SYSTEM FOR IDENTIFYING, CLASSIFYING AND MITIGATING THE ROOT CAUSES OF ROAD NETWORK OVERLOAD
Patent Information
- Authority / Receiving Office
- DE · DE
- Patent Type
- Patents
- Current Assignee / Owner
- GM GLOBAL TECHNOLOGY OPERATIONS LLC
- Filing Date
- 2022-10-10
- Publication Date
- 2026-07-09
AI Technical Summary
Traffic authorities lack comprehensive knowledge about the causes of road congestion, including recurring and non-recurring types, leading to inefficient and costly manual surveys using static sensors that provide outdated and incomplete data.
A system utilizing high-speed vehicle telemetry data from multiple automobiles to identify, classify, and mitigate road network congestion by determining the source and cause of recurring and non-recurring congestion through causal inference models and origin-destination matrix analysis.
Provides real-time, accurate, and cost-effective identification and mitigation of road congestion by leveraging vehicle telemetry data, enabling timely traffic management and infrastructure adjustments.
Smart Images
Figure 00000000_0000_ABST
Abstract
Description
CROSS-REFERENCE TO RELATED REGISTRATION
[0001] This application is a partial continuation of the previous US patent application No. 17 / 575,017, filed on January 13, 2022, the contents of which are incorporated herein in full by reference. INTRODUCTION
[0002] The present disclosure relates to road network congestion and in particular to a system and a process that use real-time high-speed vehicle telemetry data (HSVT data) collected from a “city-scale” of motor vehicles to identify the root causes of road congestion in order to mitigate it.
[0003] Traffic authorities often lack comprehensive knowledge of the causes of road congestion and its associated impact on a traffic system. Furthermore, they may lack the expertise to classify the different types of road congestion present in various locations, such as recurring and non-recurring congestion. Local authorities currently typically engage civil engineering consultants to conduct annual reviews and analyses to assess and identify fundamental deficiencies in road infrastructure and capacity limitations. These consultants may utilize partially deployed static sensors installed along the roadways to perform these annual reviews.However, a disadvantage of these static sensors is that the coverage of the data collection (both spatially and temporally) is limited, they can be time-consuming and / or expensive to use, and they often lead to incorrect recommendations due to a lack of current data.
[0004] While existing systems for mitigating road congestion achieve their intended purpose, there is a need for a new and improved system that addresses these problems. SUMMARY
[0005] According to several aspects of the present disclosure, a system for identifying, classifying, and mitigating root causes of road network congestion is provided. The system comprises multiple motor vehicles positioned at several assigned locations within a road network. Each motor vehicle has one or more sensors for generating input. Each motor vehicle also has a telematics control unit (TCU) for generating at least one location signal for a location of the assigned motor vehicle and at least one event signal for an event related to the assigned motor vehicle, wherein the at least one location signal and the at least one event signal correspond to high-speed vehicle telemetry data (HSVT data) based on the input from the at least one sensor.The system further includes a display device and a computer that communicates with the display device and the TCU of the associated vehicles. The computer contains one or more processors coupled to the TCU of the associated vehicles and receives the HSVT data from the TCU of the associated vehicles. The computer further includes a non-transient computer-readable memory (CRM) containing instructions such that the processor is programmed to identify the road network congestion in a current time slot based on the HSVT data. The processor is further programmed to determine, based on the HSVT data over a period of time, whether the road network congestion is recurring or non-recurring. The processor is further programmed to determine a source of recurring congestion and a cause of non-recurring congestion.The processor is further programmed to generate a notification signal associated with the recurring overload, the source and location of the recurring overload, the non-recurring overload and / or the cause and location of the non-recurring overload, so that the display device displays an associated information on the recurring overload, the source and location of the recurring overload, the non-recurring overload and the cause and location of the non-recurring overload in response to the display device receiving the notification signal from the processor.
[0006] According to one aspect, the sensor is at least one GPS unit, one thermocouple, one humidity sensor, one brake sensor, one airbag sensor, one ADAS module, one set of perception sensors and one motion sensor that communicates with the TCU.
[0007] According to another aspect, the processor is programmed to use a causal inference model to determine a causal relationship between a causal factor at a first location in a first time slot and the non-recurring overload at a second location in a second time slot.
[0008] According to another aspect, the processor is further programmed to follow the causal inference model according to: Y(pi,t)=∑k=1PmkY(pi,t−k)+∑k=0PnkX(pj,t−k)+∑k=0POkε(pi,t−k)+φ to use, where Y(p i , t) the non-recurring overload at location p i represented in time slot t; where ∑k=1PmkY(pi,t−k) a historical overload at the location p i represented in [t - K, t - 1]; where m k a weighted vector function for the element Y(p i, t - k) in timeslot t - k; where Y represents a vector of non-recurring congestion in historical data; where k represents a k-th historical timeslot prior to a current timeslot t; where ∑k=0PnkX(pj,t−k) a historical non-recurring overload at the location p j represented in [t - K, t]; where n k is assigned to at least one causal factor; where X represents the causal factor that contributes to the occurrence of a non-recurring overload event Y; where ∑k=0POkε(pi,t−k) an autocorrelation function between a vector Y of a non-recurring overload event and a vector X of a causative factor, which is normally represented as a normal distribution (Gaussian distribution) N(O, σ 2 ) is to be assumed; where O k a weighted function for an autocorrelation function element ε(p) i, t - k) represents; where k represents the k-th historical time slot prior to the current time slot t; and where φ represents random Gaussian noise.
[0009] According to another aspect, the processor is further programmed to determine that the causal factor at the first location in the first time slot causes the non-recurring overload at the second location in the second time slot according to a statistical equation: δ(X→Y)=ƒ(|n|,|o|)g(|m|) where δ(X → Y) represents a causality determination function to determine that the non-recurring overload event Y is caused by a contributing event X; where f(|n|, |o|) represents a function operation between a weighted vector n and a weighted vector o; where n represents a weighted vector of the vector of a contributing event; where o represents a weighted vector of the autocorrelation function between the vector of a contributing event and the non-recurring overload event; where g(|m|) represents a function operation of the weighted vector m; and where m represents a weighted vector of the non-recurring overload event.
[0010] According to another aspect, the processor is further programmed to determine the source of recurring congestion by: using an origin-destination matrix (OD matrix) derived from historical HSVT data to determine an initial route capacity estimate for a road link; using a direct measurement from a collection of HSVT data, collected, for example, in real time or over a predetermined period, such as a full day, to determine a second route capacity estimate for the road link; comparing the first and second route capacity estimates; and calibrating the processor based on a comparison between the first and second route capacity estimates.
[0011] According to another aspect, the processor is programmed to generate and use the OD matrix by: determining a spectral spatiotemporal data extrapolation (spectral TS data extrapolation) of the historical HSVT data; determining the OD matrix based on the spectral TS data extrapolation; determining a route assignment based on the OD matrix; and determining the initial route capacity estimate based on the route assignment and the OD matrix.
[0012] According to another aspect, the processor is further programmed to generate a partial origin-destination matrix (partial OD matrix) and partial trajectory data based on the HSVT data. The processor is further programmed to determine the OD matrix by: determining an origin-destination matrix estimate (OD matrix estimate) based on the partial trajectory data and the HSVT data; determining a dynamic route assignment and a route capacity estimate based on the OD matrix estimate; and determining a static route assignment and a route capacity estimate based on the partial trajectory data and the HSVT data.
[0013] According to another aspect, the processor is also programmed to determine the spectral TS data extrapolation according to a spectral decomposition: vi=arg max‖v‖=1‖Xv‖ vi=arg max‖v‖=1‖(X−∑k=1i−1XviviT)v‖
[0014] According to another aspect, the processor is further programmed to determine the spectral TS data extrapolation according to an object function formulation: X∼(r)=arg maxrank(X˜)≤r‖(X−X∼(r))‖F X∼(r)=∑i=1rσiuiviT
[0015] According to another aspect, the processor is further programmed to use an iterative approach to reduce a mean squared error (MSE) between the spectral TS data extrapolation and a basic truth.
[0016] According to several aspects of the present disclosure, a computer of a system for identifying, classifying, and mitigating root causes of road network congestion is provided. The system includes multiple motor vehicles positioned at various locations within a road network. Each motor vehicle contains one or more sensors for generating input. Each motor vehicle also contains a telematics control unit (TCU) that generates one or more location signals for a location of the associated motor vehicle and one or more event signals for an event related to the associated motor vehicle, wherein the location signal and the event signal correspond to high-speed vehicle telemetry (HSVT) data based on the input from the sensor. The computer includes one or more processors coupled to the TCUs of the associated motor vehicles, the processor receiving the HSVT data from the TCUs.The computer also contains a non-transient computer-readable memory (CRM) containing instructions such that the processor is programmed to identify road network congestion in a current time slot based on HSVT data. The processor is further programmed to determine, based on HSVT data over a period of time, whether the road network congestion is recurring or non-recurring. The processor is further programmed to determine a source of recurring congestion and a cause of non-recurring congestion.The processor is further programmed to generate a notification signal associated with the recurring overload, the source of the recurring overload, the non-recurring overload and / or the cause of the non-recurring overload, so that the display device displays an associated notification signal in response to the display receiving the notification signal from the processor.
[0017] According to one aspect, the processor is further programmed to use a causal inference model to determine a causal relationship between the causal factor at the first location in the first time slot and the non-recurring overload at the second location in the second time slot.
[0018] According to another aspect, the processor is further programmed to follow the causal inference model according to: Y(pi,t)=∑k=1PmkY(pi,t−k)+∑k=0PnkX(pj,t−k)+∑k=0POkε(pi,t−k)+φ to use, where Y(p i , t) the non-recurring overload at location p i represented in time slot t; where ∑k=1PmkY(pi,t−k) a historical overload at the location p i represented in [t - K, t - 1]; where m k a weighted vector function for the element Y(p i , t - k) in timeslot t - k; where Y represents a vector of non-recurring congestion in historical data; where k represents a k-th historical timeslot prior to a current timeslot t; where ∑k=0PnkX(pj,t−k) a historical non-recurring overload at the location p j represented in [t - K, t]; where n kis assigned to at least one causal factor; where X represents the causal factor that contributes to the occurrence of a non-recurring overload event Y; where ∑k=0POkε(pi,t−k) an autocorrelation function between a vector Y of a non-recurring overload event and a vector X of a causative factor, which is normally represented as a normal distribution (Gaussian distribution) N(O, σ 2 ) is to be assumed; where O k a weighted function for an autocorrelation function element ε(p) i , t - k) represents; where k represents the k-th historical time slot prior to the current time slot t; and where φ represents random Gaussian noise.
[0019] According to another aspect, the processor is further programmed to determine that the causal factor at the first location in the first time slot causes the non-recurring overload at the second location in the second time slot according to a statistical equation: δ(X→Y)=ƒ(|n|,|o|)g(|m|) where δ(X → Y) represents a causality determination function to determine that the non-recurring overload event Y is caused by a contributing event X; where f(|n|, |o|) represents a function operation between a weighted vector n and a weighted vector o; where n represents a weighted vector of the vector of a contributing event; where o represents a weighted vector of the autocorrelation function between the vector of a contributing event and the non-recurring overload event; where g(|m|) represents a function operation of the weighted vector m; and where m represents a weighted vector of the non-recurring overload event.
[0020] According to another aspect, the processor is further programmed to determine the source of recurring congestion by: using an origin-destination matrix (OD matrix) derived from historical HSVT data to determine an initial route capacity estimate for a road link; using a direct measurement from a collection of HSVT data to determine a second route capacity estimate for the road link; comparing the first and second route capacity estimates; and calibrating the processor based on a comparison between the first and second route capacity estimates.
[0021] According to several aspects of the present disclosure, a process for operating a system for identifying, classifying, and mitigating root causes of road network congestion. The system includes multiple motor vehicles positioned at various locations within a road network. Each motor vehicle has one or more sensors for generating input. Each motor vehicle also includes a telematics control unit (TCU) that generates one or more location signals for a location of the associated motor vehicle and one or more event signals for an event related to the associated motor vehicle, wherein the location signal and the event signal correspond to high-speed vehicle telemetry (HSVT) data based on the input from the sensor. The system further includes a computer with one or more processors coupled to the TCU of the associated motor vehicles.The computer also includes a non-transient computer-readable memory (CRM) containing instructions. The process involves identifying, using the processor, the road network congestion in a current time slot based on HSVT data. The process further involves determining, based on HSVT data during a period of time, using the processor, whether the road network congestion is recurring and / or non-recurring. The process further involves determining, using the processor, a source of the recurring congestion or a cause of the non-recurring congestion.The process includes generating, using the processor, a notification signal associated with the recurring overload, the source of the recurring overload, the non-recurring overload, or the cause of the non-recurring overload, so that the display device displays an associated signal in response to the display device receiving the notification signal from the processor.
[0022] According to one aspect, the process further includes the use of a causal inference model with the processor to determine a causal relationship between a causal factor at a first location in the first time slot and the non-recurring overload at a second location in a second time slot.
[0023] According to another aspect, the process also includes the use of the causal inference model with the processor according to: Y(pi, t)=∑k=1PmkY(pi,t−k)+∑k=0PnkX(pj,t−k)+∑k=0POkε(pi,t−k)+φ where Y(p i , t) the non-recurring overload at location p i represented in time slot t; where ∑k=1PmkY(pi,t−k) a historical overload at the location p i represented in [t - K, t - 1]; where m k a weighted vector function for the element Y(p i , t - k) in timeslot t - k; where Y represents a vector of non-recurring congestion in historical data; where k represents a k-th historical timeslot prior to a current timeslot t; where ∑k=0PnkX(pj,t−k) a historical non-recurring overload at the location p j represented in [t - K, t]; where n kis assigned to at least one causal factor; where X represents the causal factor that contributes to the occurrence of a non-recurring overload event Y; where ∑k=0POkε(pi,t−k) an autocorrelation function between a vector Y of a non-recurring overload event and a vector X of a causative factor, which is normally represented as a normal distribution (Gaussian distribution) N(O, σ 2 ) is to be assumed; where O k a weighted function for an autocorrelation function element ε(p) i , t - k) represents; where k represents the k-th historical time slot prior to the current time slot t; and where φ represents random Gaussian noise. δ(X→Y)=f(|n|,|o|)g|m| where δ(X → Y) represents a causality determination function to determine that the non-recurring overload event Y is caused by a contributing event X; where f(|n|, |o|) represents a function operation between a weighted vector n and a weighted vector o; where n represents a weighted vector of the vector of a contributing event; where o represents a weighted vector of the autocorrelation function between the vector of a contributing event and the non-recurring overload event; where g(|m|) represents a function operation of the weighted vector m; and where m represents a weighted vector of the non-recurring overload event.
[0024] According to another aspect, the process further includes the processor determining the source of the recurring congestion. The process further includes the processor using an origin-destination matrix (OD matrix) derived from historical HSVT data to determine an initial route capacity estimate for a road connection. The process further includes the processor using a direct measurement from a collection of HSVT data to determine a second route capacity estimate for the road connection. The process further includes the processor comparing the first and second route capacity estimates. The process further includes the processor being calibrated based on this comparison between the first and second route capacity estimates.
[0025] Further areas of application will become apparent from the description provided here. It should be understood that the description and specific examples serve only for illustration and are not intended to limit the scope of protection afforded by this disclosure. List of characters
[0026] The drawings described here serve only for illustration and are not intended to limit the scope of protection of the present disclosure in any way; they show: Fig. 1 a schematic view of a road network with multiple road edges and multiple road intersections, wherein an example of a system includes multiple motor vehicles traveling along the associated road edges and passing through the associated road intersections; Fig. 2 a schematic view of the system according to Fig. 1, illustrating the system comprising motor vehicles with associated telematics control units (TCUs) and a computer communicating with the TCUs for identifying, classifying, and mitigating root causes of road network congestion; and Fig. 3. A flowchart of a non-restrictive example of a process for operating the system according to Fig. 1. DETAILED DESCRIPTION
[0027] The following description is merely illustrative and is not intended to limit the present disclosure, application, or uses. Although the drawings are examples, they are not necessarily to scale, and certain features may be exaggerated to better illustrate and explain a particular aspect of an illustrative example. Any or all of these aspects may be used alone or in combination. Furthermore, the illustrative examples described herein are not intended to be complete or otherwise limiting to the exact shape and configuration shown in the drawings and disclosed in the detailed description below. Illustrative examples are described in detail with respect to the drawings as follows: In Fig. Figure 1 shows a non-restrictive example of a system 100 that uses real-time high-speed vehicle telemetry (HSVT) data collected from multiple motor vehicles 102 in a “city-scale” setting to actively identify a location of congestion, track the propagation of the congestion, and predict its evolution, enabling traffic authorities to manage traffic and / or allowing motor vehicles to take alternative routes in the face of congestion. To this end, the system 100 uses causality analysis and random forest analysis to identify the root causes of road congestion and to determine whether two spatiotemporal random events have a causal relationship with other events that contributed to the emergence of the congestion, as detailed below.In particular, System 100 identifies non-recurring causes of congestion that are attributed to aperiodic or random factors. Non-restrictive examples of these random factors may include traffic accidents, roadworks, severe weather, special events (concerts, sporting events, etc.), and shock waves. However, it is considered that non-recurring causes may be attributed to any other suitable random factor. System 100 further identifies recurring causes of congestion that are attributed to fundamental disruptions in the road infrastructure at periodic intervals or predictable times, such as rush hour or evening hours. Non-restrictive examples of these fundamental disruptions in the road infrastructure may include major road merging, insufficient road capacity, inadequate traffic management, or faulty traffic control systems.However, it is considered that recurring causes may be attributed to some other suitable underlying disruption of the road infrastructure. System 100 uses a data-driven approach to determine causal relationships in order to predict and mitigate congestion in a manner that is more accurate, timely, and cost-effective than the known static sensors currently used by civil engineers and / or traffic authorities.
[0028] According to this non-restrictive example, the system 100 contains the motor vehicles 102, which are positioned at several locations in a road network 103 to travel along the associated road edges 104 and through the associated road intersections 106. As detailed below, the system 100 can determine that a congested intersection and a congested edge, connected to each other, form a first subgraph 108 with a road congestion, and the system 100 can further determine that two congested edges, connected to each other, form a second subgraph 110 with a road congestion.
[0029] In Fig. 2 Each motor vehicle 102 contains a telematics control unit 112 (TCU) for using telematics services to generate one or more location signals for a location of the assigned motor vehicle 102 and one or more event signals for an event related to the assigned motor vehicle 102. The location signal and the event signal correspond to the high-speed vehicle telemetry data (HSVT data) based on input from one or more sensors 113, which are described below. According to this non-limiting example, the TCU 112 is a microcontroller (a complete computer on a single electronic chip), a microprocessor, or a field-programmable gate array (FPGA). The TCU 112 wirelessly connects the assigned motor vehicle 102 to cloud services or other vehicles via V2X or P2P standards over a mobile network.The TCU 112 connects and communicates with various subsystems via data and control buses (CAN) in the vehicle 102 and collects telemetry data. This data includes elements such as position, speed, engine data, and connection quality. It can also provide connectivity within the vehicle via Wi-Fi and Bluetooth, enabling an e-call function in relevant markets. The TCU 112 communicates with suitable sensors 113 of the vehicle 102 to collect telemetry data. Non-limiting examples of these components include a GPS unit 114, which tracks the latitude and longitude of the vehicle 102, allowing the TCU 112 to generate location signals based on these values.Other non-restrictive examples of these sensors 113 may include a humidity sensor 116, a brake sensor 118, an airbag sensor 120, an ADAS module 122, a perception sensor set 123, a motion sensor 124, and one or more mobile communication units 126 and an external interface for mobile communication (GSM, GPRS, Wi-Fi, WiMax, LTE, or 5G) that provides the tracked values to a centralized computer 128 or a database server, as described below. The motor vehicles 102 also include a quantity of memory 130 for storing GPS values in the case of mobile network-free zones or for intelligently storing information about the vehicle's sensor data.
[0030] As described in detail below, the System 100 can include a local road model using peer-to-peer (P2P) or edge computing, which utilizes the TCU 112 of motor vehicles communicating with the TCU 112 of other motor vehicles 102. Edge computing is a distributed computing paradigm that brings computation and data storage closer to the data sources to improve response times and save bandwidth. The System 100 can also include cloud computing with a global road model that uses a remote computer 128 or server. The computer 128 contains one or more processors 134 and a non-transient computer-readable storage medium 132 (“CRM”) containing instructions so that the processor 134 is programmed to receive the HSVT data from the sensors 113 of the associated motor vehicles 102.
[0031] Processor 134 is further programmed to identify road network congestion in a current time slot based on HSVT data. More specifically, Processor 134 is further programmed to determine, based on an analysis of HSVT data over time (e.g., days, weeks, months, etc.), whether the road network congestion is at least one recurring and one non-recurring congestion, in order to distinguish (by location, time of day, day of the week, etc.) whether the observed congestion is part of a recurring or repeating pattern. Processor 134 is programmed to use the causal inference model to establish a causal relationship between a causal factor at a first location in a first time slot and the non-recurring congestion at a second location in a second time slot, according to: Y(pi, t)=∑k=1PmkY(pi,t−k)+∑k=0PnkX(pj,t−k)+∑k=0POkε(pi,t−k)+φ to determine, whereby non-restrictive examples of random factors may include, but are not limited to, weather and related road surface conditions, traffic accidents, construction sites, shock waves, police events, special sporting / public / concert events, or other temporary road / lane closures; wherein Y(p i , t) the non-recurring overload at location p i represented in time slot t; where ∑k=1PmkY(pi,t−k) a historical overload at the location p i represented in [t - K, t - 1]; where m k a weighted vector function for the element Y(p i , t - k) in timeslot t - k; where Y represents a vector of non-recurring congestion in historical data; where k represents a k-th historical timeslot prior to a current timeslot t; where ∑k=0PnkX(pj,t−k) a historical non-recurring overload at the location p j represented in [t - K, t]; where n k is assigned to at least one causal factor; where X represents the causal factor that contributes to the occurrence of a non-recurring overload event Y; where ∑k=0POkε(pi,t−k) an autocorrelation function between a vector Y of a non-recurring overload event and a vector X of a causative factor, which is normally represented as a normal distribution (Gaussian distribution) N(O, σ 2 ) is to be assumed; where O k a weighted function for an autocorrelation function element ε(p) i, t - k); where k represents the k-th historical time slot prior to the current time slot t; and where φ represents random Gaussian noise. An average professional in the field recognizes that Eq. 1 is a non-restrictive example to show how the non-recurring overload event could be modeled as a weighted summation of several influencing factors, including: the historical non-recurring overload event, the event of a contributing factor that could cause the non-recurring overload event, the autocorrelation between the non-recurring overload event and the event of a contributing factor that could cause the non-recurring overload event, and random Gaussian noise. It is considered that other non-restrictive examples or equivalents of Eq.1 or other modeling approaches could be used to model the non-recurring overload event.
[0032] Processor 134 is further programmed to determine a source and location of recurring overload and a cause and location of non-recurring overload. Processor 134 is further programmed to determine, according to a statistical equation, that the causal factor at the first location in the first time slot causes the non-recurring overload at the second location in the second time slot. δ(X→Y)=f(|n|,|o|)g|m| where δ(X → Y) represents a causality determination function to determine that the non-recurring overload event Y is caused by a contributing event X (such as a weather condition, a construction site, a police incident, etc.); where f(|n|, |o|) represents a function operation between a weighted vector n and a weighted vector o; where n represents a weighted vector of the vector of a contributing event (as described in Eq. 1); where o represents a weighted vector of the autocorrelation function between the vector of a contributing event and the non-recurring overload event (as described in Eq. 1); where g(|m|) represents a function operation of the weighted vector m; and where m represents a weighted vector of the non-recurring overload event.
[0033] An average expert in the field recognizes that Eq. 2 is a non-restrictive example to show whether the non-recurring overload event is caused by a given contributing factor event, which could be determined by weighted factor operations assigned to several influencing factors, including: a historical non-recurring overload event, a contributing factor event that could cause the non-recurring overload event, an autocorrelation between the non-recurring overload event and the contributing factor event that could cause the non-recurring overload event, and random Gaussian noise. It is considered that other equivalents or variants of Eq.2 or other modeling approaches could be used to determine that the non-recurring overload event Y is caused by the contributing event X.
[0034] The processor 134 is further programmed to determine the source of recurring congestion by: using an origin-destination matrix (OD matrix) derived from historical HSVT data to determine an initial route capacity estimate for a road link; using a direct measurement from a collection of HSVT data (e.g., historical traffic volumes and speeds) to determine a second route capacity estimate for the road link; comparing the first and second route capacity estimates; and calibrating the processor based on a comparison between the first and second route capacity estimates.
[0035] The processor 134 is further programmed to generate and use the OD matrix by: determining a spectral spatiotemporal data extrapolation (spectral TS data extrapolation) of the historical HSVT data; determining the OD matrix, e.g., a citywide OD matrix or an area-wide OD matrix, based on the spectral TS data extrapolation; determining a route assignment based on the OD matrix; and determining the initial route capacity estimate based on the route assignment and the OD matrix.
[0036] Processor 134 is further programmed to generate a partial origin-destination matrix (partial OD matrix) and partial trajectory data based on the HSVT data, wherein the processor is further programmed to determine the OD matrix by: determining an origin-destination matrix estimate (OD matrix estimate) based on the partial trajectory data and the HSVT data; determining a dynamic route assignment and a route capacity estimate based on the OD matrix estimate; and determining a static route assignment and a route capacity estimate based on the partial trajectory data and the HSVT data. Processor 134 is further programmed to determine the spectral TS data extrapolation according to a spectral decomposition, such as principal component analysis (PCA) or singular vector decomposition (SVD). vi=arg max‖v‖=1‖Xv‖ vi=arg max‖v‖=1‖(X−∑k=1i−1XviviT)v‖
[0037] The processor 134 is further programmed to determine the spectral TS data extrapolation according to an object function formulation: X∼(r)=arg maxrank(X)≤r‖X−X∼(r)‖F X∼(r)=∑i=1rσiuiνiT
[0038] An average professional in the field recognizes that the equations above are non-restrictive examples to demonstrate a spectral decomposition procedure that can support performing data extrapolation using partial OD matrix data. However, it is considered that equivalents or variants of these equations, or other modeling approaches using spectral decomposition and reconstruction, could be employed.
[0039] The processor 134 is further programmed to use an iterative approach to reduce a mean squared error (MSE) between the spectral TS data extrapolation and a basic truth.
[0040] The processor 134 is further programmed to generate a notification signal associated with at least one location of recurrent overload, the source of the recurrent overload, the location of non-recurrent overload, and the cause of the non-recurrent overload, such that the display device indicates the location of the recurrent overload, the source of the recurrent overload, the location of the non-recurrent overload, and the cause of the non-recurrent overload based on the HSVT data in response to the display device receiving the notification signal from the processor. As described in U.S. Patent Application No.As disclosed in 17 / 575,017 (whose entire contents are included by reference), congestion is a system state identified by comparing the observed speeds and travel times on a road link with a referenced non-congestion state derived using historical HSVT data, and formally occurs when the observed traffic volume or derived throughput reaches the capacity of the road link.
[0041] According to one non-restrictive example, the display device could be a screen in a motor vehicle to inform the vehicle occupant about road congestion, enabling the occupant to drive the vehicle along an alternative route that does not experience congestion. According to another non-restrictive example, the display device could be a screen in an autonomous vehicle to inform the vehicle occupant about road congestion and to indicate that the autonomous vehicle is driving along the alternative route without congestion.According to a further, non-restrictive example, the display device may be a desktop computer monitor used by a local authority to analyze road congestion and modify traffic control infrastructure to better manage urban traffic, to dispatch first responders to an accident scene, to dispatch snowplows to clear snow-covered areas of the road network, or to take any other appropriate action.
[0042] In Fig. 4 is a non-restrictive example of a process 200 for operating the system 100 according to Fig. 1 provided. Process 200 begins in block 202 by identifying, using processor 134, the road network congestion in a current timeslot based on the HSVT data.
[0043] Block 204 further includes the determination, based on the HSVT data using processor 134, that the road network congestion is at least one of a recurring congestion or a non-recurring congestion condition.
[0044] In block 206, process 200 further includes determining, using processor 134, at least one source of recurrent overload and one cause of non-recurrent overload. More specifically, process 200 may include using a causal inference model with processor 134 to determine a causal relationship between a causal factor at a first location in the first time slot and the non-recurrent overload at a second location in a second time slot. Process 200 may include using the causal inference model with processor 134 according to equation 1 above: Process 200 further includes determining with processor 134 that the causal factor at the first location in the first time slot causes the non-recurring overload at the second location in the second time slot according to the above equation 2.
[0045] Process 200 further includes determining, with processor 134, the source of the recurring congestion by using, with the processor, an origin-destination matrix (OD matrix) to determine an initial route capacity estimate for a road link; using, with the processor, a direct measurement to determine a second route capacity estimate for the road link; comparing, with the processor, the first and second route capacity estimates; and calibrating, with the processor, the processor based on a comparison between the first and second route capacity estimates.
[0046] In block 208, process 200 further includes the generation, using processor 134, of a notification signal associated with the recurring overload, the source of the recurring overload, the non-recurring overload, and / or the cause of the non-recurring overload. The display device 136 shows an associated information about the recurring overload, the source of the recurring overload, the non-recurring overload, and the cause of the non-recurring overload based on the HSVT data in response to the notification signal received from the processor.
[0047] Computers and computing devices generally contain computer-executable instructions, which can be executed by one or more computing devices, such as those listed above. These computer-executable instructions can be compiled or interpreted from computer programs created using various programming languages and / or techniques, including, but not limited to, Java, C, C++, MATLAB, SIMUEDGE, STATEFLOW, VISUAL BA-SIC, JAVA SCRIPT, PERL, HTML, TENSORFLOW, PYTHON, PYTORCH, KE-RAS, and others, either alone or in combination. Some of these applications can be compiled and executed on a virtual machine, such as a Java Virtual Machine, a Dalvik Virtual Machine, or similar. Generally, a processor (e.g., a microprocessor) receives instructions, for example, from memory, a computer-readable medium, and so on., and executes these instructions, thereby running one or more processes, including one or more of the processes described here. Such instructions and other data can be stored and transmitted using various computer-readable media. A file in a computing device is generally a collection of data stored on a computer-readable medium, such as a storage medium, read / write memory, etc.
[0048] The CRM (also referred to as processor-readable medium) participates in providing data (e.g., instructions) that can be read by a computer (e.g., by a computer's processor). Such a medium can take many forms, including non-volatile and volatile media, but is not limited to them. Non-volatile media can include, for example, optical or magnetic disks and other permanent storage. Volatile media can include, for example, dynamic read / write memory (DRAM), which typically forms main memory. Such instructions can be transmitted through one or more transmission media, including coaxial cable, copper wire, and fiber optics, including the wires that comprise a system bus coupled to an ECU processor. Common forms of computer-readable media include, for example,a floppy disk, a flexible disk, a hard disk, a magnetic tape, any other magnetic medium, a CD-ROM, a DVD, any other optical medium, punched cards, a paper tape, any other physical medium with hole patterns, a RAM, a PROM, an EPROM, a FLASH EEPROM, any other memory chip or any other memory cartridge or any other medium that a computer can read from.
[0049] According to some examples, the system elements can be implemented as computer-readable instructions (e.g., software) on one or more computing devices, which are stored on their associated computer-readable media (e.g., disks, memory, etc.). A computer program product can include such instructions, stored on computer-readable media, for performing the functions described here.
[0050] With regard to the media, processes, systems, procedures, heuristics, etc., described herein, it should be understood that, although the steps of such processes, etc., have been described as occurring in a specific, ordered sequence, such processes can be practiced with the described steps performed in a different order than that described herein. It should further be understood that certain steps can be performed simultaneously, that other steps can be added, or that certain steps described herein can be omitted. In other words, the descriptions of the processes herein are intended to illustrate certain embodiments and should in no way be interpreted to limit the claims.
[0051] Accordingly, it should be understood that the above description is intended to be illustrative and not limiting. Many embodiments and applications, with the exception of the examples provided, would be obvious to those skilled in the art upon reading the above description. The scope of protection of the invention should not be determined with respect to the above description, but rather with respect to the appended claims, together with the full scope of protection of the equivalents to which such claims entitle. It is expected and intended that future developments will take place in the techniques discussed herein and that the disclosed systems and methods will be incorporated into such future embodiments. In summary, it should be understood that the invention is modifiable and variable and is limited only by the following claims.
[0052] All terms used in the claims shall be given their simple and ordinary meanings as understood by those skilled in the art, unless explicitly stated otherwise herein. In particular, the use of singular articles, such as "a," "the," "said," etc., should be read as representing one or more of the elements indicated, unless a claim explicitly limits this to the contrary. QUOTES INCLUDED IN THE DESCRIPTION
[0000] This list of documents cited by the applicant was automatically generated and is included solely for the reader's convenience. The list is not part of the German patent or utility model application. The DPMA accepts no liability for any errors or omissions. Cited patent literature
[0000] US 17 / 575017 [0001, 0040]
Claims
[1] System for identifying, classifying and mitigating root causes of road network congestion, comprising: Several motor vehicles positioned at several assigned locations in a road network, each of which has at least one sensor for generating an input, and each of which further has a telematics control unit (TCU) for generating at least one location signal for a location of the assigned motor vehicle and at least one event signal for an event related to the assigned motor vehicle, wherein the at least one location signal and the at least one event signal correspond to the high-speed vehicle telemetry data (HSVT data) based on the input from the at least one sensor; a display device; and a computer that communicates with the display device and the TCU of the assigned motor vehicles, the computer comprising: at least one processor that is coupled to the TCU of the associated motor vehicles and receives the HSVT data from the TCU of the associated motor vehicles; and a non-transient, computer-readable storage medium containing instructions such that at least one processor is programmed: to identify road network congestion in a current time slot based on HSVT data; to determine, based on the HSVT data during a period of time, that the road network congestion is at least one of a recurring congestion and one of a non-recurring congestion; to identify at least one source of recurring overload and one cause of non-recurring overload; and to generate a notification signal that is associated with at least one of the recurring overload, the source and location of the recurring overload, the non-recurring overload and the cause and location of the non-recurring overload, such that the display device displays an associated information on the recurring overload, the source and location of the recurring overload, the non-recurring overload and the cause and location of the non-recurring overload based on the HSVT data in response to the display device receiving the notification signal from the at least one processor. [2] System according to claim 1, wherein the at least one sensor comprises at least one GPS unit, thermocouple, humidity sensor, brake sensor, airbag sensor, ADAS module, perception sensor set and motion sensor communicating with the TCU. [3] System according to claim 2, wherein the at least one processor is programmed to use a causal inference model to determine a causal relationship between a causal factor at a first location in a first time slot and the non-recurring overload at a second location in a second time slot. [4] System according to claim 3, wherein the at least one processor is programmed to execute the causal inference model according to: Y(pi, t)=∑k=1PmkY(pi, t−k)+∑k=0PnkX(pj, t−k)+∑k=0POkε(pi, t−k)+φ to use, where Y(p i , t) the non-recurring overload at location p i represented in time slot t; where ∑k=1PmkY(pi, t−k) a historical overload at the location p i represented in [t - K, t - 1]; where m k a weighted vector function for the element Y(p i, t - k) in timeslot t - k; where Y represents a vector of non-recurring congestion in historical data; where k represents a k-th historical timeslot prior to a current timeslot t; where ∑k=0PnkX(pj, t−k) a historical non-recurring overload at the location p j represented in [t - K, t]; where n k is assigned to at least one causal factor; where X represents the causal factor that contributes to the occurrence of a non-recurring overload event Y; where ∑k=0POkε(pi, t−k) an autocorrelation function between a vector Y of a non-recurring overload event and a vector X of a causative factor, which is normally represented as a normal distribution (Gaussian distribution) N(O, σ 2 ) is to be assumed; where O k a weighted function for an autocorrelation function element ε(p) i, t - k) represents; where k represents the k-th historical time slot prior to the current time slot t; and where φ represents random Gaussian noise. [5] System according to claim 3, wherein the at least one processor is further programmed to determine, according to a statistical equation, that the causal factor at the first location in the first time slot causes the non-recurring overload at the second location in the second time slot: δ(X→Y)=ƒ(|n|,|o|)g(|m|) where δ(X → Y) represents a causality determination function to determine that the non-recurring overload event Y is caused by a contributing event X; where f(|n|, |o|) represents a function operation between a weighted vector n and a weighted vector o; where n represents a weighted vector of the vector of a contributing event; where o represents a weighted vector of the autocorrelation function between the vector of a contributing event and the non-recurring overload event; where g(|m|) represents a function operation of the weighted vector m; and where m represents a weighted vector of the non-recurring overload event. [6] System according to claim 3, wherein the at least one processor is further programmed to determine the source of the recurring overload by: Using an origin-destination matrix (OD matrix) derived from historical HSVT data to determine an initial route capacity estimate for a road link; Using a direct measurement from a collection of HSVT data to determine a second route capacity estimate for the road link; Comparing the first and second route capacity estimates; and Calibrating at least one processor based on a comparison between the first and second route capacity estimates. [7] System according to claim 6, wherein the at least one processor is programmed to generate and use the OD matrix by: Determining a spectral spatiotemporal data extrapolation (spectral TS data extrapolation) of historical HSVT data; Determining the OD matrix based on spectral TS data extrapolation and available partial OD matrix data; Determining a route assignment based on the OD matrix; and Determining the initial route capacity estimate based on the route allocation and the OD matrix. [8] System according to claim 7, wherein the at least one processor is programmed to generate a partial origin-destination matrix (partial OD matrix) and partial trajectory data based on the HSVT data, and the at least one processor is further programmed to determine the OD matrix by: Determining an origin-destination matrix estimate (OD matrix estimate) based on the partial trajectory data and the HSVT data; Determining a dynamic route assignment and a route capacity estimate based on the OD matrix estimation; and Determining a static route assignment and a route capacity estimate based on the partial trajectory data and the HSVT data. [9] System according to claim 8, wherein the at least one processor is further programmed to determine the spectral TS data extrapolation according to a spectral decomposition: νi=arg max‖ν‖=1‖Xν‖ νi=arg max‖ν‖=1‖(X−∑k=1i−1XνiνiT)ν‖ [10] System according to claim 9, wherein the at least one processor is further programmed to determine the spectral TS data extrapolation according to an object function formulation: X∼(r)=arg maxrank(X)≤r‖(X−X∼(r))‖F X∼(r)=∑i=1rσiuiνiT