Traffic influence evaluation method and system considering context state transition
By constructing a multidimensional state vector and a Markov decision process, and training the model using the Q-learning algorithm, the adaptability and interactivity issues of traditional traffic impact assessment methods in complex environments are solved, resulting in a more accurate traffic impact assessment.
Patent Information
- Application Number
- CN202511297204.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-09-11
- Publication Date
- 2025-12-12
AI Technical Summary
Traditional traffic impact assessment methods are poorly adaptable and lack interactivity when faced with complex and ever-changing traffic environments, personalized assessment needs, and high-frequency task scenarios, resulting in low assessment accuracy.
By acquiring contextual information from the tasks to be evaluated and historical assessments, a multidimensional state vector is constructed. Based on the Markov decision process and using the Q-learning algorithm, a traffic impact assessment model is trained, and the state information is dynamically updated to achieve precise policy matching.
It improves the accuracy of traffic impact assessment, reduces human trial and error, and can automatically identify the assessment status of land parcels and recommend the optimal analysis path.
Smart Images
Figure CN121122016A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of traffic impact assessment technology, specifically a traffic impact assessment method and system that considers context state transitions. Background Technology
[0002] With the continuous acceleration of urbanization, land development and construction activities are becoming increasingly frequent, and traffic impact assessment (TIA) has become an indispensable and important part of urban planning, land development and transportation infrastructure construction.
[0003] Traditional traffic impact assessment methods are mostly based on static parameter settings and fixed analysis procedures, relying on expert experience or pre-set models to perform one-way calculations of traffic indicators. However, while this method has certain advantages in terms of standardization and normalization, it suffers from poor adaptability and weak interactivity when facing complex and ever-changing traffic environments, personalized assessment needs, and high-frequency task scenarios, resulting in low accuracy in traffic impact assessments. Summary of the Invention
[0004] The present invention aims to solve at least one of the technical problems existing in the prior art; to this end, the present invention proposes a traffic impact assessment method and system that considers context state transition, in order to solve the technical problems of poor adaptability, weak interactivity and untimely updates in the prior art when facing complex and ever-changing traffic environments, personalized assessment needs and high-frequency task scenarios.
[0005] To achieve the above objectives, a first aspect of the present invention provides a traffic impact assessment method considering context state transitions, comprising: Obtain contextual information for the traffic assessment task to be evaluated and historical assessment tasks; Construct a multidimensional state vector based on contextual information from historical evaluation tasks; Constructing Markov decision processes based on multidimensional state vectors; A traffic impact assessment model is obtained by training the Markov decision process based on the Q-learning algorithm. The traffic assessment task is input into the traffic impact assessment model to obtain the assessment results.
[0006] Preferably, the construction of the multidimensional state vector based on the contextual information of the historical evaluation task includes: Several state information is extracted from the context information of the historical assessment task, and each state information is marked as a state vector to obtain a multi-dimensional state vector. The context information includes several state information, and the state information includes land use type, location level, analysis period, road capacity, population density, and public transport service level.
[0007] It should be noted that each state information corresponds to a specific traffic assessment scenario, describing all the key features in that traffic assessment scenario.
[0008] Preferably, the construction of the Markov decision process based on multidimensional state vectors includes: Extract multidimensional state vectors and construct a state space based on the multidimensional state vectors; The system acquires user action information on several state information items and constructs an action space based on this action information. The action information includes input information, modification information, analysis object adjustment information, and analysis condition adjustment information. Obtain the probability of transitioning to the next state given the current state and the selected action, and mark the corresponding probability as the state transition probability; Define the reward function: ; Set discount factor The discount factor is used to represent the degree of discount on future rewards and to weigh the immediate analytical effects of current actions against the value of potential analytical results in subsequent stages. The range of values is ; The state space, action space, state transition probability, reward function, and discount factor are constructed into a quintuple. The Markov decision process is obtained; among them, For state space, For the space of action, Let be the state transition probability. For the reward function, This is the discount factor.
[0009] It's important to note that the reward function is used to measure the effectiveness of each analytical task or stage. By providing feedback on the results of each stage, the model can adjust subsequent decision-making strategies based on user acceptance levels.
[0010] Preferably, the training of the Markov decision process based on the Q-learning algorithm includes: Extract the state space and action space to obtain several traffic evaluation cases; The Q-table is initialized based on the state space and action space; the Q-table is a second-order matrix composed of multiple Q-values, and the range of the Q-values is as follows: ; Set the learning rate, discount factor, and exploration strategy for the Markov decision process; Several traffic evaluation cases are integrated into a training set, which is then divided into several sets of training data and validation data. The Markov decision process is trained using the training data. The trained Markov decision process is tested using the validation data, and the artificial intelligence model is adjusted based on the validation results. Training stops when the Markov decision process converges.
[0011] Preferably, the exploration strategy is an ε-greedy exploration strategy.
[0012] Preferably, to avoid the sample order affecting the results, the training set is randomly shuffled before each round of training.
[0013] Preferably, the Q-learning algorithm determines the optimal strategy by continuously updating the Q-value through a core update formula, which is: ;in, Status information Select Action Information The current valuation, For learning rate, Information for the execution of the operation The instant reward obtained afterward As a discount factor, In the new state Choose the largest A worthwhile action.
[0014] A second aspect of the present invention provides a traffic impact assessment system that considers context state transitions, comprising: a data processing module, and a data acquisition module and an impact assessment module connected thereto; The data acquisition module is used to obtain contextual information about the traffic assessment task to be evaluated and historical assessment tasks. The data processing module is used to: construct a multi-dimensional state vector based on the contextual information of historical assessment tasks; construct a Markov decision process based on the multi-dimensional state vector; and train the Markov decision process based on the Q-learning algorithm to obtain a traffic impact assessment model. The impact assessment module is used to input traffic assessment tasks into the traffic impact assessment model and obtain assessment results.
[0015] A third aspect of the present invention provides a storage medium comprising a stored executable program, wherein, when the executable program is executed, it controls the device where the storage medium is located to perform the aforementioned traffic impact assessment method considering context state transitions.
[0016] A fourth aspect of the invention provides a processor for running an executable program, wherein the executable program, when running, performs the traffic impact assessment method considering context state transitions described above.
[0017] Compared with the prior art, the beneficial effects of the present invention are: This invention constructs a multi-dimensional state vector using contextual information from historical assessment tasks, builds a Markov decision process based on the multi-dimensional state vector, and trains the Markov decision process using the Q-learning algorithm to obtain a traffic impact assessment model. This enables the traffic impact assessment model to automatically identify the assessment status of land parcels and recommend the optimal analysis path, reducing human trial and error. Furthermore, the traffic impact assessment model can dynamically update state information based on the task context, tailor the action space, and achieve precise strategy matching, thereby improving the accuracy of traffic impact assessment. Attached Figure Description
[0018] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0019] Figure 1 This is a schematic diagram illustrating the principle of the traffic impact assessment reinforcement learning model of the present invention; Figure 2 This is an overall flowchart of the traffic impact assessment method considering context state transition of the present invention; Figure 3 This is a schematic diagram of the traffic impact assessment system that considers context state transitions according to the present invention. Detailed Implementation
[0020] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0021] Please see Figures 1-3 The first aspect of the present invention provides a traffic impact assessment method considering context state transitions, comprising: S1: Obtain contextual information for the traffic assessment task to be evaluated and historical assessment tasks; S2: Construct a multi-dimensional state vector based on the contextual information of historical evaluation tasks; S3: Constructing a Markov decision process based on multidimensional state vectors; S4: The Markov decision process is trained based on the Q-learning algorithm to obtain the traffic impact assessment model; S5: Input the traffic assessment task into the traffic impact assessment model to obtain the assessment results.
[0022] In this embodiment, a multi-dimensional state vector is constructed based on the contextual information of historical evaluation tasks, including: Several state information is extracted from the context information of the historical assessment task, and each state information is marked as a state vector to obtain a multi-dimensional state vector. The context information includes several state information, and the state information includes land use type, location level, analysis period, road capacity, population density, and public transport service level.
[0023] It should be noted that each state information corresponds to a specific traffic assessment scenario, describing all the key features in that traffic assessment scenario.
[0024] In this embodiment, a Markov decision process is constructed based on a multidimensional state vector, including: Extract multidimensional state vectors and construct a state space based on the multidimensional state vectors; The system acquires user action information on several state information items and constructs an action space based on this action information. The action information includes input information, modification information, analysis object adjustment information, and analysis condition adjustment information. Obtain the probability of transitioning to the next state given the current state and the selected action, and mark the corresponding probability as the state transition probability; Define the reward function: ; Set discount factor The discount factor is used to represent the degree of discount on future rewards and to weigh the immediate analytical effects of current actions against the value of potential analytical results in subsequent stages. The range of values is ; The state space, action space, state transition probability, reward function, and discount factor are constructed into a quintuple. The Markov decision process is obtained; among them, For state space, For the space of action, Let be the state transition probability. For the reward function, This is the discount factor.
[0025] It's important to note that the reward function is used to measure the effectiveness of each analytical task or stage. By providing feedback on the results of each stage, the model can adjust subsequent decision-making strategies based on user acceptance levels.
[0026] In this embodiment, the Markov decision process is trained based on the Q-learning algorithm, including: Extract the state space and action space to obtain several traffic evaluation cases; The Q-table is initialized based on the state space and action space; the Q-table is a second-order matrix composed of multiple Q-values, and the range of the Q-values is as follows: ; Set the learning rate, discount factor, and exploration strategy for the Markov decision process; Several traffic evaluation cases were integrated into a training set, with 80% of the training set used as training data and 20% as validation data. The Markov decision process was trained using the training data, and the trained Markov decision process was validated using the validation data. The artificial intelligence model was adjusted based on the validation results. Training was stopped when the Markov decision process converged.
[0027] In this embodiment, the exploration strategy is an ε-greedy exploration strategy.
[0028] In this embodiment, to avoid the sample order affecting the results, the training set is randomly shuffled before each round of training.
[0029] In this embodiment, the Q-learning algorithm determines the optimal strategy by continuously updating the Q-value through a core update formula, which is as follows: ;in, Status information Select Action Information The current valuation, For learning rate, Information for the execution of the operation The instant reward obtained afterward As a discount factor, In the new state Choose the largest A worthwhile action.
[0030] A second aspect of the present invention provides a traffic impact assessment system that considers context state transitions, comprising: a data processing module, and a data acquisition module and an impact assessment module connected thereto; Data acquisition module: used to obtain contextual information of the traffic assessment task to be evaluated and historical assessment tasks; Data processing module: used to construct multidimensional state vectors based on contextual information from historical assessment tasks; construct Markov decision processes based on multidimensional state vectors; train the Markov decision processes using the Q-learning algorithm to obtain the traffic impact assessment model; Impact Assessment Module: This module is used to input traffic assessment tasks into the traffic impact assessment model and obtain the assessment results.
[0031] A third aspect of the present invention provides a storage medium including a stored executable program, wherein, when the executable program is running, it controls the device where the storage medium is located to execute the above-described traffic impact assessment method considering context state transitions.
[0032] A fourth aspect of the present invention provides a processor for running an executable program, wherein the executable program executes the above-described traffic impact assessment method considering context state transitions.
[0033] Some of the data in the above formula are calculated by removing dimensions and taking their numerical values. The formula is the closest to the real situation obtained by software simulation of a large amount of collected data. The preset parameters and preset thresholds in the formula are set by those skilled in the art according to the actual situation or obtained through simulation of a large amount of data.
[0034] The above embodiments are only used to illustrate the technical methods of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical methods of the present invention without departing from the spirit and scope of the technical methods of the present invention.
Claims
1. A traffic impact assessment method considering context state transitions, characterized in that, include: Obtain contextual information for the traffic assessment task to be evaluated and historical assessment tasks; Construct a multidimensional state vector based on contextual information from historical evaluation tasks; Constructing Markov decision processes based on multidimensional state vectors; A traffic impact assessment model is obtained by training the Markov decision process based on the Q-learning algorithm. The traffic assessment task is input into the traffic impact assessment model to obtain the assessment results.
2. The traffic impact assessment method considering context state transition according to claim 1, characterized in that, The construction of a multidimensional state vector based on contextual information from historical evaluation tasks includes: Several state information is extracted from the context information of the historical assessment task, and each state information is marked as a state vector to obtain a multi-dimensional state vector. The context information includes several state information, and the state information includes land use type, location level, analysis period, road capacity, population density, and public transport service level.
3. The traffic impact assessment method considering context state transition according to claim 1, characterized in that, The Markov decision process based on multidimensional state vectors includes: Extract multidimensional state vectors and construct a state space based on the multidimensional state vectors; The system acquires user action information on several state information items and constructs an action space based on this action information. The action information includes input information, modification information, analysis object adjustment information, and analysis condition adjustment information. Obtain the probability of transitioning to the next state given the current state and the selected action, and mark the corresponding probability as the state transition probability; Define the reward function: ; Set discount factor The discount factor is used to represent the degree of discount on future rewards and to weigh the immediate analytical effects of current actions against the value of potential analytical results in subsequent stages. The range of values is ; The state space, action space, state transition probability, reward function, and discount factor are constructed into a quintuple. The Markov decision process is obtained; among them, For state space, For the space of action, Let be the state transition probability. For the reward function, This is the discount factor.
4. The traffic impact assessment method considering context state transition according to claim 3, characterized in that, The training of the Markov decision process based on the Q-learning algorithm includes: Extract the state space and action space to obtain several traffic evaluation cases; The Q-table is initialized based on the state space and action space; the Q-table is a second-order matrix composed of multiple Q-values, and the range of the Q-values is as follows: ; Set the learning rate, discount factor, and exploration strategy for the Markov decision process; Several traffic evaluation cases are integrated into a training set, which is then divided into several sets of training data and validation data. The Markov decision process is trained using the training data. The trained Markov decision process is tested using the validation data, and the artificial intelligence model is adjusted based on the validation results. Training stops when the Markov decision process converges.
5. A traffic impact assessment method considering context state transition according to claim 4, characterized in that, The exploration strategy is an ε-greedy exploration strategy.
6. The traffic impact assessment method considering context state transition according to claim 1, characterized in that, To avoid the sample order affecting the results, the training set is randomly shuffled before each round of training.
7. A traffic impact assessment method considering context state transition according to claim 1, characterized in that, The Q-learning algorithm determines the optimal strategy by continuously updating the Q-value through a core update formula, which is: ;in, Status information Select Action Information The current valuation, For learning rate, Information for the execution of the operation The instant reward obtained afterward As a discount factor, In the new state Choose the largest A worthwhile action.
8. A traffic impact assessment system considering context state transition, used to implement the traffic impact assessment method considering context state transition as described in any one of claims 1-7, characterized in that, include: The data processing module, and the data acquisition module and impact assessment module connected to it; The data acquisition module is used to obtain contextual information about the traffic assessment task to be evaluated and historical assessment tasks. The data processing module is used to construct a multidimensional state vector based on the context information of historical evaluation tasks. Constructing Markov decision processes based on multidimensional state vectors; A traffic impact assessment model is obtained by training the Markov decision process based on the Q-learning algorithm. The impact assessment module is used to input traffic assessment tasks into the traffic impact assessment model and obtain assessment results.
9. A storage medium, characterized in that, The storage medium stores an executable program, which, when executed by a processor, is used to implement a traffic impact assessment method considering context state transitions as described in any one of claims 1-7.
10. A processor, characterized in that, The processor is used to run an executable program, wherein the executable program executes a traffic impact assessment method considering context state transitions as described in any one of claims 1-7.