) based on the traffic density and the forecast traffic density, comprises:

comparing the traffic density with the forecast traffic density, to generate a traffic difference;

determining whether the traffic difference has exceeded a threshold, wherein the threshold is determined based on the historical traffic environment for the at least one intersection (130) ; and

selecting the traffic signal profile for the at least one intersection (130) based on the forecasted traffic density, if the traffic difference is within the threshold.

6. The method according to claim 5, further comprising: determining an optimal traffic signal profile for the at least one intersection (130), if the traffic difference exceeds the threshold, wherein the optimal traffic signal pro- file is based on the traffic difference;

identifying a traffic anomaly at the at least one intersection (130) associable to the optimal traffic signal pro- fi s

repeating the step of determining the optimal traffic signal profile if the traffic anomaly is identified; and

selecting the optimal traffic signal profile as the traffic signal profile if the traffic anomaly is not identified .

7. The method according to claim 1, further comprising: communicating the traffic signal profile of the at least one intersection (130) to neighbouring intersections ;

generating traffic signal profiles for the neighbouring intersections based on the communicated traffic signal profile of the at least one intersection (130);

iteratively generating traffic signal profiles for a plurality of intersections based on the traffic signal profiles for the neighbouring intersections; and

managing the traffic flow for the plurality of intersections based on the traffic signal profiles for the neighbouring intersections.

8. A computing device (200) for managing traffic flow in a geographical location, the computing device (200) comprising: at least one processor (202); and

a memory (204) coupled to the at least one processor (202), the memory (204) comprising:

a traffic density estimator (216) capable of determining a traffic density in real time based on a t ITS £fA.c environment for at least one intersection (130) in the geographical location, the traffic environment comprises vehicular traffic and pedestrian traffic;

a traffic prediction module (218) capable of predicting a forecast traffic density based on the traffic environment and a historical traffic environment associated with the at least one intersection (130);

a profile determination module (220) capable of selecting a traffic signal profile for the at least one intersection (130) based on the traffic density and the forecast traffic density; and

a communication unit (208) capable of communicating the traffic signal profile of the at least one intersection (130) to a traffic sign-post (140) associated with the at least one intersection (130) for managing the traffic flow in the geographical location.

9. The computing device (200) according to claim 8, further comprising :

a traffic analyzer (212) capable of determining lane oc~ cupancy for two or more lanes associated with the at least one intersection (130) based on the traffic environment;

a traffic image generator (210) capable of generating a traffic image of the traffic environment; and

a traffic object determination module (214) capable of determining traffic objects in the traffic environment based on a predetermined identification model, wherein the traffic objects include vehicle type, vehicle occupancy and pedestrian type. 10. The computing device (200) according to claim 9, wherein the traffic analyzer (212) comprises:

a comparator (222) capable of comparing the traffic density with the forecast traffic density, to generate a traffic difference; and

a threshold module (224) capable of determining whether the traffic difference has exceeded a threshold, wherein the threshold is determined based on the traffic environment for the at least one intersection (130) . 11. The computing device (200) according to claim 9 , wherein the traffic analyzer (212) comprises:

an optimal profile module (226) configured to determine an optimal traffic signal profile for the at least one intersection (130), if the traffic difference exceeds the thresh- old, wherein the optimal traffic signal profile is based on the traffic difference; and

an anomaly identifier (228) configured to identifying a traffic anomaly at the at least one intersection (130) asso- ciable to the optimal traffic signal profile determined by the anomaly identifier, wherein the optimal traffic signal profile re-determined if the traffic anomaly is identified and wherein the optimal traffic signal profile is selected as the traffic signal profile if the traffic anomaly is not identi fied .

12. The computing device (200) according to claim 8, wherein the profile determination module (220} capable of selecting the traffic signal profile for the at least one intersection (130) based on the forecasted traffic density if the traffic difference is within the threshold. 13. The computing device (200) according to claim 8, wherein the communication unit (208) comprises:

a transmitter (248) communicatively coupled to the processor (202) to transmit the traffic signal profile of the at least one intersection (130) to neighbouring intersections; and

a receiver (258) communicatively coupled to the processor (202) to receive traffic signal profiles of the neighbouring intersections, wherein the traffic signal profile of the at least one intersection (130) is based on the traffic signal profiles of the neighbouring intersections.

14. A system (100, 600) for managing traffic flow in a geographical location, the system (100, 600) comprising:

a se ver (102);

a network (108) communicatively coupled to the server

(102 ) ; and

one or more computing devices (200) as claimed in claims 8-15, communicatively coupled to the server (102) via the network (108) and wherein the one or more computing devices (200) are associated with one or more intersections (130) in the geographical location.

15. The system (100, 600) according to claim 14, further comprising :

a plurality of capture devices (120, 122, 124) communicatively coupled to the server (102) via the network (108), the plurality of capture devices (120, 122, 124) are one of manned devices and unmanned devices selected comprising image sensors, a motion sensors, a GPS devices and communication devices .

16. The system (100) according to claim 14, wherein the server (102) comprises:

a database (110) comprising data associated with the traffic environment, wherein the traffic environment comprises data associated with weather, air quality, time, day, scheduled events and unscheduled events for the geographical location comprising a plurality of intersections;

a traffic model generator (112) capable of generating a predetermined identification model based on the traffic environment; and

a forecast model generator (114) capable of generating a forecast model based on a historical traffic environment.