DESC:FORM 2
THE PATENTS ACT 1970
(39 of 1970)
&
The Patents Rules, 2003
COMPLETE SPECIFICATION
(See section 10 and rule 13)

1. TITLE OF THE INVENTION: A SYSTEM AND METHOD FOR COGNITIVE ASSESSMENT OF RISK FOR DRIVERS AND DRIVER SAFETY MEASUREMENT FRAMEWORK
2. APPLICANT:

(a) Name : Nervanik AI Labs Pvt. Ltd.
(b) Nationality : Indian
(c) Address : B 129, New Shreejinagar, Sector - 5, Nirnaynagar, Ahmedabad - 382481, Gujarat, INDIA.

PROVISIONAL

The following specification describes the invention. þ COMPLETE

The following specification particularly describes the invention and the manner in which it is to be performed.

Field of The Invention
The present invention relates to a system and method for cognitive assessment of risk for drivers and driver safety measurement framework. The present invention improves the existing driver scoring framework by a comprehensive 360° analysis of a driver's risk index. The present invention measures various driving conditions and gives an appropriate weightage to each of the factors while computing the score with the best intention to eliminate bias in rating driving patterns.
Background of The Invention
Driving is a motor task that requires significant visual guidance and attention. Any additional complex visual-motor secondary tasks performed while driving, such as dialing a phone, texting/browsing the internet, or reaching for an object, impede the ability of a driver to successfully complete the task of driving. Studies have shown show that secondary tasks involving manual typing, texting, dialing, reaching for an object, or reading are dangerous. The use of a cell phone is associated with a quadrupling of the risk of injury and property damage. Cellular telephone use while driving is a risk factor, but the magnitude of risk is unknown, especially with the use smart-phones that are essentially hand-held internet-accessible computers, allowing access, for example, to social media platforms.
Drivers today are equipped with smart-phones that provide them with navigational support. However, smart-phones are playing a major role in causing distractions during a trip. Every smartphone activity, be it using it for navigational support or texting or calling, both in-hand usage and hands-free usage have risks associated with it. Smart-phone usage results in cognitive distraction and eye gaze deviations from the road that lowers the driver’s attention state while driving. In-hand phone usage inhibits both the attention state as well as vehicle maneuvering abilities through the steering wheel.
In addition to distractions, traditionally recognized driver risks like “hard braking”, “hard acceleration”, “hard cornering”, “rapid lane changing” and “rapid overtaking” impact a driver's overall accidental risk. Continuous Tailgating, lower time to collision (TTC) and lower distance gap from the leading vehicle increases the chance of crashes. Such risks need to be identified and measured for rating a driver’s performance and identifying the required remediation actions for improving overall driver and fleet safety.
As per the study by Marketing and Development Research Associates (MDRA), in India an average driver drives around 12 hours daily and on average covers a distance of 417km daily. In the same study, about half of the drivers responded that they drive even when feeling fatigued or sleepy. The study found that over 2/3rd drivers feel overburdened with assignments and resort to speeding to meet their deadlines. This results in lower driver reaction time in emergency situations and increased risk for road accidents and collisions. Thus driving under fatigue/drowsy state need to be identified and necessary preventive actions should be taken to minimise driver fatigue, both through real-time trip monitoring, coaching/awareness sessions and better work allocation.
Adherence to legal frameworks, rules and regulations laid down by law enforcement authorities is important for an ordered and rule based driving behaviour. Correspondingly, driver’s tendency to strictly follow the laid down rules need to be measured and any shortcoming or carelessness among drivers towards traffic signs and signals be remedied on priority.
The monitoring and behavioral profiling of drivers have an increasing relevance in the application of vehicle telematics. Vehicle telematics is the integration of wireless communications, monitoring systems, and location devices to provide real-time spatial and performance data of a vehicle or a fleet of vehicles. The challenges of using telematics in fleet management are the transformation of data into actionable information. Telematics provides a large amount of data that for example report the individual vehicle's location and performance. However, it does not directly provide useful information about its operational condition or efficiency. Fleet managers lack efficient computational hardware and algorithms to transform large amounts of telematics data into more useful predictive information on the condition of a fleet, an individual vehicle, or vehicle components. Telematics technology providers do not provide a clear methodology for the integration the collected data.
In the telematics industry, “Driver Scores” have long been employed to measure the risk associated with a driver’s driving behaviour. The current approaches majorly identify risky events in categories of hard braking, speeding, hard acceleration and hard cornering. A limited set of approaches include a measure of driver’s fatigue/drowsiness. These events are given an individual event score. Individual event scores are then aggregated using a simple or weighted average across different categories to calculate the overall driver score.
Late night driving can have a significant impact on driver scores and overall safety of a driver. Studies have shown that likelihood of accidents increases significantly during late night driving due to factors such as reduced visibility, driver fatigue, and impaired driving. According to the National Highway Traffic Safety Administration (NHTSA) 50% of all fatal crashes that occur at night involve alcohol-impaired driving. In addition, the risk of being involved in a fatal crash is three times higher during night time hours compared to daytime hours.
Further, weather and lighting conditions can heavily impact the driver’s driving performance. Low visibility conditions require higher vigilance and attentiveness and consequently risk associated with distracted driving and other risky driving events increase manifold. Rains can affect the grip of tires on the roads affecting the braking distance. Thus weather and visibility state is determined with respect to rain, wind speed, lighting conditioning, time of day for scoring drivers.
Advanced driver assistance systems (ADAS) can enhance the driving ability of drivers and contribute to safer traffic conditions. Despite their potential benefits, some drivers are hesitant to adopt these systems. Acceptance of new technologies is a complex issue that depends on various psychological and practical factors. Researchers have explored the acceptance of ADAS by drawing on several theories of human behavior and technology acceptance. A recent driver acceptance model consists of five components (attitude, perceived usefulness, endorsement, affordability, and compatibility) that were able to accurately predict 85% of the variability in drivers’ willingness to accept driver support systems. An important factor that needs to be addressed is how to create a trustworthy relationship between the driver and the system. If a driver mishandles the device or tries to turn it off while driving, this could be considered a violation of safe driving practices and may result in a lower driver score. On the other hand, compliance to alerts results in improvisation of scores.
In current approaches, context of the event/driver condition is often not considered while calculating the score. Even in driver scoring approaches that consider the context in which the events occurred, the parameters considered in defining the context are very restricted, considering just the weather or traffic factors. Each individual risky event is considered in isolation resulting in the compounding effect of co-occurrence of 2 or more hazardous events on driver’s risk being completely ignored. Due to its limited nature as described, driver score has largely been found to be used for gamification or leader board purposes and is easily manipulated. The limitations of the driver scoring approaches currently available makes a case for developing a more comprehensive driver scoring framework.
Hence, the present invention presents a view to improve the exiting driver scoring framework by a comprehensive 360-degree analysis of a driver's risk index.

Object of The Invention
The main object of the present invention is to disclose a system and method for cognitive assessment of risk for drivers and driver safety measurement framework to overcome the problem associated with conventional approaches of calculating a driver score by providing a comprehensive 360° analysis of a driver’s risk index.
Yet another object of the present invention is to provide a driver assessment framework by calculating driving state measurements with extensive number of data points for generating score.
Yet another object of the present invention is to identify the driver’s fatigue/drowsy state while driving and increase the driver’s reaction time in emergency situations to take preventive actions to minimise the risk for road accidents.
Further object of the present invention is to measure the comprehensive number of factors describing the risky events; vehicle state, road state and weather state to calculate the route risk that is combined with the cognitive risk assessment for driver route allocation.
Yet another object of the present invention is to monitor driver’s behaviour and detect signs of fatigue, such as changes in lane position or the frequency of steering corrections and provide alerts to driver for take a break or adjust their driving behaviour using the information.
Further object of the present invention is to provide a feedback mechanism via personalized driver coaching and incorporating behavioral change response of post coaching into scoring framework.
Further object of the present invention is to measure the compounding impact on risk of co-occurring hazardous driving event.
Another object of the present invention is to provide a penalty as well as reward based scoring system for universal use case of driver accreditation, driver coaching and driver rating.
These and other objects will be apparent based on the disclosure herein.
Summary of The Invention
The present invention relates to a system and method for cognitive assessment of risk for drivers and driver safety measurement framework. It provides uniqueness as a driver assessment framework are i) driving state measurements with extensive number of data points are consider for generating score, ii) feedback mechanism via personalized driver coaching, iii) incorporating behavioral change response post coaching into scoring framework, iv) penalty as well as reward based scoring system, v) measures the compounding impact on risk of co-occurring hazardous driving event, and vi) scoring framework for universal use case of driver accreditation, driver coaching and driver rating. In the present invention, the risk score is measured by a comprehensive number of factors describing the driver’s state, risky event state, vehicle state, road state and weather state. The present invention measures driver’s state through psychometric assessment tools to assess psychological and physiological fitness of the driver. Assessment of personality traits affecting driving risk such as conscientiousness, sensation seeking and anger hostility is done for preliminary profiling.
In addition, the present invention assesses motivation level, fatigue and stress as these significantly affect on-road driving performance. The present invention considers a behavioral change, an attitude and responsiveness to coaching while assessing driver’s psychological fitness. In the present invention, the computer vision models are trained to perform alcohol-level tests and detect driving under influence.
In the present invention, the Context Factor (CF) is evaluate by receiving the data from the road segment characteristics/(r), traffic levels/(tr), weather parameters/(w), visibility levels/(vi) and vehicle characteristics/(ve). Further, the event severity analysis device evaluates the events by the deviations from the threshold events/(dt) and the time duration/(de) of the events.
The present invention identifies the compounding events that are co-occurring of multiple risky driving events are through a clustering module and on penalized based the severity of these events in these final score. In the present invention, the CARDs include the analysis of compounding impact on the drivers risk due to co-occurring events. Compound events are recognized based on a multi-level AI System that consists of clustering module and anomaly detection algorithm. The module identifies zones of high concentration of risky driving events and evaluates the risk involved and a high weightage is given to such risk and high penalty deducted from the overall CARDs score.
Brief Description of The Drawings
Other objects, advantages and novel features of the invention will become apparent from the following detailed description of the present embodiment when taken in conjunction with the accompanying drawings.
Fig. 1 illustrates a process flow diagram for identifying high risk driving events using EdgeAI of a method for cognitive assessment of risk for drivers and driver safety measurement framework according to the present invention.
Fig. 2 illustrates a functional block diagram of apparatus for a system for cognitive assessment of risk for drivers and driver safety measurement framework according to the present invention.
Fig. 3 illustrates a process block diagram of a context based on road segment characteristics, traffic, weather and vehicular characteristics for a method for cognitive assessment of risk for drivers and driver safety measurement framework according to the present invention.
Fig. 4 illustrates a functional block diagram of apparatus for a system for cognitive assessment of risk for drivers and driver safety measurement framework according to the present invention.
Fig. 5 illustrates a process block diagram of an event score calculated from the determined events and contexts for a method for cognitive assessment of risk for drivers and driver safety measurement framework according to the present invention.
Fig. 6 illustrates a process block diagram of trip and overall score calculation for a method for cognitive assessment of risk for drivers and driver safety measurement framework according to the present invention.
Fig. 7 illustrates a functional block diagram of apparatus for a system for cognitive assessment of risk for drivers and driver safety measurement framework according to the present invention.
Fig. 8 illustrates a process block diagram of potential impact of a method for cognitive assessment of risk for drivers and driver safety measurement framework according to the present invention.
Fig. 9 illustrates a flow diagram of method for enforcing the penalties on the driver and/or incentivizes the driver based on the driver’s driving history according to the present invention.
Fig. 10 illustrates a heatmap of the example of the system based on driver’s performance calculated for a month according to the present invention.
Detailed Description of The Invention
Before explaining the present invention in detail, it is to be understood that the invention is not limited in its application to the details of the construction and arrangement of parts illustrated in the accompany drawings. The invention is capable of other embodiment, as depicted in different figures as described above and of being practiced or carried out in a variety of ways. It is to be understood that the phraseology and terminology employed herein is for the purpose of description and not of limitation.
It is to be noted that a system and method for cognitive assessment of risk for drivers and driver safety measurement framework describes herein is within the scope of present invention.
As used herein, the term “CARDs”, refers to cognitive assessment of risk for drivers.
As used herein, the term “ADAS”, refers to an advance driver-assistance system.
As used herein, the term “DADS”, refers to driver alertness detection system.
As used herein, the term "module", refers to as the unique and addressable components of the software implemented in hardware which can be solved and modified independently without disturbing (or affecting in very small amount) other modules of the software implemented in hardware.
As used herein, the term “device”, refers to a unit of hardware, outside or inside the case or housing that is capable of providing input or of receiving output or of both.
As used herein, the term "database" refers to either a body of data, a relational database management system (RDBMS), or to both. The database includes any collection of data including hierarchical databases, relational databases, flat file databases, object-relational databases, object oriented databases, and any other structured collection of records or data that is stored in a computer system. The above examples are example only, and thus are not intended to limit in any way the definition and/or meaning of the term database.
As shown in Fig. 1, it illustrates a process flow diagram for identifying high risk driving events using EdgeAI of a method for cognitive assessment of risk for drivers and driver safety measurement framework. It is to be noted that, the present invention uses IMU (Inertial Measurement Unit) based and video based telematics data with advanced machine learning module to identify these risks and trigger alerts to drivers to avoid accidents. The present invention has device means to measures the attentiveness to the real time alerts and detects improper driver actions like hard braking, hard acceleration, hard cornering, traffic signal violation, drowsy driving, and poor response time and distracted driving.
As shown in Fig. 2, the system for cognitive assessment of risk for drivers and driver safety measurement framework comprises, a camera device (1) configured to capture both time-series based data and video data of the driver’s actions and road side movements and generates collision warning alerts, a GNSS (Global Navigation Satellite Systems) and IMU (Inertial Measurement Unit) sensors (11) configured to identify greater variety of risky driving events and context mapping along with the compounding impact of compound event interaction, an AI powered road monitoring unit (12) configured to identify hard-braking, tailgating, swerving maneuvers, hard cornering, hard acceleration, over-speeding, traffic violations, distracted driving, drowsy driving, illegal overtaking and compounding events, an AI powered driver monitoring unit (13) configured to identify distraction events of drivers, collision alerts, face recognition, assess the psychological and physiological fitness of driver, a cloud API (Application Programming Interface) services (14) configured to provide real-time traffic and weather details on a map as the trip's external conditions to a trip weather indices device (15) and a traffic indices device (16), a trip weather indices device (15) configured to determine the weather and visibility state with respect to rain, wind speed, lighting conditioning and time of day for scoring drivers, humidity and visibility to understand the ambient conditions while driving, a trip traffic indices device (16) configured to determine route, traffic flow, traffic density and daily average traffic volumes, a context weight traffic-weather-route vulnerability zone device (17) configured to determine weights for measuring the impact of context factors on driver safety levels by calculating the marginal impact of each context parameter on the driver’s risk based on the data received from the camera device (1), the GNSS and IMU sensors (11), the AI powered road monitoring unit (12), the AI powered driver monitoring unit (13), the cloud API services (14), the trip weather indices device (15) and the trip traffic indices device (16) and an events/event-clusters detection (18) configured to identify the performance in high risk zones to cluster routes into graded risk profiles into non-graded profile weights and penalties for unsafe driving in each of these risk-graded route zones based on the data received from the AI powered road monitoring unit (12) and the AI powered driver monitoring unit (13).
Geospatial data from service providers is used for analysis of road segments vulnerability mapping and road characteristics. The IMU sensor (11) data is filtered and thereafter threshold based analytics and anomaly detection models are used to identify risky events.
The roadside condition is assessed in terms of availability of clear zone for recovery of errant vehicles. The widths of the clear zone, lateral offset against obstructions, presence of median barrier are assessed to characterize roadside condition. The road curvature is assessed in terms of horizontal curvature radius, transition curves between straight path and curves, and vertical curves. It considers the route gradation into high to low accident risk zones historical crash data on the route, the road architecture with respect to presence of intersections, roundabouts, crossways, diverging and merging lanes is done.
Therefore in CARDs framework, road characteristics are extensively measured based on road cross-section, roadside condition and road curvature as road cross-section is characterised by parameters like travel lane width, skid resistance of road surface, width of shoulder, type and material of road shoulder, pavement types and texture, pavement distresses or faults like rutting, polishing, bleeding and also dirty pavements cause poor skid resistances of road surfaces.
The CARDs measures driver’s state through psychometric assessment tools to assess psychological and physiological fitness of the driver. The assessment of personality traits affecting driving risk such as conscientiousness, sensation seeking and anger hostility is done for preliminary profiling. In addition, the CARDs assess motivation level, fatigue and stress as these significantly affect on-road driving performance. The behavioral change and attitude and responsiveness to coaching are also considered while assessing driver’s psychological fitness. The alcohol-level tests are also performed and computer vision models are trained to detect driving under influence. All these factors go into describing the driver’s state for scoring of driver’s in CARDs.
The driving behaviour is measured by identification of risky driving events like hard braking, tailgating, swerving maneuvers, hard cornering, hard acceleration, over speeding, traffic violations, distracted driving, drowsy driving and illegal overtaking. Compounding events, that is, co-occurring of multiple risky driving events, are identified through a clustering module and penalized based on the severity of these events in these final score.
The, CARDs takes into consideration how long the driver has been driving at night. To calculate a driver’s score in this context, factors such as the driver’s speed, the frequency of sudden braking or accelerating, and their overall level of attentiveness is considered. Other factors include i) distance driven during late night hours, ii) compliance with traffic laws and iii) previous incidents or accidents that occurred during late night driving.
CARDs program with both IMU and video-based telematics allows greater variety of risky driving event identification and context mapping, along with compounding impact of event interaction, making it a unique and comprehensive driver score framework.
Now as shown in Fig. 3, it illustrates a process block diagram of context based on road segment characteristics, traffic, weather and vehicular characteristics. The present invention considers the impact of contextual factors like route characteristics, real time traffic levels, weather and visibility information, vehicular load and performance profile. Route characteristics are derived by segmenting the trip into road segments and analysing the road segment inherent risk profile. Key road segments based on junction details like roundabouts or 3/4 way crossroads, presence of bends, vertical or horizontal curvature, terrain ruggedness, presence of high accident risk zones, road category and road surface characteristics are identified and a driver’s entire trip route is mapped out. To further elaborate the impact of trip route mapping on analysing driver’s risk, consider the case of events like hard braking.
The vehicle is assessed for its safety preparedness based on presence of features such as driver alertness detection system (DADS), automatic braking systems, infrared night vision systems, adaptive headlamps, reverse backup sensors, adaptive cruise control, lane departure warning systems, deflation detection systems, tire pressure monitoring systems, traction control systems, electronic stability control, emergency brake assist, assured clear distance ahead. It considers the driver’s familiarity and training for capacity building for safe driving using these vehicle safety features. The vehicle is assessed for the presence of seatbelts, airbags and laminated windshield as well for its overall safety rating.
As shown in Fig. 4, the system additionally comprises a database AI interface device (2), a historical accident pattern on route device (21), drivers CARDs on specific route device (22) and a route’s risk profile device (23). The database AI interface device (2) configured to transfer and/or receive the data-sets from a historical accident pattern on route device (21), drivers CARDs on specific route device (22) and a route’s risk profile device (23). A route risk profiling device (24) to identify the suitable route to the driver by analyzing parameters received from the historical accident pattern on route device (21), drivers CARDs on specific route device (22), the route’s risk profile device (23), the driver and vehicle fright details device (31) and the route roster device (32). With respect to the Fig. 4, a dynamic route risk zoning device (25) configured to analyze the route risks based on the data received from the route risk profiling device (24) and based on the risk calculation. The dynamic route risk zoning device (25) measures road characteristics based on road cross-section, roadside condition and road curvature as characterised by parameters i.e. travel lane width, skid resistance of road surface, width of shoulder, type and material of road shoulder, pavement types and texture, pavement distresses or faults i.e. rutting, polishing, bleeding and dirty pavements cause poor skid resistances of road surfaces. A route driver recommender system (26) configured to recommend the suitable route to the driver through analysing the parameters by a route risk profiling device (24) and enable the driver based on the data received from the route risk profiling device (24) and weighted urgency, hazardous/ non-hazardous nature and value of goods across long haul routes.
The CARDs score is augmented with the route risks weighted driver's driving performance on the previous journeys and adherence to the halt zones and accident zone warnings, pre-determined from route risk analysis. Events generated in predetermined high risk zones are given higher weights. Performance in high risk zones are analyzed by advanced clustering module to cluster routes into graded risk profiles with non-linear graded weights and penalties for unsafe driving in each of these risk-graded route zones are deducted from score. Route risks are analyzed based on the GIS data for routes, spatial pattern of historical vehicular accident data, and previous journeys data (however not limited to, weather condition, road condition, highways, city, speed limits, road-type distributions, previous crash risk zones as per trips data collected and stored in database from on-vehicle camera device).
Based on the above route risk calculation and based on the weighted urgency, hazardous/non-hazardous nature and value of goods the driver’s route allocation recommender engine enable driver allocation across long haul routes. Recommending halt durations required for attentive driving, along with routes having halt locations for long haul trips, garages en-route and avoiding accident prone locations. This enables improved fleet safety, efficiency for the fleet operators on specific routes which are computed by the set of rules. The cumulative result of the above gives top drivers, minimum driver switching, recommended drivers along the route, and specialization of drivers on high risk and difficult terrain routes. The overall benefit is reflected via efficient fuel consumption and increased safe turn-around-time for the trip.
As shown in Fig. 4, the system additionally comprises an OLI (Online Local Installer) application device (3) configured for client side data entry, a driver and vehicle fright details device (31) configured to analysing the vehicle state for score through vehicle brand, model, wheelbase, size, weight, registration year and vehicle maintenance history and a route roster device (32). The OLI application device (3) configured to provide data of the driver and vehicle to the route risk profiling device (24).
The factual vehicular and driver details are obtained from customers through a customized data exchange platform. The vehicular details like model, year of manufacture, maintenance service history, and engine power are obtained for estimation of braking efficiency, braking distance, maneuverability, etc. The real-time traffic and weather details are obtained from cloud API service providers which are used to map the trip's external conditions. The AI powered in-cabin dual camera device (1) captures both time-series based IMU sensor data and video data of driver actions and road side movements. This 360 degree data collection pipeline powers CARDs set of rules.
With respect to the Fig. 2, Fig. 4 and Fig. 5, an AI powered in-vehicle coaching device (5) configured to receive data from the events severity analysis device (4), the events/event-clusters detection device (18), the dynamic route zoning device (25) and the route driver recommender system (26). A real time alerts and coaching device (51) to deliver coaching manually as well as virtual driving skill improvement suggestions to each person through driver coach based on the data received from the AI powered in-vehicle coaching device (5). A driver responsiveness to alerts and coaching device (52) configured to measure coaching interventions, driver’s driving performance and behavioral change after receiving coaching to reflect the score based on the data received from the real time alerts and coaching device (51). The driver responsiveness alert and coaching device (52) provides data to AI powered in-vehicle coaching device (5) for in-vehicle coaching.
The CARDs have an in-built mechanism for feedback based performance review on driver driving behavior. Using an AI powered in-vehicle coaching device (5) recommendation engines, clustering of drivers for coaching needs past performance on CARDs framework and driver’s profile, a personalized coaching module is developed for each driver.
The AI powered in-vehicle coaching device (5) consists of a voice assistant and a GPU (Graphics Processing Unit) processor. The device delivers real-time coaching interventions while enroute based on the real time trip analysis for events and trip score. The pre-trip determined graded risks zone is dynamically updated with real-time signals for traffic density, weather, and collision risks. Based on the updated risk level for immediately approaching segments of road, coaching interventions to navigate high risk accidental zones are delivered to the driver. The driver’s response to these alerts and time taken are incorporated in updating the driver’s and CARDs profile score.
Within the CARDs framework, the coaching is delivered to each person through a driver coach who delivers manual as well as virtual driving skill improvement suggestions to be acted upon. The CARDs framework consider the coaching interventions a driver has received, and the score reflects the driver’s driving performance and behavioral change after receiving coaching. With respect to the Fig. 2, in the Fig. 5, an events severity analysis device (4) receives the data from the context weight traffic-weather-route vulnerability zone device (17) and events and event clusters detection device (18) for the analysis of the risky events.
Now as shown in Fig. 6, the CARDs scoring methodology identifies the hazardous driving events using video based and IMU based telematics. An AI models are used for identification of driver’s posture changes, detect phone usage, detect eye gaze deviation from road, measure face orientation changes. The road side camera of the camera device (1) video feed is analyzed by trained AI models to identify traffic signs and signals, road terrain, road alignment, object in vehicle’s path and calculate distance gap and time to collision (TTC) etc. The driving behaviour is measured by identification of risky driving events like hard braking, tailgating, swerving maneuvers, hard cornering, hard acceleration, over speeding, traffic violations, distracted driving, drowsy driving and illegal overtaking. The camera device (1) provides collision alerts and driver’s response time.
Along with the events, its context is identified. The weights for measuring the impact of context factors on driver safety levels are determined by calculating the marginal impact of each context parameter on driver’s risk. Once the risky events and its context are identified, event severity is determined by calculating the deviations from the safe limit threshold. A penalty is levied for each of the events that increase the driver’s risk of accident.
The Context Factor (CF) is evaluate by receiving the data from the road segment characteristics, traffic levels, weather parameters, visibility levels and vehicle characteristics. Here, the Context Factor (CF) denoted as:
CF = f(r,tr,w,vi,ve)
Wherein, the context factor involving the multivariable functions as f(r) is road segment characteristics, f(tr) is traffic levels, f(w) is weather parameters, f(vi) is visibility levels and f(ve) is vehicle characteristics.
Further, the event severity analysis device (4) evaluates the events by the deviations from the threshold events and the time duration of the events. Here, the event severity is denoted as:
E = f(dt,de)
Wherein, the event severity involving the functions as f(d??) and f(d??) are the deviations from event threshold and time duration of the event.
Using the context factor and event severity value, a penalty for each event is calculated. If at least two events occur within the same time window, the compounding impact of interaction of the events on accident risk is calculated which results in a higher penalty for such event clusters. Rewards are calculated by identifying good driving actions of the driver, such as a smooth navigation in high traffic area, a dangerous turn, immediate braking in case of a sudden pedestrian coming in front of visible from sides, etc.
It is to be noted that the CARD’s measures the accuracy of detection by random sampling based manual annotation of events and calculating the consequent KPI (Key Performance Index) for each event type to determine the accuracy of the set of rules.
In the present invention, the method for measuring the KPI for a given timeframe is generated by following a series of steps. First, annotated events are filtered, including only those that have been marked as valid or invalid by annotators. Second, the total number of events, valid events, and invalid events are tabulated for each event type. Third, the KPI is then calculated using the precision formula, where KPI (precision) is equal to the number of valid events divided by the total number of events. Instances where the total number of events is less than 50 are filtered out as a placeholder value. Finally, the KPI score at the 50th percentile is taken for each event type, representing the KPI for the event type in the given timeframe.
The total number of alerts and events in a trip are normalized based on the same KPI to have accurate measure of actual events and alerts.
Normalized Event Count=Event Count*Event KPI
E.g., if there are 100 distraction alerts in a trip, and KPI for Distraction is 87%, total number of distraction that will be considered for scoring would be;
Normalized Event Count=100*0.87=87
Penalties for alerts and events are calculated based on weights assigned to each event type,
Alert Penalty = w*(e^((normalized alert count per km) -1 )))*10
Event Penalty = w*(e^((normalized event count per 100 km) -1 )))*10
Since the score is based on alerts and events, and each alert and events has its own weight in the score, and each alert/ event is scaled down as per the KPI, the penalty is calculated as weighted average of both alert and event for each event type,
Penalty (e)=((alert penalty (e)+event penalty (e)))/((sum of weights for e))
Where e is each event type in the score.
Compound Events are recognized based on a multi-level AI System that consists of clustering module and anomaly detection set of rules. The module identifies zones of high concentration of risky driving events and evaluates the risk involved and a high weightage is given to such risk and high penalty deducted from the overall CARDs score.
Now as shown in Fig. 5 and Fig. 7, a trip score device (6) configured to receive data from the event severity analysis device (4) and the driver responsiveness alert and coaching device (52) to determine the weighted aggregate based on the data received from the trip score device (6) by a weighted average of trip score device (61). The weighted average of trip score device (61) configured to generate a CARDs score of all risky events, penalties and rewards based on the data received from the trip score device (6) which are then normalized for driving distance.
In embodiments of the present invention, the final card score is generated by performing following method steps;
generating collision warning alerts through capturing time-series based data and video data of driver’s actions and road side movements by a camera device (1) and sending the data to a GNSS and IMU sensors (11), an AI powered road monitoring unit (12) and an AI powered driver monitoring unit (13);
identifying risky driving events and context mapping along with the compounding impact of compound event interaction by the GNSS and IMU sensors (11);
identifying hard-braking, tailgating, swerving maneuvers, hard cornering, hard acceleration, over-speeding, traffic violations, distracted driving, drowsy driving, illegal overtaking and compounding events by the AI powered road monitoring unit (12);
identifying distraction events of drivers, collision alerts, face recognition and assessing the psychological and physiological fitness of the driver by the AI powered driver monitoring unit (13);
providing real-time traffic and weather details on a map as the trip's external conditions to a trip weather indices device (15) and a traffic indices device (16) by a cloud API services (14);
receiving data from the trip weather indices device (15) and the trip traffic indices device (16) by a context weight traffic-weather-route vulnerability zone device (17);
evaluating data from the road segment characteristics, the traffic levels, the weather parameters, the visibility levels and the vehicle characteristics to get the context factor by the context weight traffic-weather-route vulnerability zone device (17) and transferring data to an events severity analysis device (4);
identifying performance of driver in high risk zones to cluster routes into graded risk profiles into non-graded profile weights and penalties for unsafe driving in each of these risk-graded route zones based on the data received from the AI powered road monitoring unit (12) and the AI powered driver monitoring unit (13) by an events/event-clusters detection (18) and transferring data to the events severity analysis device (4) and an AI powered in-vehicle coaching (5);
evaluating events based on the data received from the context weight traffic-weather-route vulnerability zone device (17) and events/event-clusters detection (18) by the event severity analysis device (4);
sending data to a historical accident pattern on route device (21), drivers CARDs on specific route device (22) and a route’s risk profile device (23) by a database AI interface device (2);
sending data to a driver and vehicle fright details device (31) and a route roster device (32) by an OLI application device (3);
identifying suitable route to the driver by analysing the parameters based on the data received from the historical accident pattern on route device (21), drivers CARDs on specific route device (22), the route’s risk profile device (23), the driver and vehicle fright details device (31) and the route roster device (32) by the route risk profiling device (24) and transferring data to a dynamic route risk zoning device (25) and a route driver recommender system (26);
analyzing route risks based on the data received from the route risk profiling device (24) by the dynamic route risk zoning device (25);
enabling the driver based on the data received from the route risk profiling device (24) by a route driver recommender system (26);
receiving data from the events severity analysis device (4), the events/event-clusters detection device (18), the dynamic route zoning device (25) and the route driver recommender system (26) by the AI powered in vehicle device (5);
providing a real-time alerts and coaching to driver based on the data received from the AI powered in-vehicle coaching (5) by a real time alerts and coaching device (51);
measuring driver responsiveness alerts based on the data received from the real time alerts and coaching device (51) by a driver responsiveness to alerts and coaching device (52);
measuring accuracy of detection by random sampling based manual annotation of events and calculating the consequent KPI (Key Performance Index) for each event type to determine the accuracy of the set of rules;
receiving data from the event severity analysis device (4) and the driver responsiveness alert and coaching device (52) by a trip score device (6);
determine weighted aggregate based on the data received from the trip score device (6) by a weighted average of trip score device (61); and
generating a final CARDs score of all risky events, penalties and rewards based on data received from the trip score device (6) by the weighted average of trip score device (61).
The collision prevention and responsive action taken after collision warning alerts generated by the camera device (1) gives rewards in the CARDs and helps in improving scores of drivers. The swift and safe maneuvering on damaged roads, difficult terrain and in accident prone zones is rewarded with an increase in score. This improvement on traditional scoring frameworks that only penalize the risky driving events detected while positive emergency responses requiring hard braking are either not accounted for or clubbed with other risky driving events and penalized in the final score.
As shown Fig. 8, the CARDs score is intended for use as a platform for comprehensive assessment of driver’s driving behavior. It can be used by fleet services for enlisting high performing drivers based on driver’s CARDs score. This will revolutionize the current process of hiring drivers which is highly ad-hoc and unorganized with as much as 90% of driver’s being untrained. The current driver hiring process or the lack of any such process leads to untrained, inexperienced and risky driver being hired for long hauls on roads which compromises the overall road safety. The drivers seeking to establish their driving skill credentials can enlist with the CARDs and generate CARDs score for themselves to get enlisted as prospective drivers/partners for fleets and logistics aggregators. The fleet owners can use CARDs to identify their driver's coaching needs and work allocation based on route-driver fit. The auto insurance companies can use CARDs to determine annual subscription premiums based on CARDs score and reward good drivers with discounts on premium.
Now, as shown in Fig. 9, it illustrates the flow diagram of enforcing the penalties on the driver and/or incentivizing the driver based on the driver’s driving history. The embodiment of the present invention calculates KPI of events for the past days (i.e. 7 days). After this, it counts events for each trip in conjunction with KPIs. Later, the system retrieves the driver coaching information from the database AI interface device (2). After this, the system retrieves instances of late-night driving and behavioural information from the AI powered road monitoring unit (12) and the AI powered driver monitoring unit (13). Further, enforcing penalties for high-risk events, taking into consideration the weather and vehicle profile as well as co-occurring events on the driver. Incentivize driver for complying with alerts and demonstrating an active approach to driving and coaching.
The present invention assigns bonus scores to drivers who effectively comply with alerts and can incentivize safe driving behavior. It encourages drivers to trust the system and adhere to warnings more frequently. Further, the CARD’s assign bonus scores to drivers who i) consistently maintain a safe distance from the vehicle ahead or adhere to the speed limit; ii) complete training courses that focus on the effective and proper use of the system; and iii) regularly review the feedback and take steps to correct any unsafe driving behavior.
Now, as shown in Fig. 10, it illustrates the heatmap based on the driver’s performance calculated for a month. In this example of the present invention, each trip is assigned a score when the trip ends based on alerts, events, and the driver’s behaviour during that trip. Other information, such as location and total driving time and distance, is also considered during the trip.
For each driver, a cumulative score is calculated for each month using the system of present invention. These scores help in identifying riskiest or safest drivers and provide necessary driving coaching to the drivers.
The present invention is the first driver scoring approach that considers a comprehensive number of factors while scoring drivers. It assesses drivers based on their psychological and physiological fitness as well as their driving behaviour with respect to the frequency and severity of risky driving events based on the context of road characteristics, traffic characteristics, and weather characteristics. High risk accident zones are identified based on assessment of road architecture and network, and historical accidental data.
The present invention driver scoring framework is the most comprehensive exercise in driver risk assessment taking into factors to measure driver’s psychological and physiological fitness, personality traits affecting driving risk such as conscientiousness, sensation seeking and anger hostility, motivation level, fatigue and stress. The driving behaviour like by identification of risky driving events like hard braking, tailgating, swerving maneuvers, hard cornering, hard acceleration, over speeding, traffic violations, distracted driving, drowsy driving, illegal overtaking are identified and analyzed for CARDs score. The assessment of road condition, travel lane width, skid resistance of road surface, width of shoulder, type and material of road shoulder, pavement types and texture, pavement distresses or faults, width of the clear zone, lateral offset against obstructions, presence of median barrier id some to understand the impact of road characteristics on drivers risk.
Assessment of vehicle in terms of vehicle brand, model, wheelbase, size and weight, safety preparedness based on presence of features such as driver alertness detection system (DADS), automatic braking systems, infrared night vision systems, adaptive headlamps, reverse backup sensors, adaptive cruise control, lane departure warning systems, deflation detection systems, tire pressure monitoring systems, traction control systems, electronic stability control, emergency brake assist, assured clear distance ahead is done have a comprehensive understanding of vehicular impact on drivers risk.
Additionally, assessment of route, traffic flow, traffic density, daily average traffic volumes, weather conditions with respect to wind, rain, humidity, visibility is done for understanding ambient conditions while driving. Thus CARDs will reflect the widest scope of factors taking into consideration for driver risk assessment.
The present invention has beneficial advantages that the CARDs will act as a catalyst for the development of drivers' wellness programs and helps in addressing the driver’s need for skill coaching, physical and mental well-being and will incentivize rule based on-road driver behavior.
The invention has been explained in relation to specific embodiment. It is inferred that the foregoing description is only illustrative of the present invention and it is not intended that the invention be limited or restrictive thereto. Many other specific embodiments of the present invention will be apparent to one skilled in the art from the foregoing disclosure.
All substitution, alterations and modification of the present invention which come within the scope of the following claims are to which the present invention is readily susceptible without departing from the invention. The scope of the invention should therefore be determined not with reference to the above description but should be determined with reference to appended claims along with full scope of equivalents to which such claims are entitled.

List of Reference Numerals
1 Camera Device
11 GNSS And IMU Sensors
12 AI Powered Road Monitoring Unit
13 AI Powered Driver Monitoring Unit
14 Cloud API Services
15 Trip Weather Indices Device
16 Trip Traffic Indices Device
17 Context Weights Traffic weather Route Vulnerability Zone Device
18 Events/Events-Clusters Detection Device
2 Database AI Interface Device
21 Historical Accident Pattern on Route Device
22 Drivers CARDs on Specific Routes Device
23 Routes’ Risk Profiling Device
24 Route Risk Profiling Device
25 Dynamic Route Risk Zoning Device
26 Route Driver Recommender System
3 OLI Application Device
31 Driver And Vehicle Fright Details Device
32 Route Roster Device
4 Event Severity Analysis Device
5 AI Powered In-Vehicle Coaching Device
51 Real Time Alerts And Coaching Device
52 Driver’s Responsiveness To Alerts And Coaching Device
6 Trip Score Device
61 Weighted Average of Trip Score Device

,CLAIMS:We Claim:
A system for cognitive assessment of risk for drivers and driver safety measurement framework comprises,
a camera device (1) configured to capture both time-series based data and video data of the driver’s actions and road side movements and generates collision warning alerts;
a GNSS (Global Navigation Satellite Systems) and IMU (Inertial Measurement Unit) sensors (11) configured to identify greater variety of risky driving events and context mapping along with the compounding impact of compound event interaction;
an AI powered road monitoring unit (12) configured to identify hard-braking, tailgating, swerving maneuvers, hard cornering, hard acceleration, over-speeding, traffic violations, distracted driving, drowsy driving, illegal overtaking and compounding events;
an AI powered driver monitoring unit (13) configured to identify distraction events of drivers, collision alerts, face recognition, assess the psychological and physiological fitness of driver;
a cloud API (Application Programming Interface) services (14) configured to provide real-time traffic and weather details on a map as the trip's external conditions to a trip weather indices device (15) and a traffic indices device (16);
the trip weather indices device (15) configured to determine weather and visibility state with respect to rain, wind speed, lighting conditioning, time of day for scoring drivers, humidity and visibility to understand the ambient conditions while driving;
the trip traffic indices device (16) configured to determine route, traffic flow, traffic density and daily average traffic volumes;
a context weight traffic-weather-route vulnerability zone device (17) configured to determine weights for measuring the impact of context factors on driver safety levels by calculating the marginal impact of each context parameter on the driver’s risk based on the data received from the camera device (1), the GNSS and IMU sensors (11), the AI powered road monitoring unit (12), the AI powered driver monitoring unit (13), the cloud API services (14), the trip weather indices device (15) and the trip traffic indices device (16);
an events/event-clusters detection (18) configured to identify the performance of driver in high risk zones to cluster routes into graded risk profiles into non-graded profile weights and penalties for unsafe driving in each of these risk-graded route zones based on the data received from the AI powered road monitoring unit (12) and the AI powered driver monitoring unit (13);
an events severity analysis device (4) configured to evaluate the events based on the data received from the context weight traffic-weather-route vulnerability zone device (17) and events/event-clusters detection (18);
a database AI interface device (2) configured to transfer and/or receive data-sets from a historical accident pattern on route device (21), drivers CARDs on specific route device (22) and a route’s risk profile device (23);
an OLI application device (3) configured for client side data entry;
a driver and vehicle fright details device (31) configured to analyze the vehicle state for score through vehicle brand, model, wheelbase, size, weight, registration year and vehicle maintenance history;
a route roster device (32);
a route risk profiling device (24) configured to identify the suitable route to the driver by analyzing parameters received from the historical accident pattern on route device (21), drivers CARDs on specific route device (22), the route’s risk profile device (23), the driver and vehicle fright details device (31) and the route roster device (32);
a dynamic route risk zoning device (25) configured to analyze route risks based on the data received from the route risk profiling device (24) and based on the risk calculation;
a route driver recommender system (26) configured to enable the driver based on the data received from the route risk profiling device (24) and weighted urgency, hazardous/non-hazardous nature and value of goods across long haul routes;
an AI powered in-vehicle coaching device (5) configured to receive data from the events severity analysis device (4), the events/event-clusters detection device (18), the dynamic route zoning device (25) and the route driver recommender system (26);
a real time alerts and coaching device (51) configured to deliver coaching manually as well as virtual driving skill improvement suggestions to each person through driver coach based on the data received from the AI powered in-vehicle coaching device (5);
a driver responsiveness to alerts and coaching device (52) configured to consider coaching interventions, driver’s driving performance and behavioral change after receiving coaching to reflect the score based on the data received from the real time alerts and coaching device (51);
a trip score device (6) configured to receive the data from the event severity analysis device (4), the driver responsiveness alert and coaching device (52); and
a weighted average of trip score device (61) configured to generate a CARDs score of all risky events, penalties and rewards based on the data received from the trip score device (6).
The system as claimed in claim 1, wherein the IMU sensors (11) measures attentiveness to the real time alerts and detect improper driver actions including hard braking, hard acceleration, hard cornering, traffic signal violation, drowsy driving, and poor response time and distracted driving.
The system as claimed in claim 2, wherein the IMU sensor (11) identifies risky events by filtering the data and using threshold based analytics and anomaly detection models.
The system as claimed in claim 1, wherein the AI powered driver monitoring unit (13) consists of a psychometric assessment tools to assess motivation level, fatigue and stress, behavioral change, attitude and responsiveness of driver.
The system as claimed in claim 1, wherein the compound events are recognized based on a multi-level AI system that consists of clustering module and anomaly detection set of rules.
The system as claimed in claim 1, wherein the dynamic route risk zoning device (25) measures road characteristics based on road cross-section, roadside condition and road curvature.
The system as claimed in claim 1, wherein the OLI application device (3) provides data of the driver and vehicle to the route risk profiling device (24) through the driver and vehicle fright details device (31) and the route roster device (32).
The system as claimed in claim 1, wherein the context factor (CF) is evaluated by receiving data from the road segment characteristics, the traffic levels, the weather parameters, the visibility levels and the vehicle characteristics through the context weight traffic-weather-route vulnerability zone device (17) using the following equation:
CF = f(r,tr,w,vi,ve)
The system as claimed in claim 1, wherein the event severity analysis device (4) evaluates events by the deviations from the threshold events and the time duration of the events using the following equation:
E = f(dt,de)
The system as claimed in claim 1, wherein the AI powered in-vehicle coaching device (5) consisting of a voice assistant and a GPU processor configured to deliver real-time coaching interventions while enroute based on the real time trip analysis for events and trip score.
The system as claimed in claim 1, wherein the driver responsiveness alert and coaching device (52) provides data to AI powered in-vehicle coaching device (5) for in-vehicle coaching.
The system as claimed in claim 1, wherein the weighted average of trip score device (61) determines weighted aggregate by receiving the data from the trip score device (6).
A method for cognitive assessment of risk for drivers comprises following method steps:
generating collision warning alerts through capturing time-series based data and video data of driver’s actions and road side movements by a camera device (1) and sending the data to a GNSS and IMU sensors (11), an AI powered road monitoring unit (12) and an AI powered driver monitoring unit (13);
identifying risky driving events and context mapping along with the compounding impact of compound event interaction by the GNSS and IMU sensors (11);
identifying hard-braking, tailgating, swerving maneuvers, hard cornering, hard acceleration, over-speeding, traffic violations, distracted driving, drowsy driving, illegal overtaking and compounding events by the AI powered road monitoring unit (12);
identifying distraction events of drivers, collision alerts, face recognition and assessing the psychological and physiological fitness of the driver by the AI powered driver monitoring unit (13);
providing real-time traffic and weather details on a map as the trip's external conditions to a trip weather indices device (15) and a traffic indices device (16) by a cloud API services (14);
receiving data from the trip weather indices device (15) and the trip traffic indices device (16) by a context weight traffic-weather-route vulnerability zone device (17);
evaluating data from the road segment characteristics, the traffic levels, the weather parameters, the visibility levels and the vehicle characteristics to get the context factor by the context weight traffic-weather-route vulnerability zone device (17) and transferring data to an events severity analysis device (4);
identifying performance of driver in high risk zones to cluster routes into graded risk profiles into non-graded profile weights and penalties for unsafe driving in each of these risk-graded route zones based on the data received from the AI powered road monitoring unit (12) and the AI powered driver monitoring unit (13) by an events/event-clusters detection (18) and transferring data to the events severity analysis device (4) and an AI powered in-vehicle coaching (5);
evaluating events based on the data received from the context weight traffic-weather-route vulnerability zone device (17) and events/event-clusters detection (18) by the event severity analysis device (4);
sending data to a historical accident pattern on route device (21), drivers CARDs on specific route device (22) and a route’s risk profile device (23) by a database AI interface device (2);
sending data to a driver and vehicle fright details device (31) and a route roster device (32) by an OLI application device (3);
identifying suitable route to the driver by analysing the parameters based on the data received from the historical accident pattern on route device (21), drivers CARDs on specific route device (22), the route’s risk profile device (23), the driver and vehicle fright details device (31) and the route roster device (32) by the route risk profiling device (24) and transferring data to a dynamic route risk zoning device (25) and a route driver recommender system (26);
analyzing route risks based on the data received from the route risk profiling device (24) by the dynamic route risk zoning device (25);
enabling the driver based on the data received from the route risk profiling device (24) by a route driver recommender system (26);
receiving data from the events severity analysis device (4), the events/event-clusters detection device (18), the dynamic route zoning device (25) and the route driver recommender system (26) by the AI powered in vehicle device (5);
providing a real-time alerts and coaching to driver based on the data received from the AI powered in-vehicle coaching (5) by a real time alerts and coaching device (51);
measuring driver responsiveness alerts based on the data received from the real time alerts and coaching device (51) by a driver responsiveness to alerts and coaching device (52);
measuring accuracy of detection by random sampling based manual annotation of events and calculating the consequent KPI (Key Performance Index) for each event type to determine the accuracy of the set of rules;
receiving data from the event severity analysis device (4) and the driver responsiveness alert and coaching device (52) by a trip score device (6);
determine weighted aggregate based on the data received from the trip score device (6) by a weighted average of trip score device (61); and
generating a final CARDs score of all risky events, penalties and rewards based on data received from the trip score device (6) by the weighted average of trip score device (61).
The method as claimed in claim 13, wherein determining weather and visibility state with respect to rain, wind speed, lighting conditioning and time of day for scoring drivers, humidity and visibility to understand the ambient conditions while driving by the trip weather indices device (15).
The method as claimed in claim 13, wherein determining route, traffic flow, traffic density and daily average traffic volumes by the trip traffic indices device (16).
The method as claimed in claim 13, wherein determining context weights for measuring the impact of context factors on driver safety levels by calculating the marginal impact of each context parameter on the driver’s risk by the context weight traffic-weather-route vulnerability zone device (17).
The method as claimed in claim 13, wherein measuring the road characteristics based on road cross-section, roadside condition and road curvature and based on the risk calculation by the dynamic route risk zoning device (25).
The method as claimed in claim 15, wherein the KPI for a timeframe is generated by following a series of steps;
filtering annotated events including only those that have been marked as valid or invalid by annotators;
tabulating total number of events, valid events, and invalid events for each event type; and
calculating KPI using the following precision formula.
Normalized Event Count=Event Count*Event KPI
The method as claimed in claim 14, wherein the penalties for alerts and events are calculated based on weights assigned to each event type by using following formula.
Alert Penalty = w*(e^((normalized alert count per km) -1 )))*10
Event Penalty = w*(e^((normalized event count per 100 km) -1 )))*10
The method as claimed in claim 15, wherein the penalty is calculated as weighted average of both alert and event for each event type,
Penalty (e)=((alert penalty (e)+event penalty (e)))/((sum of weights for e) )
where (e) is each event type in the score.
The method as claimed in claim 13, wherein the method for generating penalties and/or rewards comprises following steps:
calculating KPI of the events for past days;
counting events for each trip in conjunction with KPIs;
retrieving driver coaching information from the database AI interface device (2);
retrieving instances of late night driving and behavioural information from the AI powered road monitoring unit (12) and the AI powered driver monitoring unit (13);
enforcing penalties for high–risk events through considering weather and vehicle profile as well as co-occurring events on the driver; and/or
incentivize driver for complying with alerts and demonstrating an active approach for driving and coaching.
Dated this on 11th Day of April, 2023.