Autonomous driving car-following control method based on brain-like continuous learning

Abstract

The invention discloses an autonomous driving car-following control method based on brain-like continuous learning. The method comprises the steps of firstly obtaining a driving data set under an incremental task; inputting the driving data set under the incremental task into the vehicle following control model trained by the previous task for continuous learning to obtain a vehicle following control model trained under the incremental task; acquiring driving data of the following vehicle in real time, inputting the driving data into the trained vehicle following control model under the incremental task to obtain real-time predicted driving data, feeding back the real-time predicted driving data to the following vehicle, and adjusting the following behavior by the following vehicle according to the predicted driving data; and acquiring the driving data of the following vehicle after the following behavior is adjusted in real time, and taking the acquired driving data as the driving data of the next task. The generalization ability and stability of the automatic driving system are enhanced, the adaptive capacity of the vehicle following control system in the dynamic traffic environment is remarkably improved, and a safer and more efficient autonomous driving solution is provided.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of intelligent transportation, and particularly relates to an autonomous driving car-following control method based on brain-like continuous learning. BACKGROUND

[0002] With the rapid development of technology, autonomous driving technology has made significant progress. However, despite this, autonomous driving systems still face many challenges in dealing with complex and unpredictable driving environments. In particular, the "long tail problem", which means that some rare but critical driving scenarios may not be fully represented in the training data, makes the system lack sufficient adaptability when facing these rare scenarios, thereby increasing the safety risk.

[0003] Current autonomous driving systems mainly rely on offline training methods, which require a large amount of training data to teach the system how to deal with various driving situations. However, these systems have difficulty in dealing with rare or unforeseen events, especially when these situations are not fully demonstrated in the training data. In addition, the static nature of traditional models limits their dynamic adaptability after deployment, which is crucial for achieving high adaptability and safety in the real world.

[0004] To solve these problems, the concept of "continuous learning" has emerged as a promising solution. Continuous learning enables models to adapt to new data without forgetting previously learned knowledge. The challenge lies in balancing "stability" and "plasticity".

[0005] In the prior art, the mainstream method of continuous learning still has the following significant defects when dealing with complex scenarios such as autonomous driving car-following, which have strong time series and fuzzy task boundaries. First, the means of memory regulation is too static. The mainstream regularization method (such as elastic weight consolidation) relies on static anchoring of the historical parameter space, and cannot perform fine-grained selective forgetting and updating according to the task time sequence and the contribution of the parameters to the current learning, resulting in difficulty in balancing adaptability learning and historical knowledge retention. Second, the existing methods are too single in the protection dimension of knowledge. Existing methods focus on constraining individual parameters to approach the historical optimal value, but ignore the collaborative changes of functionally related parameter groups in the neural network during continuous learning, affecting the coherent transfer of knowledge and the generalization performance of the model. Finally, the learning process of the current method is seriously disconnected from the actual task dynamics. The memory retention strength of most methods is a fixed value set in advance, and cannot be adaptively adjusted according to the difficulty of new tasks (such as the comparison of new and old task losses) or the relevance between tasks. This makes the model prone to performance fluctuations or forgetting problems when facing complex and variable continuous tasks, especially when encountering new tasks with significantly increased difficulty or low relevance. SUMMARY

[0006] To solve the problems in the background art, the present application provides an autonomous driving car-following control method based on brain-like continuous learning, which solves the technical problems of serious forgetting of historical knowledge, insufficient adaptation to new tasks and unstable overall control performance of existing automatic driving car-following models when continuously learning new driving scenarios due to rigid memory mechanism, single knowledge protection dimension and disconnection between learning process and task dynamics.

[0007] The technical solution adopted by the present application is: One kind autonomous driving car-following control method based on brain-like continuous learning: S1, collect driving data of car-following vehicles under incremental tasks through vehicle-mounted sensors, and construct driving data set under incremental tasks according to driving data.

[0008] S2, input driving data set under incremental tasks into vehicle car-following control model trained under the previous task for continuous learning, and obtain vehicle car-following control model trained under incremental tasks.

[0009] S3, collect driving data of car-following vehicles in real time, input driving data into vehicle car-following control model trained under incremental tasks for processing to obtain real-time predicted driving data.

[0010] S4, feed back predicted driving data to car-following vehicles in real time, and car-following vehicles adjust car-following behavior according to predicted driving data.

[0011] S5, collect driving data of car-following vehicles after adjusting car-following behavior in real time, and collect driving data as driving data of next task.

[0012] The step S1 is specifically: S11, collect continuous time frame driving data of several car-following vehicles under incremental tasks through vehicle-mounted sensors.

[0013] S12, for each car-following vehicle, divide continuous time frame driving data into first time period driving data and second time period driving data in a preset proportion in time sequence, take first time period driving data as input driving data, and take second time period driving data as label driving data, thereby constructing single driving data sample.

[0014] S13, several car-following vehicles corresponding driving data samples are summarized to obtain driving data set under incremental tasks.

[0015] Each frame driving data of car-following vehicles includes speed of car-following vehicles, acceleration of car-following vehicles, speed of followed vehicles, acceleration of followed vehicles and distance between two vehicles.

[0016] The vehicle following control model includes several time-series prediction modules and an integration module; each time-series prediction module uses an LSTM network; and each LSTM network uses the MSE loss function during training.

[0017] The specific process of continuous learning for the vehicle car-following control model is as follows: D1. The driving dataset under the incremental task is input into each time series prediction module that has been trained in the previous task for training, resulting in several trained time series prediction modules. Each trained time series prediction module is then frozen.

[0018] D2. The driving dataset under the incremental task is input into each frozen time series prediction module. Each frozen time series prediction module outputs the corresponding predicted driving data. All predicted driving data are then input into the integration module for integration processing to obtain integrated predicted driving data. The integration module is continuously updated based on the integrated predicted driving data, the labeled driving data in the driving dataset, and the loss function, thereby obtaining the vehicle car-following control model trained under the incremental task.

[0019] The integration module is configured according to the following formula: I=Σ i=1 o (g i ·X i ) g i =softmax(ResNet1D([X1,…,X n ])) Where I represents the integrated predicted driving data obtained by the integration module; i is the index; o is the total number of time-series prediction modules; g i The output of the i-th time series prediction module is assigned a weight; X i Represents the output of the i-th time series prediction module; X1, X i and X n These represent the outputs of the 1st, 1st, and 1st time series prediction modules, respectively.

[0020] The loss function of the integration module is set according to the following formula: J LA (θ LA )=L B LA +Ω MSUM LA (θ LA )+Ω GMMFM LA (θ LA )+Ω H (g) L BLA =(1 / n)Σ i=1 n (I i -Y i ) 2 Oh MSUM LA (i LA )=λ MSUM S t<T S m=1 d (a t ·g e,m |θ LA,m -θ e,m |) g e,m =((|▽ θLA,m ·L e |) / (S m=1 d |▽ θLA,m L e |))·(L e / L past ) Oh GMMFM LA (i LA )=λ GMMFM S t<T S m=1 d [c GMMFM g m (i LA,m -θ LA,m * ) 2 +S j=1 d l m,j |θ LA,m -θ LA,j |] g m =(|▽ θLA,m ·L past |) / (S m=1 d |▽ θLA,m ·L past |) Oh H (g)=-λ H S i=1 o (g i log(g i )) Among them, J LA (i LA ) for the total loss function of the integrated module; i LAFor the set of parameters in the integration module; L B LA Ω is the task loss function; MSUM LA (θ LA ) represents the memory-selective update loss function; Ω GMMFM LA (θ LA ) represents the global memory transfer and feedback loss function; Ω H (g) represents the weight entropy regularization loss function; n is the sample size of a single training batch; i is the index; I i Y represents the predicted driving data corresponding to the i-th sample; i Let λ be the label driving data corresponding to the i-th sample; MSUM The weights are selectively updated for memory purposes; t is the index, representing the t-th historical task; α t Let be the temporal preservation coefficient of the t-th historical task; T be the index of the incremental task; m be the index, m∈{1,…,d}, and d be the parameter set θ. LA The total number of trainable parameters in θ; LA,m θ represents the m-th parameter value in the integration module during the incremental task training process. e,m To determine the m-th parameter value in the empty network during the incremental task training process when the integrated module is replaced with an empty network; ζ e,m For the migration weights of incremental tasks; L e For the loss during incremental task training in an empty network, the loss function of the empty network is the task loss function; L past The optimal loss in the historical task; |▽ θLA,m ·L e | is the parameter θ in the incremental task LA,m The magnitude of the gradient; Σ m=1 d |▽ θLA,m ·L e | represents the sum of gradients for all parameters during the incremental task training process; λ GMMFM To record the hyperparameters of global memory transfer and feedback; γ GMMFM θ is a hyperparameter used to control the strength of global memory transfer. LA,m * λ represents the m-th parameter value corresponding to the optimal loss in the historical task; j represents the index, j∈{1,…,d}; m,j θ is the feedback transfer factor between the m-th and j-th parameters in the integration module. LA,j ζ represents the j-th parameter value in the integration module during the incremental task training process. m For the migration weight of historical tasks; λ H g represents the entropy regularization coefficient; iThe output of the i-th time series prediction module contributes a weight; log is the logarithmic function; o is the total number of time series prediction modules.

[0021] Step S3 specifically involves: S31. Collect N consecutive frames of driving data from the N-1 frames before the current time to the current time frame.

[0022] S32. Input N consecutive frames of driving data into each time-series prediction module in the vehicle following control model, and each time-series prediction module outputs the corresponding predicted driving data.

[0023] S33. All predicted driving data are processed in the integration module to obtain integrated predicted driving data, which is the final real-time predicted driving data.

[0024] II. A computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the above-described method.

[0025] 3. A computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of the above method.

[0026] The innovation of this invention lies in processing the driving data of following vehicles, introducing a biomimetic mechanism similar to a brain-like neural system, simulating the structure and memory and forgetting mechanism of the fruit fly olfactory neural system, designing a multi-level continuous learning architecture, and combining the memory selective update mechanism, global memory transfer and feedback mechanism, and weight entropy regularization mechanism proposed in this invention. This effectively improves the model's continuous learning ability, enhances the model's adaptability in long-tail events, realizes the continuous learning and online optimization of vehicles in dynamic traffic environments for autonomous driving systems, and improves the accuracy and stability of vehicle following behavior. It has stronger practical significance than traditional static models.

[0027] The beneficial effects of this invention are: This invention employs a multi-layered continuous learning architecture based on a brain-like continuous learning mechanism. The incremental learning method enables the system to learn and optimize online in real time, enhancing its ability to cope with long-tail events and complex scenarios. Simultaneously, by combining a selective memory update mechanism and a global memory transfer and feedback mechanism, it effectively balances learning new tasks with remembering old tasks, avoiding catastrophic forgetting. This invention enhances the generalization ability and stability of autonomous driving systems, significantly improves the adaptability of vehicle following control systems in dynamic traffic environments, and provides a safer and more efficient autonomous driving solution. Attached Figure Description

[0028] Figure 1 This is a flowchart of the present invention.

[0029] Figure 2 A flowchart for continuous learning.

[0030] Figure 3 This is a schematic diagram of the biomimetic architecture of the multi-level continuous learning model inspired by the fruit fly brain in Example 1.

[0031] Figure 4 This is a diagram showing the results of continuous learning of the vehicle following control model in Example 1.

[0032] Figure 5 The above are simulated speed heatmaps of the 25-vehicle convoy following scenario in Example 1 and Comparative Examples 1-3. Detailed Implementation

[0033] The present invention will now be described in more detail with reference to the accompanying drawings and embodiments. However, the present invention is not limited thereto. For those skilled in the art, several improvements and modifications can be made without departing from the principles of the present invention, and these improvements and modifications are also considered to be within the scope of protection of the present invention. Contents not described in detail in this specification are prior art known to those skilled in the art.

[0034] Example 1:

[0035] like Figure 1 As shown, the autonomous driving following control method of this embodiment includes the following steps: S1. Collect driving data of following vehicles under incremental tasks through on-board sensors, and construct a driving dataset under incremental tasks based on the driving data.

[0036] like Figure 2 As shown, the continuous learning process is divided into historical tasks and incremental tasks. Historical tasks are the tasks that the model has learned and trained on, while incremental tasks are the tasks that the current model needs to learn.

[0037] S11. Collect continuous time frame driving data of several following vehicles under incremental tasks through the vehicle's onboard sensors.

[0038] Each frame of driving data for each following vehicle includes the following vehicle's speed, the following vehicle's acceleration, the speed of the vehicle being followed, the acceleration of the vehicle being followed, and the distance between the two vehicles.

[0039] S12. For each following vehicle, divide the continuous time frame driving data into a first time period driving data and a second time period driving data in chronological order. Use the first time period driving data as input driving data and the second time period driving data as labeled driving data to construct a single driving data sample.

[0040] S13. The driving data samples corresponding to several following vehicles are summarized to obtain the driving dataset under the incremental task.

[0041] S2. Input the driving dataset under the incremental task into the vehicle car-following control model based on brain-like continuous learning trained in the previous task for continuous learning, and obtain the vehicle car-following control model trained under the incremental task; the vehicle car-following control model includes several time-series prediction modules and integration modules. In this embodiment, mimicking the structure and function of the γMB system in the Drosophila nervous system, a multi-layered continuous learning architecture is designed. This includes an integration module and multiple parallel temporal prediction modules, each with its own independent learning capability. Figure 3 As shown, this embodiment is inspired by the parallel processing structure of the five regions of the γMB system in the fruit fly brain, and designs a biomimetic architecture (vehicle car-following control model) using a multi-level continuous learning model. This architecture simulates the mechanism of different functional regions working together in a biological neural system, consisting of five parallel LSTM networks (denoted as L...). B –L F ) and an integration module L A Composition: Five LSTM learning modules independently learn driving data of different feature dimensions, each with an independent forgetting rate to simulate the differential regulation of γ regions by DPM neurons in the fruit fly brain; integration module L A ResNet(1D) is used as the specific implementation of the weighted fusion network, which integrates the outputs of the five learning modules based on the optimal weights obtained from the core training network. The MSE loss function is used during the training of each LSTM network.

[0042] The specific process of continuous learning for the vehicle car-following control model is as follows: D1. The driving dataset under the incremental task is input into each time series prediction module that has been trained in the previous task for training, resulting in several trained time series prediction modules. Each trained time series prediction module is then frozen.

[0043] D2. The driving dataset under the incremental task is input into each frozen time series prediction module. Each frozen time series prediction module outputs the corresponding predicted driving data. All predicted driving data are then input into the integration module for integration processing to obtain integrated predicted driving data. The integration module is continuously updated based on the integrated predicted driving data, the labeled driving data in the driving dataset, and the loss function, thereby obtaining the vehicle car-following control model trained under the incremental task.

[0044] The goal of the integration module is to achieve optimal synergy between different learning modules through weight fusion, while maintaining the model's balanced learning ability for incremental and historical tasks. The integration module is configured according to the following formula: I=Σ i=1 o (g i ·X i ) g i =softmax(ResNet1D([X1,…,X n ])) Where I represents the integrated predicted driving data obtained by the integration module; i is the index; o is the total number of time-series prediction modules; g i X represents the output contribution weight corresponding to the i-th time series prediction module, which is obtained by normalizing all predicted driving data from ResNet (1D) after softmax; softmax() represents the softmax function; ResNet1D([]) represents the one-dimensional ResNet network model; X i Represents the output of the i-th time series prediction module; X1, X i and X n These represent the outputs of the 1st, 1st, and 1st time series prediction modules, respectively.

[0045] The loss function for the integration module is set according to the following formula: J LA (θ LA )=L B LA +Ω MSUM LA (θ LA )+Ω GMMFM LA (θ LA )+Ω H (g) L B LA =(1 / n)Σ i=1 n (I i -Y i ) 2 Ω MSUM LA (θ LA )=λ MSUM Σ t<T Σ m=1 d (α t ·ζ e,m |θ LA,m -θ e,m |) ζ e,m =((|▽ θLA,m ·L e |) / (Σm=1 d |▽ θLA,m ·L e |))·(L e / L past ) Ω MSUM LA (θ LA The loss function ζ is a memory-selective update function that simulates the selective retention and forgetting mechanism of memory regulated by Rac1 activity in the fruit fly brain. It dynamically allocates parameter update intensity based on parameter importance and the relative difficulty of historical task loss and incremental task loss. Parameters with larger gradients are considered more important to the current incremental task and given higher transfer weights. e,m This allows for more significant updates; simultaneously, the greater the ratio of incremental task loss to historical task loss, the stronger the overall migration. Temporal preservation coefficient α t This approach incorporates the biological principle of memory decay over time, resulting in a lower weighting of historical tasks on current parameters. Its advantages include: achieving refined parameter-level memory management by differentially forgetting old memories based on the importance of parameters in the current task and the temporal relationship of tasks, thus more effectively mitigating catastrophic forgetting. Simultaneously, it enhances the model's adaptability to incremental tasks by focusing on targeted updates to parameters important for the current task, accelerating the model's convergence speed on new tasks.

[0046] Ω GMMFM LA (θ LA )=λ GMMFM Σ t<T Σ m=1 d [γ GMMFM ζ m (θ LA,m -θ LA,m * ) 2 +Σ j=1 d λ m,j |θ LA,m -θ LA,j |] ζ m =(|▽ θLA,m ·L past |) / (Σ m=1 d |▽ θLA,m ·L past |) Ω GMMFM LA (θ LAThe function is a global memory transfer and feedback loss function, which mimics the memory integration and consolidation process across brain regions and time scales in the fruit fly brain. This function constrains the historically optimal parameters, with the constraint strength determined by the importance ζ of the parameters in the historical task. m and global memory transfer strength coefficient γ GMMFM This collaborative decision-making process protects important historical knowledge from being overwritten. Simultaneously, a feedback transfer constraint is introduced between parameters, using a feedback transfer factor λ. m,j This method measures and constrains the consistency of changes in functionally coupled parameter pairs (m,j) during incremental learning, simulating the cooperative adaptation characteristics of functionally related neuron groups in neural networks. Its advantages include: achieving global knowledge consolidation by explicitly anchoring historically optimal parameters, effectively combating historical performance degradation caused by learning new tasks. Simultaneously, it emphasizes cooperative transfer between parameters; the feedback transfer term constrains not only individual parameters but also the relationships between parameters, enabling the model to maintain the beneficial associations learned within or between functional modules when adapting to new tasks, enhancing the coherence of knowledge transfer and the model's generalization ability.

[0047] Ω H (g)=-λ H Σ i=1 o (g i log(g i )) Ω H (g) is the weight entropy regularization loss function, designed to prevent the "information collapse" phenomenon where the weight distribution becomes overly concentrated in one or a few modules when the integration module weights and fuses the outputs of various time-series prediction modules. By minimizing the negative entropy of the weight distribution g, it encourages the weight vector to tend towards a uniform distribution. Its advantages include: promoting diversity in module utilization, ensuring that all parallel time-series prediction modules contribute to the final decision, and avoiding insufficient model capacity utilization due to the dominance of a single module. Furthermore, it improves the robustness of the model; the dispersion of weights reduces the risk of overfitting the model to a single module or specific feature dimension, thereby improving the model's stability when facing out-of-distribution data or noise.

[0048] Among them, J LA (θ LA ) represents the total loss function of the integration module; θ LA For the set of parameters in the integration module; L B LA Ω is the task loss function; MSUM LA (θ LA ) represents the memory-selective update loss function; Ω GMMFM LA (θ LA) represents the global memory transfer and feedback loss function; Ω H (g) represents the weight entropy regularization loss function; n is the sample size of a single training batch; i is the index; I i Y represents the predicted driving data corresponding to the i-th sample in the training batch; i λ represents the label driving data corresponding to the i-th sample in the training batch; MSUM The preset memory-selective update weights (hyperparameters controlling the intensity of forgetting) are used; t is the index, representing the t-th historical task; α t Let be the temporal preservation coefficient of the t-th historical task, used to reflect the biological law of memory decay over time; T is the index of the incremental task; m is the index, m∈{1,…,d}, and d is the parameter set θ. LA The total number of trainable parameters in θ; LA,m θ represents the m-th parameter value in the integration module during the incremental task training process. e,m To determine the m-th parameter value in the empty network during the incremental task training process when the integrated module is replaced with an empty network; ζ e,m L represents the migration weight for the incremental task, indicating the intensity at which the m-th parameter value needs to be updated in the incremental task. It is determined based on the importance of the parameter in the incremental task. e For the loss during incremental task training in an empty network, the loss function of the empty network is the task loss function; L past The optimal loss in the historical task; |▽ θLA,m ·L e | is the parameter θ in the incremental task LA,m The magnitude of the gradient indicates the importance of the parameter in the incremental task; a larger gradient means that the parameter contributes more to the loss function and should be updated more strongly. m=1 d |▽ θLA,m ·L e | is the sum of gradients for all parameters during incremental task training, used to normalize the gradient of individual parameters; L e / L past λ represents the ratio of the incremental task's loss to the historical task's loss, indicating the difficulty of the incremental task. A higher incremental task loss results in a higher transfer weight, and vice versa. GMMFM To record the hyperparameters of global memory transfer and feedback; γ GMMFM θ is a hyperparameter used to control the strength of global memory transfer. LA,m * Let be the m-th parameter value corresponding to the optimal loss in the historical task, used to characterize the existing knowledge that should be retained; j represents the index, j∈{1,…,d}, and d is the parameter set θ. LA The total number of trainable parameters in the λ range; m,jθ is the feedback transfer factor between the m-th and j-th parameters in the integration module during the incremental task training process, used to measure the functional coupling strength between the m-th and j-th parameters; LA,j ζ represents the j-th parameter value in the integration module during the incremental task training process. m λ represents the transfer weight for historical tasks, indicating the strength of the m-th parameter value that needs to be remembered in the historical task; H This represents the entropy regularization coefficient, used to adjust the balance of weights across modules, preventing information collapse caused by excessively high weights in a single module, and ensuring the diversity and robustness of model learning; g i The output of the i-th time series prediction module contributes a weight; log is the logarithmic function; o is the total number of time series prediction modules.

[0049] In this embodiment, the memory forgetting mechanism is modeled after the memory forgetting mechanism regulated by the small G protein Rac1 in the fruit fly brain. Simulating the effect of Rac1 activity on memory retention time, a memory-selective update loss function and a global memory transfer and feedback loss function are incorporated into the model's loss function. By adjusting the active forgetting rate, the relationship between new task learning and historical task memory is balanced, thereby avoiding catastrophic forgetting and enhancing the model's adaptability. The memory-selective update loss function selects an appropriate forgetting rate for historical memories based on the relevance and importance of the input data, where the forgetting rate β is expressed by the following formula: λ MSUM =β / (1-β) The memory selective update loss function and the global memory transfer and feedback loss function are compared through λ. GMMFM and λ MSUM The two hyperparameters form a complementary control mechanism. Specifically, λ... GMMFM From λ MSUM It can be directly derived that: λ GMMFM =1-λ MSUM Furthermore: When training the baseline model (Model 0) corresponding to the vehicle car-following control model, the integration module uses L... B LA +Ω H (g) Loss.

[0050] S3. Real-time acquisition of driving data from following vehicles, inputting the driving data into the vehicle following control model trained under the incremental task for processing to obtain predicted driving data.

[0051] S31. Collect N consecutive frames of driving data from the N-1 frames before the current time to the current time frame.

[0052] S32. Input N consecutive frames of driving data into each time-series prediction module in the vehicle following control model, and each time-series prediction module outputs the corresponding predicted driving data.

[0053] S33. All predicted driving data are processed in the integration module to obtain integrated predicted driving data, which is the final real-time predicted driving data.

[0054] S4. Feedback the predicted driving data obtained in real time to the following vehicles, and the following vehicles adjust their following behavior according to the predicted driving data.

[0055] S5. Collect driving data of the following vehicle in real time after adjusting its following behavior, and use the collected driving data as the driving data for the next task.

[0056] Furthermore, after real-time prediction and data collection for a period of time, the current car-following vehicle is no longer predicted. The next task is treated as a new incremental task, and steps S1-S2 are repeated to obtain a new car-following control model trained under the new incremental task. Then, steps S3-S5 are executed, and this process is repeated in sequence to achieve continuous learning.

[0057] This embodiment also provides a computer device, including a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the steps of the above-described method.

[0058] This embodiment also provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the steps of the above-described method.

[0059] In this embodiment, experiments were conducted using highD data collected on an open highway in a certain country and data collected on January 19, 2013, on a 3.2km stretch of Innovation Avenue in a certain region, showing vehicles following each other. The total data collection time was 269 minutes, with a frame rate of 10 frames per second. Data including the speed and acceleration of both the following and being followed vehicles, and the distance between them, were obtained.

[0060] In this embodiment, the driving data for the first time period and the driving data for the second time period are each 50 frames, with k taking 10 frames, for a total of 6 seconds of driving data. A total of over 110,000 following vehicles and 43.9 million data points were collected, of which 40 million data points were used for training model 0 (the baseline model), and 3.9 million data points were used for continuous learning.

[0061] In this embodiment, highD data collected on a highway in a certain country will be used as training data for model 0 (baseline model), and platoon following experiment data conducted in a certain region will be used as data for incremental tasks. The experiment is divided into 3 groups, with each group consisting of a platoon of 25 vehicles completing a total of 12 single-trip driving tasks. Each single trip includes three typical driving scenarios: rapid acceleration, constant speed driving, and rapid deceleration. At the same time, different vehicle speeds are set in the constant speed driving phase of each single trip to increase data diversity.

[0062] like Figure 4 The diagram shows the results of the vehicle car-following control model in this embodiment under 36 tasks. The lines of different colors represent the model's iterative process, with lighter lines indicating later iterations. As the model is trained with incremental tasks, its car-following results gradually approach real-world driving data and stabilize. In other words, as the continuous learning process progresses, the model's car-following level continuously improves and remains stable, indicating that the model has continuous learning capabilities, good adaptability, and can achieve smooth operation in road environments.

[0063] Comparative Example 1: This comparative example uses the same method as Example 1 to conduct an experiment with a car-following convoy of 25 vehicles. The car-following control model in this comparative example uses the model of this invention, and only the baseline model (Model 0) needs to be trained; no further learning is performed for subsequent incremental tasks.

[0064] The comparative vehicle following control model's simulation results in a 25-vehicle platoon are as follows: Figure 5 As shown in (c), the consistency indicators with the real data are compared in Table 1. Through Figure 5 As shown in Table 1, the performance of the vehicle following control model in this comparative example is far lower than that of the vehicle following control model used in Example 1.

[0065] Comparative Example 2: This comparative example uses the same method and dataset as Example 1 to conduct a 25-vehicle platoon car-following simulation experiment. However, in this comparative example, the vehicle car-following control model adopts a reinforcement learning (RL) model.

[0066] The comparative vehicle following control model's simulation results in a 25-vehicle platoon are as follows: Figure 5 As shown in (d), the consistency indicators with the real data are compared in Table 1. Through Figure 5 As shown in Table 1, the performance of the vehicle following control model in this comparative example is far lower than that of the vehicle following control model used in Example 1.

[0067] Comparative Example 3: This comparative example uses the same method and dataset as Example 1 to conduct a 25-vehicle platoon car-following simulation experiment. However, in this comparative example, the vehicle car-following control model adopts a physical model-driven intelligent driver model (IDM).

[0068] The consistency index of the vehicle following control model in this comparative example with the real data in the 25-vehicle queuing simulation is shown in Table 1. Table 1 shows that the performance of the vehicle following control model in this comparative example is far lower than that of the vehicle following control model used in Example 1.

[0069] As shown in Table 1, the consistency index between the simulated and real car-following scenarios of each model queue reveals that the simulation results of the brain-inspired continuous learning model of this invention are close to the real data, significantly outperforming the initial model, reinforcement learning model, and IDM model. These results demonstrate that the biomimetic continuous learning model can effectively support car-following incremental task learning and is significantly superior to traditional data-driven methods and physical models in terms of plasticity.

[0070] Table 1

[0071] Depend on Figure 5 A comparison of the speed heatmap results of the 25-vehicle platoon following scenario simulations of the various models shows that the method of this invention can enable the model to maintain the integrity of existing knowledge while dynamically adapting to new tasks through a multi-level error correction mechanism, significantly improving the stability of the vehicle following system. Its performance advantages are particularly prominent in the rear of the platoon and in long-term following scenarios.

[0072] Comparative Example 4: This comparative example is implemented using the same method as Example 1, but the loss function of the integration module is set as follows: J LA (θ LA )1=L B LA The results are shown in Table 2.

[0073] Comparative Example 5: This comparative example is implemented using the same method as Example 1, but the loss function of the integration module is set as follows: J LA (θ LA )2=L B LA +Ω MSUM LA (θ LA The results are shown in Table 2.

[0074] Comparative Example 6: This comparative example is implemented using the same method as Example 1, but the loss function of the integration module is set as follows: J LA (θLA )3=L B LA +Ω GMMFM LA (θ LA The results are shown in Table 2.

[0075] Comparative Example 7: This comparative example is implemented using the same method as Example 1, but the loss function of the integration module is set as follows: J LA (θ LA )4=L B LA + Ω H (g) The results are shown in Table 2.

[0076] Comparative Example 8: This comparative example is implemented using the same method as Example 1, but the loss function of the integration module is set as follows: J LA (θ LA )5=L B LA +Ω MSUM LA (θ LA )+Ω GMMFM LA (θ LA The results are shown in Table 2.

[0077] Comparative Example 9: This comparative example uses the same method as Example 1, but the loss function of the integration module is set as follows: J LA (θ LA )6=L B LA +Ω GMMFM LA (θ LA )+Ω H (g) The results are shown in Table 2.

[0078] Comparative Example 10: This comparative example is implemented using the same method as Example 1, but the loss function of the integration module is set as follows: J LA (θ LA )7=L B LA +Ω MSUM LA (θ LA )+Ω H (g) The results are shown in Table 2.

[0079] Table 2

[0080] This invention proposes a brain-inspired continuous learning-based autonomous driving car-following control method, inspired by the neural mechanisms of biological organisms, particularly the fruit fly brain, which exhibits remarkable adaptability and memory retention through active forgetting and memory plasticity. By simulating these biological processes, this invention enables autonomous vehicles to continuously learn and adapt in dynamic driving environments, especially in modeling vehicle car-following behavior. The method integrates multiple temporal prediction modules with different forgetting rates, ensuring the stability of learned tasks while flexibly incorporating new driving scenarios.

[0081] The above embodiments are merely preferred embodiments provided to fully illustrate the present invention, and the scope of protection of the present invention is not limited thereto. Equivalent substitutions or modifications made by those skilled in the art based on the present invention are all within the scope of protection of the present invention. The scope of protection of the present invention is defined by the claims.

Claims

1. An autonomous driving car-following control method based on brain-like continuous learning, characterized in that, Includes the following steps: S1. Collect driving data of the following vehicles under the incremental task through vehicle-mounted sensors, and construct the driving dataset under the incremental task based on the driving data; S2. Input the driving dataset under the incremental task into the vehicle car-following control model trained in the previous task for continuous learning, and obtain the vehicle car-following control model trained under the incremental task. S3. Real-time acquisition of driving data of following vehicles, input of driving data into the vehicle following control model trained under incremental task for processing to obtain real-time predicted driving data. S4. Feedback the real-time predicted driving data to the following vehicles, and the following vehicles adjust their following behavior according to the predicted driving data. S5. Collect driving data of the following vehicle in real time after adjusting its following behavior, and use the collected driving data as the driving data for the next task.

2. The autonomous driving following control method based on brain-like continuous learning according to claim 1, characterized in that, Step S1 specifically involves: S11. Collect continuous time frame driving data of several following vehicles under incremental tasks through on-board sensors; S12. For each following vehicle, the continuous time frame driving data is divided into a first time period driving data and a second time period driving data in a preset ratio according to the time sequence. The first time period driving data is used as the input driving data and the second time period driving data is used as the label driving data, thereby constructing a single driving data sample. S13. The driving data samples corresponding to several following vehicles are summarized to obtain the driving dataset under the incremental task.

3. The autonomous driving following control method based on brain-like continuous learning according to claim 2, characterized in that: Each frame of driving data for the following vehicle includes the speed of the following vehicle, the acceleration of the following vehicle, the speed of the vehicle being followed, the acceleration of the vehicle being followed, and the distance between the two vehicles.

4. The autonomous driving following control method based on brain-like continuous learning according to claim 1, characterized in that: The vehicle following control model includes several timing prediction modules and an integration module; each timing prediction module uses an LSTM network.

5. The autonomous driving following control method based on brain-like continuous learning according to claim 4, characterized in that, The specific process of continuous learning for the vehicle car-following control model is as follows: D1. The driving dataset under the incremental task is input into each time series prediction module that has been trained in the previous task for training, resulting in several trained time series prediction modules. Each trained time series prediction module is then frozen. D2. The driving dataset under the incremental task is input into each frozen time series prediction module. Each frozen time series prediction module outputs the corresponding predicted driving data. All predicted driving data are then input into the integration module for integration processing to obtain integrated predicted driving data. The integration module is continuously updated based on the integrated predicted driving data, the labeled driving data in the driving dataset, and the loss function, thereby obtaining the vehicle car-following control model trained under the incremental task.

6. The autonomous driving following control method based on brain-like continuous learning according to claim 5, characterized in that: The integration module is configured according to the following formula: I=S i=1 o (g i ·X i ) g i =softmax(ResNet1D([X1,…,X n ])) Where I represents the integrated predicted driving data obtained by the integration module; i is the index; o is the total number of time-series prediction modules; g i The output of the i-th time series prediction module is assigned a weight; X i Represents the output of the i-th time series prediction module; X1, X i and X n These represent the outputs of the 1st, 1st, and 1st time series prediction modules, respectively.

7. The autonomous driving following control method based on brain-like continuous learning according to claim 5, characterized in that: The loss function of the integration module is set according to the following formula: J LA (i LA )=L B LA +Oh MSUM LA (i LA )+Ω GMMFM LA (i LA )+Ω H (g) L B LA =(1 / n)Σ i=1 n (I i -Y i ) 2 Oh MSUM LA (i LA )=λ MSUM S t<T S m=1 d (a t ·g e,m |θ LA,m -θ e,m |) g e,m =((|▽ θLA,m ·L e |) / (S m=1 d |▽ θLA,m L e |))·(L e / L past ) Oh GMMFM LA (i LA )=λ GMMFM S t<T S m=1 d [c GMMFM g m (i LA,m -θ LA,m * ) 2 +S j=1 d l m,j |θ LA,m -θ LA,j |] g m =(|▽ θLA,m ·L past |) / (S m=1 d |▽ θLA,m ·L past |) Oh H (g)=-λ H S i=1 o (g i log(g i )) Among them, J LA (θ LA ) represents the total loss function of the integration module; θ LA For the set of parameters in the integration module; L B LA Ω is the task loss function; MSUM LA (θ LA ) represents the memory-selective update loss function; Ω GMMFM LA (θ LA ) represents the global memory transfer and feedback loss function; Ω H (g) represents the weight entropy regularization loss function; n is the sample size of a single training batch; i is the index; I i Y represents the predicted driving data corresponding to the i-th sample; i Let λ be the label driving data corresponding to the i-th sample; MSUM The weights are selectively updated for memory purposes; t is the index, representing the t-th historical task; α t Let be the temporal preservation coefficient of the t-th historical task; T be the index of the incremental task; m be the index, m∈{1,…,d}, and d be the parameter set θ. LA The total number of trainable parameters in θ; LA,m θ represents the m-th parameter value in the integration module during the incremental task training process. e,m The value of the m-th parameter in the empty network during the incremental task training process, α, when the integration module is replaced with an empty network. t ζ is the timing preservation coefficient; e,m For the migration weights of incremental tasks; L e For the loss during incremental task training in an empty network, the loss function of the empty network is the task loss function; L past The optimal loss in the historical task; |▽ θLA,m ·L e | is the parameter θ in the incremental task LA,m The magnitude of the gradient; Σ m=1 d |▽ θLA,m ·L e | represents the sum of gradients for all parameters during the incremental task training process; λ GMMFM To record the hyperparameters of global memory transfer and feedback; γ GMMFM θ is a hyperparameter used to control the strength of global memory transfer. LA,m * λ represents the m-th parameter value corresponding to the optimal loss in the historical task; j represents the index, j∈{1,…,d}; m,j θ is the feedback transfer factor between the m-th and j-th parameters in the integration module. LA,j ζ represents the j-th parameter value in the integration module during the incremental task training process. m For the migration weight of historical tasks; λ H g represents the entropy regularization coefficient; i The output of the i-th time series prediction module contributes a weight; log is the logarithmic function; o is the total number of time series prediction modules.

8. The autonomous driving following control method based on brain-like continuous learning according to claim 5, characterized in that, Step S3 specifically involves: S31. Collect N consecutive frames of driving data from the N-1 frames before the current time to the current time frame; S32. Input N consecutive frames of driving data into each time-series prediction module in the vehicle following control model, and each time-series prediction module outputs the corresponding predicted driving data. S33. All predicted driving data are processed in the integration module to obtain integrated predicted driving data, which is the final real-time predicted driving data.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 8.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 8.

Images