Description:FIELD OF THE INVENTION
[001] The present invention relates to a method for generating test scenarios. More particularly, the present invention relates to a method for generating test scenarios for testing a vehicle.

BACKGROUND OF THE INVENTION
[002] Generally, navigation and safety features of vehicles are tested in a controlled environment before introduction in the marketplace. Conventionally, test scenarios are manually generated based on an understanding of potential use cases and edge cases (unusual and rare scenarios), which leads to limited test coverage, human bias, and potential oversight of certain edge cases. Further, manually conducting tests on actual roads and environments to assess vehicle performance is expensive, time-consuming and offers limited control over riding parameters and environmental factors. Accordingly, in some conventional approaches, vehicles are tested in simulated environments that attempt to replicate real-world scenarios. However, there exists an inherent difficulty in creating truly realistic simulations and there is a significant variance of test scenarios with real-world scenarios.
[003] Conventional testing methods generally do not cover a broad range of real-world scenarios involving dynamically varying parameters such as traffic conditions and weather conditions. Further, test scenarios generated by conventional testing methods often rely on predetermined settings, potentially overlooking the variability encountered in real-world driving. Thus, testing under a limited set of real-world scenarios may not fully validate the robustness and adaptability of the vehicles. Additionally, conventional testing methods generally do not account riding behaviour, riding preferences and riding experiences in simulations, leading to potential usability issues in real-world scenarios. Therefore, conventional testing methods often fall short in capturing the complexity and realism of real-world scenarios and user experiences, leading to potential gaps in the evaluation of vehicle performance.
[004] Thus, there is a need in the art for a method for generating test scenarios for testing a vehicle which addresses at least the aforementioned problems.

SUMMARY OF THE INVENTION
[005] In one aspect, the present invention relates to a method for generating test scenarios for testing a vehicle. The method has the steps of procuring user data from one or more first databases, the user data including one or more of riding parameters and demographic data of one or more riders; analysing the procured user data to classify each of the one or more riders into one or more predefined profiles; procuring real-time environmental parameters pertaining to one or more of a geographical location, a traffic condition, and a weather condition; and generating a plurality of test scenarios based on the classification of the one or more riders and the real-time environmental parameters.
[006] In an embodiment of the invention, the procured user data is analysed by a K-Means clustering module to classify the one or more riders into one or more predefined profiles.
[007] In an embodiment of the invention, the plurality of test scenarios is generated by a Recurrent Neural Networks (RNN) module.
[008] In a further embodiment of the invention, the method has the steps of procuring feedback data pertaining to riding experiences from one or more second databases; processing the feedback data pertaining to the riding experiences; extracting the processed feedback data for generating one or more use cases; simulating interactions of the vehicle with the one or more use cases for updating the user data; and generating the plurality of test scenarios based on the updated user data obtained by the simulated interactions of the vehicle with the one or more use cases.
[009] In an embodiment of the invention, processing the feedback data pertaining to the riding experiences is executed by a Natural Language Processing (NLP) module.
[010] In an embodiment of the invention, generating the one or more use cases is executed by a Named Entity Recognition (NER) module in conjunction with a text generation module.
[011] In an embodiment of the invention, interactions of the vehicle with the one or more use cases for updating the user data is simulated by a Reinforcement Learning module.
[012] In an embodiment of the invention, the Reinforcement Learning module includes a Deep Q-Networks (DQN) module.
[013] In a further embodiment of the invention, the method has the steps of identifying edge cases by a predictive analytics module configured to analyse historical data of the one or more riders, and incorporating the identified edge cases into the plurality of test scenarios.
[014] In a further embodiment of the invention, the method has the step of emulating the riding parameters of the one or more riders by a Generative Adversarial Networks (GAN) module for generating the plurality of test scenarios.
[015] In a further embodiment of the invention, the method has the steps of dynamically adjusting one or more testing parameters of the vehicle, optimising the one or more testing parameters corresponding to one or more predefined performance metrics, and generating the plurality of test scenarios based on the optimised one or more testing parameters of the vehicle.
[016] In an embodiment of the invention, the one or more testing parameters of the vehicle includes at least one of a speed of the vehicle, loads acting on the vehicle, and a throttle response of the vehicle.
[017] In an embodiment, the one or more predefined performance metrics includes a fuel efficiency of the vehicle, handling stability of the vehicle, and responsiveness of the vehicle.
[018] In an embodiment of the invention, the steps of dynamically adjusting the one or more testing parameters of the vehicle and optimising the one or more testing parameters corresponding to the one or more predefined performance metrics are executed by a Bayesian Optimisation module.
[019] In a further embodiment of the invention, the method has the steps of analysing sensory data procured by one or more sensors by a Convolutional Neural Networks (CNN) module in real-time, and generating the plurality of test scenarios based on the sensory data.
[020] In an embodiment of the invention, the sensory data procured by the one or more sensors includes at least one of still images and a video feed.
[021] In a further embodiment of the invention, the method includes the steps of continuous learning, by an Online Gradient Descent (OGD) module, based on feedback input by the one or more riders and the real-time environmental parameters; and generating the plurality of test scenarios based on the continuous learning by the OGD module.

BRIEF DESCRIPTION OF THE DRAWINGS
[022] Reference will be made to embodiments of the invention, examples of which may be illustrated in accompanying figures. These figures are intended to be illustrative, not limiting. Although the invention is generally described in context of these embodiments, it should be understood that it is not intended to limit the scope of the invention to these particular embodiments.
Figure 1 is a flow diagram illustrating method steps for generating test scenarios for testing a vehicle, in accordance with an embodiment of the invention.
Figure 2 is a flow diagram illustrating further method steps for generating test scenarios for testing the vehicle, in accordance with a further embodiment of the invention.
Figure 3 is a block diagram of a system for generating test scenarios for testing the vehicle, in accordance with an embodiment of the invention.

DETAILED DESCRIPTION OF THE INVENTION
[023] The present invention generally relates to a method for generating test scenarios. More particularly, the present invention relates to a method for generating test scenarios for testing a vehicle. Throughout the specification, the term “test scenarios” corresponds to a broad description of test conditions or situations encountered by the vehicle during testing. Test scenarios help identify different paths and/or conditions that need to be tested to evaluate safety and performance of the vehicle. It should be understood that the method as illustrated may be applicable for testing the vehicle such as a two-wheeled vehicle, a three-wheeled vehicle, a four-wheeled vehicle, or any multi-wheeled vehicle as required.
[024] Figure 1 illustrates a method 100 for generating test scenarios for testing a vehicle, in accordance with an embodiment of the invention. At step 102 of the method 100, user data is procured from one or more first databases. The user data includes one or more riding parameters and demographic data of one or more riders. In an embodiment, the term “riding parameters” encompasses multifaceted ways in which the one or more riders interact with the vehicle. The riding parameters comprise of measurable parameters, personal choices and preferences, and adaptive responses to external conditions.
[025] The measurable parameters may include but not limited to a speed of the vehicle, position of a brake lever in the vehicle, position of a gear shifter lever in the vehicle and the like. Additionally, the riding parameters correspond to riding behaviour and riding preferences of the one or more riders of the vehicle. For instance, riding behaviour may correspond to a sedate riding behaviour (associated with gradual acceleration and deceleration, and gradual braking of the vehicle) or an aggressive riding behaviour (associated with rapid acceleration and deceleration, and rapid braking) or a combination of sedate riding behaviour and aggressive riding behaviour. Riding preferences corresponds to subjective choices of the one or more riders while riding the vehicle. As an example, partially engaging a brake lever of the vehicle while traversing an undulated terrain may be a riding preference of a novice rider.
[026] At step 104, the procured user data is analysed to classify each of the one or more riders into one or more predefined profiles. The “one or more predefined profiles” refers to one or more categories characterized by specific riding parameters and demographic parameters.
[027] In an embodiment, clustering techniques are adapted to classify the one or more riders into one or more predefined profiles based on riding habits and preferences of the one or more riders. In an embodiment, the procured user data is analysed by a K-Means clustering module to classify the one or more riders into one or more predefined profiles. The K-Means clustering module iteratively groups the user data based on closeness of each data point of the user data to “centroids” representing each of the one or more predefined profiles, thereby classifying the one or more riders into one or more predefined profiles.
[028] In an embodiment, the one or more predefined profiles may include “Urban Commuter”, “Adventure Rider” and “City Deliverer”. In the embodiment, a rider who prioritises quick and efficient navigation in dense traffic conditions in urban areas may be classified as an “Urban Commuter”. A rider who explores diverse terrains and requires route customisation may be classified as an “Adventure Rider”. A rider who prefers optimised routes for timely deliveries in urban areas may be classified as a “City Deliverer”. However, it should be appreciated that the aforementioned profiles are merely exemplary, and that the one or more predefined profiles are not limited thereto. In an embodiment, an individual rider may be classified into multiple predefined profiles.
[029] At step 106, real-time environmental parameters pertaining to one or more of a geographical location, a traffic condition, and a weather condition are procured. Traffic condition corresponds to a status of traffic at a specified timeframe and location. Weather condition corresponds to atmospheric conditions that describe the state of the atmosphere at a specified timeframe, and typically includes parameters such as temperature, humidity, wind speed, cloud cover, precipitation and so on. In an embodiment, the real-time environmental parameters may be procured using a Global Positioning System (GPS) sensor, a traffic level sensor, and a weather sensor. In an embodiment, one or more of the GPS sensor, the traffic level sensor and the weather sensor may be configured as clusters of sensors.
[030] At step 108, a plurality of test scenarios is generated based on the classification of the one or more riders (into the one or more predefined profiles) and the real-time environmental parameters. In an embodiment, machine learning techniques such as Reinforcement Learning can be implemented for the generation of the plurality of test scenarios.
[031] In an embodiment, the plurality of test scenarios is generated by a Recurrent Neural Networks (RNN) module. The RNN module is a specific type of artificial neural network architecture capable of processing sequential data and incorporating contextual information over time. The RNN module is configured to analyse sequential or time-series data associated with the real-time environmental parameters for generating the plurality of test scenarios. As an example, one of the generated plurality of test scenarios can be “Urban Commuter navigating through heavy traffic in downtown during rainy weather”.
[032] Figure 2 illustrates further method steps for generating test scenarios for testing the vehicle, in accordance with a further embodiment of the invention. According to the method, at step 110, feedback data pertaining to riding experiences is procured from one or more second databases. In an embodiment, the one or more second databases correspond to one or more websites or forums hosting reviews and/or discussions on riding experiences of various vehicle users. Examples of feedback data include feedback surveys or reviews on vehicle performance, vehicle safety systems, accident reports, incident data, etc.
[033] At step 112, the feedback data pertaining to the riding experiences is processed by a processing unit. Processing the feedback data involves organising the feedback data procured from the one or more second databases and identifying themes, trends, or patterns in the procured feedback data.
[034] In an embodiment, processing the feedback data pertaining to the riding experiences is executed by a Natural Language Processing (NLP) module. The NLP module is configured to process and extract insights from natural language expressions from the one or more second databases to identify common usage patterns, issues, and scenarios associated with vehicles in general.
[035] At step 114, the processed feedback data is extracted for generating one or more use cases. Here, the term “one or more use cases” corresponds to one or more descriptions of vehicle interactions with external entities (i.e., vehicle users and environmental parameters).
[036] In an embodiment, generation of the one or more use cases is executed by a Named Entity Recognition (NER) module in conjunction with a text generation module (for e.g., a Generative Pre-training Transformer module). The NER module typically uses machine learning techniques to analyse textual information. The NER module is configured to tokenise or break down the textual information into individual words or phrases (termed as “tokens”), analyse each token and its surrounding tokens for clues like part-of-speech and prior mentions of the surrounding tokens, and systematically arrange contextually relevant tokens. The text generation module then generates coherent use cases from the contextually relevant tokens. Thereby, unstructured feedback data can be transformed into contextually relevant use cases by the NER module in conjunction with the text generation module.
[037] As an example, the NLP module may extract an insight from the one or more second databases that riders emphasize the need for a quick route change option during unexpected road closures. Correspondingly, the NER module in conjunction with the text generation module may generate a use case which reads “rider quickly changes navigation route due to unexpected road closures”.
[038] At step 116, interactions of the vehicle with the one or more use cases are simulated for updating the user data. In an embodiment, the simulation of the interactions of the vehicle with the one or more use cases is executed in a virtual environment.
[039] At step 118, the plurality of test scenarios is generated based on the updated user data obtained by the simulated interactions of the vehicle with the one or more use cases. By accounting the updated user data corresponding to the one or more use cases, realism of the generated plurality of test scenarios is enhanced.
[040] In an embodiment, interactions of the vehicle with the one or more use cases for updating the user data is simulated by a Reinforcement Learning module. In a further embodiment, the Reinforcement Learning module comprises a Deep Q-Networks (DQN) module. The Reinforcement Learning module is configured to adapt an iterative process, enabling learning and adaptation of the simulation based on the one or more use cases generated over time. As an example, the Reinforcement Learning module may learn to simulate more cautious riding in adverse weather conditions upon receiving one or more use cases corresponding to riding behaviour in adverse weather conditions.
[041] In a further embodiment of the invention, edge cases (rare and unusual scenarios) are identified by a predictive analytics module configured to analyse historical data of the one or more riders. The historical data of the one or more riders may include rare incidents and unusual scenarios, such as but not limited to road accidents. As an example, an edge case may correspond to a case where a temporary roadblock is encountered due to an event. One or more machine learning models can be adapted to predict scenarios that might not be immediately obvious but are crucial for comprehensive testing. The identified edge cases are incorporated into the plurality of test scenarios, thereby enhancing the coverage of testing.
[042] In an embodiment, the predictive analytics module may implement “Random Forest”, an ensemble learning method capable of handling complex relationships in the historical data, making it suitable for predicting potential edge cases based on the historical data of the one or more riders.
[043] In a further embodiment of the invention, the riding parameters of the one or more riders are emulated by a Generative Adversarial Networks (GAN) module for generating the plurality of test scenarios. The GAN module is adapted to emulate diverse rider behaviours, including variations in riding habits, preferences and riding skill levels. The GAN generated emulations allow for the generation of test scenarios that are specifically relevant to individual riders or rider groups. For example, a test scenario which reads “riders with minimal riding experience struggle with complex navigation routes” may be emulated by the GAN module. By mimicking diverse riding parameters, the GAN module can introduce unforeseen actions or reactions into the plurality of test scenarios. This can help identify potential shortcomings in vehicle performance and safety features that might otherwise be missed. Additionally, if the user data is limited, the GAN module can generate additional synthetic data that shares characteristics with the user data, thereby providing an augmented dataset comprising the user data and the synthetic data. This augmented dataset can be used to train the Reinforcement Learning module, potentially leading to improved learning outcomes in the generation of the plurality of test scenarios. Thereby, the GAN module generates synthetic data to emulate diverse test scenarios for usability testing.
[044] In a further embodiment of the invention, the method 100 for generating test scenarios comprises the steps of dynamically adjusting one or more testing parameters of the vehicle, optimising the one or more testing parameters corresponding to one or more predefined performance metrics, and generating the plurality of test scenarios based on the optimised one or more testing parameters of the vehicle. In an embodiment, the one or more testing parameters of the vehicle comprises at least one of a speed of the vehicle, loads acting on the vehicle, and a throttle response of the vehicle. In an embodiment, the one or more predefined performance metrics is at least one of a fuel efficiency of the vehicle, handling stability of the vehicle, and responsiveness of the vehicle. In an exemplary embodiment, the vehicle speed can be dynamically increased to test responsiveness of the vehicle at higher speeds.
[045] In an embodiment, the steps of dynamically adjusting the one or more testing parameters of the vehicle and optimising the one or more testing parameters corresponding to the one or more predefined performance metrics are executed by a Bayesian Optimisation module. In the embodiment, the Bayesian Optimisation module explores a parameter space comprising the one or more testing parameters, taking into account different values and combinations of the one or more testing parameters. The Bayesian Optimisation module evaluates a plurality of cases associated with the parameter space with the one or more predefined performance metrics (for e.g., fuel efficiency, handling stability, responsiveness etc.). The Bayesian Optimisation module then generates a memory bank termed as a “Bayesian model” which encompasses relationships between the one or more testing parameters and the one or more performance metrics. Using the Bayesian model, the Bayesian Optimisation module iteratively predicts which testing parameter values and/or combinations are likely to yield the best performance in accordance with the predefined one or more performance metrics. Thereby, the Bayesian Optimisation module effectively optimises the one or more testing parameters by intelligently exploring the parameter space based on observed outcomes.
[046] In a further embodiment of the invention, one or more sensors are provided in the vehicle to procure sensory data pertaining to an environment surrounding the vehicle. The sensory data procured by the one or more sensors is analysed by a Convolutional Neural Networks (CNN) module. The plurality of test scenarios is then generated based on the sensory data. In an embodiment, the sensory data procured by the one or more sensors comprises at least one of still images and a video feed. The CNN module is configured to analyse the sensory data to extract relevant information about the environment surrounding the vehicle, such as obstacles on the road, condition of the road (e.g. road with undulations and potholes), lane markings on the road, and traffic dynamics in real-time. Thereby, the CNN module facilitates identification of potential safety hazards and adaptation of safety scenarios in the plurality of test scenarios. As an example, a safety scenario which reads “automatically engage collision avoidance when an obstacle is detected at close proximity” may be adapted in the plurality of test scenarios.
[047] In a further embodiment of the invention, the method 100 involves continuous learning, by an Online Gradient Descent (OGD) module, based on feedback input the one or more riders and the real-time environmental parameters. The plurality of test scenarios is then generated based on the continuous learning by the OGD module. The OGD module implements online learning to continuously update its machine learning model by adapting to data streams corresponding to the feedback input by the one or more riders and the real-time environmental parameters. By updating the simulation based on feedback of the one or more riders, the OGD module facilitates continuous learning of new road features and rider preferences.
[048] In a further embodiment, the OGD module implements online learning to continuously update its machine learning model based on technological developments and changing road conditions. Consequently, the OGD module ensures relevancy and effectiveness of the plurality of test scenarios over time.
[049] In a further embodiment of the invention, reviews from automobile experts and experienced riders are procured for validating the generated plurality of test scenarios. A rule-based logic may be implemented to capture the expertise of the automobile experts and experienced riders in validating the plurality of test scenarios. The user data may further be updated based on the procured reviews from automobile experts and experienced riders for refining the plurality of test scenarios. This ensures that the generated plurality of test scenarios conform with human intuition and capture the intricacies of real-world vehicle usage.
[050] In an embodiment, a plurality of test cases is extracted from the plurality of test scenarios for testing the vehicle. Each test case corresponds to a detailed set of conditions or variables under which performance of the vehicle is tested. Each test case includes preconditions, test inputs, expected outcomes, and steps to execute the test.
[051] Figure 3 illustrates a system 300 for implementing the method 100 for generating test scenarios for testing the vehicle, in accordance with an embodiment of the invention. The system 300 comprises the vehicle 302 mounted with one or more first sensors 304. The one or more first sensors 304 are configured to procure information pertaining to the measurable parameters of the vehicle and the one or more testing parameters of the vehicle 302.
[052] The system 300 comprises a database server 306 which includes the one or more first databases and the one or more second databases. The database server 306 is communicably coupled to the one or more first sensors 304. Particularly, the one or more first databases of the database server 306 are communicably coupled to the one or more first sensors 304. Further, the database server 306 is communicably coupled to one or more second sensors 308. Particularly, the one or more first databases of the database server 306 are communicably coupled to the one or more second sensors 308. The one or more second sensors 308 are configured to procure the real-time environmental parameters.
[053] The one or more first databases of the database server 306 comprises data pertaining to the riding parameters and the real-time environmental parameters. The one or more second databases of the database server 306 comprise the feedback data pertaining to riding experiences of various users.
[054] In an embodiment, the database server 306 further comprises historical data of the one or more riders. The historical data of the one or more riders may include rare incidents and unusual scenarios, such as but not limited to road accidents. In an embodiment, the historical data of the one or more riders may be stored in a third database of the database server 306. In an embodiment, the database server may comprise traditional databases or NoSQL databases depending on the data structure.
[055] The system 300 comprises a central server 310 communicably coupled to the database server 310. The central server 310 comprises an input/output unit 312, a memory 314, a transceiver 316, and a processor 318.
[056] The input/output unit 312 is configured to facilitate communication between the processor 318 and one or more peripheral devices (not shown) connected to the central server 310. The one or more peripheral devices may include a keyboard configured to receive reviews from experts and experienced riders, a display device configured to display the simulation of the vehicle 302, a pointing device configured to manipulate the simulation of the vehicle 302, one or more storage devices (not shown), one or more network interfaces (not shown), and so on.
[057] The memory 314 is configured to store instructions to be executed by the processor 318. The memory 314 may include one or more of a volatile memory and a non-volatile memory.
[058] The transceiver 318 is configured to facilitate communication between the database server 306 and the processor 318. Furthermore, the transceiver 318 may also facilitate bidirectional communication between the processor 318 and a network (not shown).
[059] The processor 318 comprises a machine learning unit 320 and a test scenario generation unit 322. The machine learning unit 320 comprises of various modules implementing machine learning methods, such as the K-Means clustering module, the Recurrent Neural Networks (RNN) module, the Natural Language Processing (NLP) module, the Named Entity Recognition (NER) module, the text generation module, the Reinforcement Learning module, the Deep Q-Networks (DQN) module, the Generative Adversarial Networks (GAN) module, the Convolutional Neural Networks (CNN) module, the Online Gradient Descent (OGD) module, and so on. The test scenario generation unit 322 is communicatively coupled to the machine learning unit 320. The test scenario generation unit 322 is configured to generate the plurality of test scenarios based on instructions received from the machine learning unit 320.
[060] In an embodiment, the processor 318 comprises a dedicated Graphics Processing Unit (GPU) to accelerate the simulation environment. In a further embodiment, hardware accelerators like Tensor Processing Units (TPUs) or specialised GPUs can be used to expedite the execution of machine learning models, for real-time data processing and test scenario generation.
[061] Advantageously, the method of the present invention adopts artificial intelligence driven profile analysis of one or more riders and dynamic scenario generation based on classification of the one or more riders and the real-time environmental parameters, thereby ensuring a broad and realistic set of test scenarios. Thus, the present invention addresses the problem of limited scenario coverage encountered in manual test scenario generation in the conventional art.
[062] Further, automated nature of the method of the present invention reduces manual effort for test scenario generation and testing, leading to more efficient resource utilisation compared to conventional testing methods that may be labour intensive. The automated nature of the method coupled with efficient resource utilisation contributes to cost-effectiveness and time efficiency in the testing process.
[063] By updating the user data with the one or more use cases generated from the feedback data pertaining to riding experiences (by the NLP module and the NER module in conjunction with the text generation module) the present invention takes user experiences into consideration in the generation of the plurality of test scenarios. Consequently, realism of the plurality of test scenarios is enhanced.
[064] Further, simulating the interactions of the vehicle with the one or more use cases by the Reinforcement Learning module (such as the DQN module) allows for adaptive learning and optimisation of the plurality of test scenarios based on riding experiences. This ensures efficient testing in a controlled virtual environment, reducing safety risks associated with real-world testing.
[065] By employing predictive analytics module implementing an ensemble learning method such as “Random Forest”, potential edge cases can be identified and incorporated in the plurality of test scenarios. Consequently, the present invention addresses the problem of overlooking rare and critical scenarios, thereby enhancing the robustness of testing.
[066] Further, the GAN module enables the emulation of diverse rider behaviours, including variations in riding habits, preferences, and riding skill levels. This ensures a more accurate representation of rider interactions with the vehicle compared to conventional usability testing methods.
[067] By employing the Bayesian Optimisation module, the method of the present invention effectively optimises the one or more parameters by intelligently exploring the parameter space based on observed outcomes, thereby ensuring relevancy of the generated plurality of test scenarios with the one or more performance metrics.
[068] Further, the CNN module enables identification of potential safety hazards for adaptation of safety scenarios in the plurality of test scenarios. Thereby, comprehensiveness of testing is enhanced compared to conventional testing methods.
[069] Incorporating continuous learning techniques such as Online Gradient Descent, the OGD module adapts to changing conditions and updates its models based on real-world data, consequently facilitating continuous learning of rider preferences and new road features. This ensures that the method of the present invention remains adaptive and relevant over time, thereby overcoming the challenge of adaptability of conventional testing methods that deal with static data.
[070] By procuring reviews from automobile experts and experienced riders for validating the plurality of test scenarios, credibility and reliability of the testing is enhanced. Such a human-in-the-loop validation step ensures a balance between automated testing and human expertise. Consequently, subjectivity and potential biases associated with automated testing are overcome by the method of the present invention.
[071] Further, the method of the present invention effectively balances realism and control in the simulation. Such an approach allows for effective testing in a controlled environment while incorporating real-world dynamics. By incorporating the aforementioned technical advantages, the present invention provides a comprehensive and effective solution for testing the vehicle, contributing to improved performance, safety, and usability.
[072] In light of the abovementioned advantages and the technical advancements provided by the disclosed method, the claimed steps as discussed above are not routine, conventional, or well understood in the art, as the claimed steps enable the following solutions to the existing problems in conventional technologies. Further, the claimed steps clearly bring an improvement in the functioning of the system itself as the claimed steps provide a technical solution to a technical problem.
[073] Furthermore, one or more computer-readable storage media may be utilized in implementing embodiments consistent with the present disclosure. A computer-readable storage medium refers to any type of physical memory on which information or data readable by a processor may be stored. Thus, a computer-readable storage medium may store instructions for execution by one or more processors, including instructions for causing the processor(s) to perform steps or stages consistent with the embodiments described herein. The term “computer-readable medium” should be understood to include tangible items and exclude carrier waves and transient signals, i.e., be non-transitory. Examples include random access memory (RAM), read-only memory (ROM), volatile memory, non-volatile memory, hard drives, CD ROMs, DVDs, flash drives, disks, and any other known physical storage media.
[074] While the present invention has been described with respect to certain embodiments, it will be apparent to those skilled in the art that various changes and modification may be made without departing from the scope of the invention as defined in the following claims.

List of Reference Numerals
100: Method
300: System
302: Vehicle
304: One or more first sensors
306: Database server
308: One or more second sensors
310: Central server
312: Input/output unit
314: Memory
316: Transceiver
318: Processor
320: Machine Learning unit
322: Test scenario generation unit
, Claims:WE CLAIM:
1. A method for generating test scenarios for testing a vehicle, the method comprising the steps of:
procuring user data from one or more first databases, the user data including one or more of riding parameters and demographic data of one or more riders;
analysing the procured user data to classify each of the one or more riders into one or more predefined profiles;
procuring real-time environmental parameters pertaining to one or more of a geographical location, a traffic condition, and a weather condition; and
generating a plurality of test scenarios based on the classification of the one or more riders and the real-time environmental parameters.

2. The method as claimed in claim 1, wherein the procured user data is analysed by a K-Means clustering module to classify the one or more riders into one or more predefined profiles.

3. The method as claimed in claim 1, wherein the plurality of test scenarios is generated by a Recurrent Neural Networks (RNN) module.

4. The method as claimed in claim 1, comprising the steps of:
procuring feedback data pertaining to riding experiences from one or more second databases;
processing the feedback data pertaining to the riding experiences;
extracting the processed feedback data for generating one or more use cases;
simulating interactions of the vehicle with the one or more use cases for updating the user data; and
generating the plurality of test scenarios based on the updated user data obtained by the simulated interactions of the vehicle with the one or more use cases.

5. The method as claimed in claim 4, wherein processing the feedback data pertaining to the riding experiences is executed by a Natural Language Processing (NLP) module.

6. The method as claimed in claim 4, wherein generating the one or more use cases is executed by a Named Entity Recognition (NER) module in conjunction with a text generation module.

7. The method as claimed in claim 4, wherein interactions of the vehicle with the one or more use cases for updating the user data is simulated by a Reinforcement Learning module.

8. The method as claimed in claim 7, wherein the Reinforcement Learning module comprises a Deep Q-Networks (DQN) module.

9. The method as claimed in claim 1, comprising the steps of:
identifying edge cases by a predictive analytics module configured to analyse historical data of the one or more riders; and
incorporating the identified edge cases into the plurality of test scenarios.

10. The method as claimed in claim 1, comprising the step of emulating the riding parameters of the one or more riders by a Generative Adversarial Networks (GAN) module for generating the plurality of test scenarios.

11. The method as claimed in claim 1, comprising the steps of:
dynamically adjusting one or more testing parameters of the vehicle;
optimising the one or more testing parameters corresponding to one or more predefined performance metrics; and
generating the plurality of test scenarios based on the one or more testing parameters of the vehicle.

12. The method as claimed in claim 11, wherein the one or more testing parameters of the vehicle comprises at least one of a speed of the vehicle, loads acting on the vehicle and a throttle response of the vehicle.

13. The method as claimed in claim 11, wherein the one or more predefined performance metrics comprises a fuel efficiency of the vehicle, handling stability of the vehicle, and responsiveness of the vehicle.

14. The method as claimed in claim 11, wherein the steps of dynamically adjusting the one or more testing parameters of the vehicle and optimising the one or more testing parameters corresponding to the one or more predefined performance metrics is executed by a Bayesian Optimisation module.

15. The method as claimed in claim 1, comprising the steps of:
analysing sensory data procured by one or more sensors by a Convolutional Neural Networks (CNN) module in real-time; and
generating the plurality of test scenarios based on the sensory data.

16. The method as claimed in claim 15, wherein the sensory data procured by the one or more sensors comprises at least one of still images and a video feed.

17. The method as claimed in claim 1, comprising the steps of:
continuous learning, by an Online Gradient Descent (OGD) module, based on feedback input by the one or more riders and the real-time environmental parameters; and
generating the plurality of test scenarios based on the continuous learning by the OGD module.

Dated this 19th day of February 2024
TVS MOTOR COMPANY LIMITED
By their Agent & Attorney

(Nikhil Ranjan)
of Khaitan & Co
Reg No IN/PA-1471