Operation control method and device and unmanned vehicle

Abstract

The invention provides an operation control method and device and an unmanned vehicle, and relates to the technical field of unmanned driving, and the method comprises the steps: obtaining the environment information of an operation region where a first vehicle is located, and the first vehicle is an unmanned vehicle; identifying the environment information to obtain identification information, wherein the identification information comprises an operation scene where the first vehicle is located and a driving style and a driving intention of a second vehicle in the operation area; risk assessment is carried out according to reference information to obtain the risk level of the current risk of the first vehicle, and the reference information comprises the identification information; predicting an action to be taken by the second vehicle according to the driving style and the driving intention of the second vehicle; and generating a control instruction for controlling the first vehicle according to the risk level of the current risk of the first vehicle and the predicted action to be taken by the second vehicle.

Description

Technical Field

[0001] This disclosure relates to the field of unmanned driving technology, and in particular to an operation control method, device and unmanned vehicle. Background Technology

[0002] With the deepening of smart mine construction, autonomous driving technology has become a key development direction for improving mine transportation efficiency and ensuring production safety. Autonomous driving systems typically rely on high-precision positioning, environmental perception, and control algorithms, and have been effectively applied in the relatively closed environment of mining areas.

[0003] In related technologies, most autonomous driving systems make decisions primarily based on collision risk assessment, such as calculating time-to-collision (TTC) and safe distance. The core logic of the decision-making is to brake or avoid a collision if one is predicted to occur; otherwise, continue driving along a predetermined trajectory. Summary of the Invention

[0004] The inventors noted that there are limitations in the relevant technologies, especially in the transition phase where manned and unmanned vehicles coexist, and when facing extremely complex working conditions in mines, the decision-making capabilities of the autonomous driving systems in the relevant technologies appear particularly rigid and insufficient, making it difficult to meet the requirements for safe, efficient, and smooth transportation in certain specific scenarios (such as mining scenarios).

[0005] First, it cannot handle the complex interactions in mixed environments.

[0006] In mixed-traffic scenarios, the behavior of human-driven vehicles is highly unpredictable and human-like. Human drivers rely on experience, visual communication, and anticipation of intent to coordinate. For example, a conservative driver will proactively yield to an aggressive driver, and an empty vehicle will understand the significant risks of a fully loaded vehicle going downhill and will give way. The autonomous driving systems in related technologies lack a deep understanding of these driving styles, intentions, and scenario-specific risks. They can only make mechanical binary (safety / danger) judgments, unable to make smooth and efficient collaborative decisions like humans, resulting in abrupt interactions and potentially triggering emergency braking, impacting transportation efficiency.

[0007] Secondly, the asymmetric risks in specific scenarios are ignored.

[0008] Certain scenarios may inherently possess "asymmetry." For example, a fully loaded vehicle on a downhill slope has extremely high inertia, resulting in a significantly longer braking distance and a much higher risk of loss of control compared to an empty vehicle or a vehicle on a flat road. Based on this common sense, human drivers would proactively grant a fully loaded vehicle going downhill more right-of-way, even if no collision occurs. Autonomous driving systems in related technologies rely solely on immediate collision assessments and cannot understand this deep, asymmetric risk distribution. They might make dangerous decisions to maintain a safe distance even when necessary, posing serious safety hazards.

[0009] Secondly, there is a lack of contextualized understanding of driving intentions.

[0010] Some transportation scenarios have specific contexts (such as loading areas, unloading areas, sharp bends, and long slopes) and task states (empty or fully loaded). Human drivers anticipate the intentions of other vehicles based on this contextual information. For example, a vehicle ahead lingering at a loading point may be about to reverse into a parking space; a vehicle on the right, being empty, may be preparing to enter the loading area. Autonomous driving systems in related technologies can only identify the position and speed of obstacles; they cannot integrate contextual and state information to infer deeper intentions, resulting in a lack of foresight in decision-making and thus impacting transportation efficiency.

[0011] In view of this, the present disclosure proposes the following solutions to improve the safety and efficiency of unmanned vehicle operations.

[0012] According to a first aspect of the present disclosure, a job control method is provided, comprising: acquiring environmental information of a job area where a first vehicle is located, wherein the first vehicle is an unmanned vehicle; identifying the environmental information to obtain identification information, wherein the identification information includes the job scenario where the first vehicle is located, and the driving style and driving intention of a second vehicle within the job area; performing a risk assessment based on reference information to obtain a risk level of the current risk of the first vehicle, wherein the reference information includes the identification information; predicting the actions that the second vehicle will take based on the driving style and driving intention of the second vehicle; and generating control commands for controlling the first vehicle based on the risk level of the current risk of the first vehicle and the predicted actions that the second vehicle will take.

[0013] In some embodiments, generating control instructions for controlling the first vehicle based on the risk level of the current risk situation of the first vehicle and the predicted actions that the second vehicle will take includes: determining a control strategy for controlling the first vehicle based on the risk level of the current risk situation of the first vehicle and the predicted actions that the second vehicle will take, the control strategy including acceleration, deceleration, maintaining current speed or steering; and generating control instructions for controlling the first vehicle based on the risk level of the current risk situation of the first vehicle and the control strategy.

[0014] In some embodiments, determining a control strategy for controlling the first vehicle based on the risk level of the current risk situation of the first vehicle and the predicted actions to be taken by the second vehicle includes: determining the candidate strategy from the strategy combination with the largest value of the benefit function among multiple strategy combinations as the control strategy, wherein each strategy combination consists of one candidate strategy from the candidate strategy set of the first vehicle and the predicted actions to be taken by the second vehicle, wherein the benefit function comprehensively quantifies safety benefits, efficiency benefits and comfort benefits, and the higher the risk level of the current risk situation of the first vehicle, the greater the weight of the safety benefits.

[0015] In some embodiments, generating control instructions for controlling the first vehicle based on the risk level of the current risk of the first vehicle and the control strategy includes: planning the motion trajectory of the first vehicle based on the risk level of the current risk of the first vehicle and the control strategy; converting the motion trajectory into control quantity instructions; and sending the control quantity instructions to the actuators of the first vehicle, wherein the control quantity includes throttle control quantity, brake control quantity, and steering control quantity.

[0016] In some embodiments, the identification information may also include the load status of the second vehicle.

[0017] In some embodiments, the reference information further includes at least one of the relative speed and relative distance between the first vehicle and the second vehicle.

[0018] In some embodiments, performing a risk assessment based on reference information to obtain the risk level of the current situation of the first vehicle includes: determining a matching risk level from a risk level knowledge base that matches each piece of information in the reference information, the risk level knowledge base being pre-built based on historical accidents; and determining the matching risk level as the risk level of the current situation of the first vehicle.

[0019] In some embodiments, the environmental information includes first information, which includes the speed, longitudinal acceleration, lateral acceleration, and lateral deviation from the desired lane of the second vehicle; the driving style is obtained by identifying the first information.

[0020] In some embodiments, the first information may also include the distance between the second vehicle and the vehicle in front.

[0021] In some embodiments, the environmental information includes second information, which includes the trajectory curvature, lateral acceleration, and lateral velocity of the second vehicle; the driving intention is obtained by recognizing the second information.

[0022] In some embodiments, the second information may also include the turn signal status and wheel angle of the second vehicle.

[0023] In some embodiments, the environmental information further includes third information, which includes the location information and map information of the first vehicle; the operation scenario is obtained by identifying the third information.

[0024] In some embodiments, the work area is a mining work area; the work scenario includes road slope, road type, and work area type.

[0025] According to a second aspect of the present disclosure, a job control apparatus is provided, including a module configured to perform the job control method described in any of the above embodiments.

[0026] According to a third aspect of the present disclosure, a job control apparatus is provided, comprising: a memory; and a processor coupled to the memory, the processor being configured to execute the job control method described in any of the above embodiments based on instructions stored in the memory.

[0027] According to a fourth aspect of the present disclosure, an unmanned vehicle is provided, including: the operation control device described in any of the above embodiments.

[0028] According to a fifth aspect of the present disclosure, a computer-readable storage medium is provided, including computer program instructions, wherein the computer program instructions, when executed by a processor, implement the job control method described in any of the above embodiments.

[0029] According to a sixth aspect of the present disclosure, a computer program product is provided, including a computer program, wherein when the computer program is executed by a processor, it implements the job control method described in any of the above embodiments.

[0030] In this embodiment, by identifying environmental information of the work area, identification information is obtained, including the current work scenario of the first vehicle and the driving style and intention of the second vehicle. Risk assessment of this identification information accurately determines the risk level of the first vehicle's current situation. Then, based on the risk level of the first vehicle's current situation and the predicted actions of the second vehicle, control commands for controlling the first vehicle are generated. In this approach, by mimicking the thinking of a human driver and engaging in brain-like decision-making and dynamic game theory, the first vehicle can operate according to its own risk situation and the predicted actions of other vehicles, thus improving work efficiency while ensuring safety.

[0031] Other features, aspects, and advantages of this disclosure will become clear from the following detailed description of exemplary embodiments with reference to the accompanying drawings. Attached Figure Description

[0032] To more clearly illustrate the technical solutions in the embodiments of this disclosure 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 this disclosure. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0033] The accompanying drawings form part of this specification, illustrating exemplary embodiments of the present disclosure, and together with the specification serve to explain the principles of the present disclosure.

[0034] This disclosure will become clearer with reference to the accompanying drawings and the following detailed description, in which:

[0035] Figure 1 This is a flowchart illustrating a job control method according to some embodiments of the present disclosure.

[0036] Figure 2 This is a schematic diagram illustrating a model structure for identifying driving style and driving intention according to some embodiments of the present disclosure.

[0037] Figure 3 This is a schematic diagram of a model structure for identifying work scenarios and load conditions according to some embodiments of the present disclosure.

[0038] Figure 4 This is a schematic diagram illustrating the structure of a work control device according to some embodiments of the present disclosure.

[0039] Figure 5 This is a schematic diagram illustrating the structure of a work control device according to other embodiments of the present disclosure. Detailed Implementation

[0040] The technical solutions of the embodiments of this disclosure will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this disclosure, and not all embodiments. Based on the embodiments of this disclosure, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this disclosure.

[0041] Unless otherwise specifically stated, the relative arrangement, numerical expressions, and values ​​of the components and steps set forth in these embodiments do not limit the scope of this disclosure.

[0042] At the same time, it should be understood that, for ease of description, the dimensions of the various parts shown in the accompanying drawings are not drawn according to actual scale.

[0043] Techniques, methods, and equipment known to those skilled in the art may not be discussed in detail, but where appropriate, such techniques, methods, and equipment should be considered part of the specification.

[0044] In all examples shown and discussed herein, any specific values ​​should be interpreted as merely exemplary and not as limitations. Therefore, other examples of exemplary embodiments may have different values.

[0045] It should be noted that similar labels and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be discussed further in subsequent figures.

[0046] Figure 1 This is a flowchart illustrating a job control method according to some embodiments of the present disclosure.

[0047] In step 102, environmental information of the work area where the first vehicle is located is obtained. Here, the first vehicle is an unmanned vehicle. In some embodiments, the work area is a mining work area. In this case, the first vehicle may be an unmanned mining truck.

[0048] In some embodiments, environmental information may include sensing data, such as lidar point clouds, camera images, millimeter-wave radar data, or data obtained by fusing lidar point clouds, camera images, and millimeter-wave radar data.

[0049] In other embodiments, environmental information may include information about traffic participants within the work area, such as location, speed, acceleration, and heading angle. For example, traffic participant information can be obtained via V2X (Vehicle-to-X) technology. Traffic participants may include a first vehicle and other vehicles, which may be unmanned or manned vehicles.

[0050] In some other embodiments, environmental information may include map information of the work area, etc.

[0051] In step 104, environmental information is identified to obtain identification information. Here, the identification information includes the work scenario where the first vehicle is located, and the driving style and driving intention of the second vehicle within the work area.

[0052] It should be noted that the second vehicle can be any vehicle within the work area other than the first vehicle. The second vehicle can be a manned vehicle or an unmanned vehicle. For manned vehicles, driving style and driving intention can be understood as the driver's driving style and driving intention; for unmanned vehicles, driving style and driving intention can be understood as the driving style and driving intention of the unmanned driving system.

[0053] For example, machine learning models can be used to identify environmental information.

[0054] In some embodiments, the work scenario includes road slope, road type, and work area type. Road slope may be, for example, uphill, downhill (e.g., a long downhill), or flat; road type may be, for example, straight or curved (e.g., a sharp curve). Work area type may be, for example, a loading area, an unloading area, or a main road. For example, the current work scenario of the first vehicle can be determined by combining high-precision map information with the location information of the first vehicle (e.g., GPS data).

[0055] In some embodiments, the driving style can be aggressive or conservative. For example, based on LSTM (Long Short-Term Memory) network, the continuous historical trajectory sequence of the second vehicle (e.g., features such as lateral sway and rate of change of acceleration) can be modeled and analyzed to identify the driving style of the second vehicle as aggressive or conservative in real time online.

[0056] In some embodiments, the driving intention may be to turn left, turn right, go straight, or pull over to the side of the road. For example, an LSTM with an attention mechanism can be used to predict the driving intention of the second vehicle based on the dynamic characteristics of the second vehicle.

[0057] In some embodiments, the identification information may also include other information. As some implementations, the identification information may also include the load status of the second vehicle. The load status can be empty or fully loaded. For example, a convolutional neural network (CNN) can be used to analyze visual images of the second vehicle to identify the shape of the second vehicle's load space (e.g., cargo bed) to determine the load status of the second vehicle.

[0058] In some embodiments, multiple pieces of information from the identification information can be fused into a feature vector. This feature vector can reflect the working scenario of the first vehicle, as well as the driving style and driving intention of the second vehicle within the working area. In some embodiments, the feature vector can also reflect the load status of the second vehicle.

[0059] In step 106, a risk assessment is performed based on reference information to determine the current risk level of the first vehicle. Here, the reference information includes the identification information identified in step 104, meaning that the risk assessment is performed at least based on the identification information.

[0060] In some embodiments, the identification information includes not only the work scenario in which the first vehicle is located, and the driving style and driving intention of the second vehicle within the work area, but also the load status of the second vehicle. In certain scenarios (e.g., downhill scenarios), the risk level determined when the second vehicle is fully loaded is higher than the risk level determined when the second vehicle is unloaded. Thus, by combining the load status of the second vehicle, the risk level of the first vehicle's current situation can be assessed more accurately.

[0061] In some embodiments, the reference information may also include other information; that is, in addition to identification information, other information may be combined for risk assessment. As some implementations, the reference information may also include at least one, such as one or both, of the relative speed and relative distance between the first and second vehicles. The relative speed and relative distance between the first and second vehicles can reflect the likelihood of a collision between them. When conducting a risk assessment, by further combining the relative speed and / or relative distance between the first and second vehicles, the risk level of the first vehicle's current situation can be assessed more accurately.

[0062] In some embodiments, the feature vectors described above can be parsed based on a prior risk level knowledge base to quantify the risk level of the first vehicle's current situation. For example, the risk level may include a risk label and a risk score to more accurately reflect the current risk situation of the first vehicle.

[0063] In step 108, the actions that the second vehicle will take are predicted based on the second vehicle's driving style and driving intentions.

[0064] One approach is to use driving style and driving intention as prior information to predict the actions that the second vehicle will take.

[0065] The actions taken by the second vehicle will vary depending on the driving style and intention. For example, if the second vehicle is driven aggressively and its intention is to turn left, the predicted action might be to accelerate and turn left. Conversely, if the second vehicle is driven conservatively and its intention is to turn left, the predicted action might be to decelerate and turn left.

[0066] In step 110, control commands for controlling the first vehicle are generated based on the risk level of the first vehicle's current risk and the predicted actions that the second vehicle will take.

[0067] In some embodiments, a control strategy for controlling the first vehicle is first determined based on the risk level of the current situation of the first vehicle and the predicted actions that the second vehicle will take; then, control commands for controlling the first vehicle are generated based on the risk level of the current situation of the first vehicle and the control strategy. Here, the control strategy includes acceleration, deceleration, maintaining the current vehicle speed, or steering. In some embodiments, the control strategy may also include cooperative action.

[0068] As one implementation method, dynamic game theory decision-making can be carried out based on the risk level of the first vehicle's current situation and the predicted actions that the second vehicle will take, in order to determine the control strategy for controlling the first vehicle.

[0069] For example, when the control strategy determined by the game is to slow down and the risk level is extremely high, the generated control instruction is a proactive yielding instruction to reduce safety risks; when the determined control strategy is to maintain and the risk level is low, the generated control instruction is to maintain the original driving plan to improve work efficiency; when the determined control strategy is to cooperate, the generated control instruction is a cooperative passage instruction (e.g., alternating passage at intersections) to achieve efficient passage through cooperation and improve work efficiency while ensuring safety.

[0070] Generating control commands is equivalent to transforming the output of an abstract, game-theoretic control strategy into concrete, executable high-level instructions. For example, based on preset rules, a control strategy can be mapped to a high-level behavioral instruction rich in interactive intent—that is, a control command. This transformation process can be implemented using behavior trees, for instance. Behavior trees can define transition rules between different driving behavior states, ensuring the rationality and interpretability of the output control commands, ultimately realizing the transformation from numerical calculation to intelligent behavior.

[0071] In the above embodiments, by identifying environmental information of the work area, identification information is obtained, including the current work scenario of the first vehicle and the driving style and intention of the second vehicle. Risk assessment of this identification information accurately determines the risk level of the first vehicle's current situation. Then, based on the risk level of the first vehicle's current situation and the predicted actions of the second vehicle, control commands for controlling the first vehicle are generated. In this approach, by mimicking the thinking of a human driver and engaging in brain-like decision-making and dynamic game theory, the first vehicle can perform subsequent operations based on its own risk situation and the predicted actions of other vehicles, thereby improving work efficiency while ensuring safety.

[0072] In some embodiments, the second vehicle is a manned vehicle. Thus, by identifying the characteristics of traffic participants in mixed-traffic scenarios involving both manned and unmanned vehicles, and making intelligent dynamic game-theoretic decisions based on these characteristics, the coordination problem in such scenarios can be solved, improving safety and operational efficiency.

[0073] In some embodiments, the motion trajectory of the first vehicle is planned according to the risk level and control strategy of the current risk situation of the first vehicle; the motion trajectory is converted into control quantity instructions, and the control quantity instructions are sent to the actuators of the first vehicle. Here, the control quantities include throttle control quantities, brake control quantities, and steering control quantities.

[0074] For example, a Model Predictive Control (MPC) algorithm is used to receive the control strategy. Taking the current vehicle state as initial conditions, and comprehensively considering dynamic constraints, road boundaries, and the positions of other vehicles, a smooth and comfortable motion trajectory that meets the control strategy's requirements is generated through rolling optimization. For instance, the MPC controller can use this motion trajectory to convert it into specific throttle, brake, and steering control values ​​in real time, and then send them to the actuators of the first vehicle. The actuators receive the commands, ultimately achieving safe, efficient, and human-like driving behavior.

[0075] In some embodiments, the environmental information obtained in step 102 includes one or more of first information, second information, and third information. For example, driving style is obtained by identifying the first information, driving intention is obtained by identifying the second information, and work scenario is obtained by identifying the third information.

[0076] The following examples illustrate how to identify the first, second, and third information to obtain driving style, driving intention, and work scenario.

[0077] First, let's introduce the identification of primary information and driving style.

[0078] As one implementation method, the first information includes the second vehicle's speed, longitudinal acceleration, lateral acceleration, and lateral deviation from the desired lane. Thus, the driving style of the second vehicle can be accurately identified based on the first information.

[0079] As one implementation method, the first information may also include the distance between the second vehicle and the vehicle in front. In this way, the driving style of the second vehicle can be more accurately identified based on the first information.

[0080] Figure 2 This is a schematic diagram illustrating a model structure for identifying driving style and driving intention according to some embodiments of the present disclosure.

[0081] like Figure 2 As shown, the model includes a bidirectional LSTM and a feedforward neural network.

[0082] For example, the first information within the first time window can be obtained. The trajectory dataset of the second vehicle can be represented as X = {X1,X2,...,Xt,...,Xn}, where X1,X2,...,Xt,...,Xn represent the first information at different times.

[0083] For a bidirectional LSTM, at time t, the forward LSTM can obtain information up to Xt. , ,…, Backward LSTM can obtain information after time t. , ,…, The two hidden layer output vectors and Combining the information can yield the entire sequence information of the bidirectional LSTM at time t.

[0084] For a feedforward neural network, the input at time t is the forward output of the bidirectional LSTM at time t. and the reverse output at the same time The feature vectors are composed of these features. For example, ReLU (Linear rectification function) can be used as the activation function of the hidden layer, and Softmax (normalization exponent) function can be used as the activation function of the output layer.

[0085] The two neurons in the output layer represent the probability values ​​of "aggressive" and "conservative" respectively, and the sum of all output probabilities is 1. For example, [0.85, 0.15] means that the model judges the second vehicle to have an 85% probability of being aggressive and a 15% probability of being conservative.

[0086] Next, we will introduce the recognition of second information and driving intention.

[0087] As one implementation method, the second information may include the trajectory curvature, lateral acceleration, and lateral velocity of the second vehicle. Thus, the driving intention of the second vehicle can be accurately identified based on this second information.

[0088] In other implementations, the second information may also include the turn signal status and wheel angle of the second vehicle. Thus, the driving intention of the second vehicle can be more accurately identified based on this second information.

[0089] In some embodiments, a similar approach can be adopted. Figure 2 The model shown is used to identify driving intentions. For example, in Figure 2Following the bidirectional LSTM shown, an attention layer is added. This layer automatically learns and calculates the weights for each time step, using tanh and softmax as activation functions. Additionally, a fully connected layer can be placed before the output layer in the feedforward neural network following the attention layer. This layer performs a non-linear transformation on the context vector generated by the attention layer, ultimately mapping it to the driving intention classification. In this case, the output of the output layer represents the probability values ​​of intentions such as "left turn," "right turn," "straight ahead," and "lane keeping."

[0090] Similarly, second information can be obtained within a second time window. This second time window might be, for example, two seconds prior to the current moment. In some embodiments, the length of the second time window is shorter than the length of the first time window. Considering that the prediction of driving intention depends on the latest dynamic changes and focuses more on recent key actions, shortening the time window and introducing an attention mechanism can more accurately identify the driving intention of a second vehicle.

[0091] Next, we will introduce the identification of third information and work scenarios.

[0092] As one implementation method, the third information includes the location information of the first vehicle and map information. The map information is, for example, high-precision map information.

[0093] Figure 3 This is a schematic diagram of a model structure for identifying work scenarios and load conditions according to some embodiments of the present disclosure.

[0094] like Figure 3 As shown, the model includes a convolutional neural network with two convolutional layers (convolutional layer 1 and convolutional layer 2), two pooling layers (pooling layer 1 and pooling layer 2), and one fully connected layer. The convolutional kernel size of the convolutional layers is, for example, 5x5, and the stride is, for example, 1. The pooling window size of the pooling layers is, for example, 2x2. The fully connected layer outputs a feature vector, which is then used by an SVM (Support Vector Machine) for classification.

[0095] Input data can include visual features, localization and mapping features, obstacle features, etc. A convolutional neural network is used as the backbone network for feature extraction to extract features such as the type of the second vehicle and the shape of the cargo bed from the camera to determine the load status of the second vehicle.

[0096] The location and map information can be used to determine the scene in which the first vehicle is located, i.e., the operational scene. The high-precision map is used to determine whether the current road segment of the first vehicle is uphill, downhill, or flat; and whether the first vehicle is in a specific scene such as a loading area, unloading area, main road, or sharp bend. In some embodiments, the pitch and roll angles of the first vehicle can also be obtained to help determine the load status of the first vehicle and the road inclination, thereby helping to determine the risk level of the first vehicle. For example, the pitch and roll angles of the first vehicle can be obtained through an IMU (Inertial Measurement Unit).

[0097] The model fuses and classifies these features, ultimately producing scene classification output and load classification output. Scene classification output may include, for example, uphill roads, downhill roads, curves, loading areas, unloading areas, sharp turns, and straight roads. Load classification output may include fully loaded or empty.

[0098] The following section describes how to determine the implementation method of the control strategy for controlling the first vehicle based on the risk level of the first vehicle's current situation and the predicted actions that the second vehicle will take.

[0099] In some embodiments, the control strategy is determined by constructing a payoff function. Specifically, the candidate strategy from the strategy combination with the highest payoff function value among multiple strategy combinations can be determined as the control strategy. Here, each strategy combination consists of one candidate strategy from the candidate strategy set of the first vehicle and the predicted action to be taken by the second vehicle. It is understood that the predicted action to be taken by the second vehicle is equivalent to the strategy adopted by the second vehicle, and the candidate strategy set of the first vehicle includes, for example, acceleration, deceleration, maintaining the current speed, or steering. In some embodiments, the candidate strategy set may also include cooperative action.

[0100] The payoff function comprehensively quantifies safety, efficiency, and comfort benefits, with a higher weight given to safety benefits as the risk level of the first vehicle increases. Different strategy combinations correspond to different payoff function values; the payoff function value is maximized when the first vehicle adopts a particular candidate strategy, resulting in the highest overall benefit. In this case, the candidate strategy adopted by the first vehicle can be determined as the control strategy.

[0101] Understandably, in the above approach, the traffic scenario can be formalized as a game theory model when making dynamic game decisions. In this model, the first and second vehicles are considered as players in the game, each possessing a set of potential strategies. Each possible strategy combination corresponds to a value of a payoff function, which comprehensively quantifies the safety, efficiency, and comfort benefits resulting from the strategy's execution. The core input for safety payoff comes from the risk level; in high-risk scenarios, the weight of safety payoff in the total payoff is significantly amplified, almost dominating the final decision direction. The decision-making algorithm not only relies on calculating the payoff of the first vehicle but also infers the most likely strategy adopted by the second vehicle. Finally, by solving the Nash equilibrium game model, the optimal strategy that maximizes the overall expected payoff of the first strategy, given the prediction of the second vehicle's actions, can be calculated.

[0102] The following section introduces some methods for determining risk levels.

[0103] In some embodiments, risk assessment based on reference information to obtain the risk level of the current situation of the first vehicle includes: determining a matching risk level from a risk level knowledge base that matches each piece of information in the reference information, the risk level knowledge base being pre-built based on historical accidents; and determining the determined matching risk level as the risk level of the current situation of the first vehicle.

[0104] For example, a risk level knowledge base can be established based on historical accident data, expert experience, and the characteristics of the work area, encompassing various scenario risk levels. This risk level knowledge base can include predefined "IF-THEN" rules.

[0105] For example, if (other vehicle's load is full and the work scenario is downhill and the other vehicle's driving intention is not deceleration), then the risk level is extremely high (weight = 0.95). Another example is if (other vehicle's driving style is aggressive and the other vehicle's driving intention is cutting in and the relative distance is less than the safe distance), then the risk level is high (weight = 0.75). Yet another example is if (work scenario is loading area and the other vehicle's driving intention is reversing), then the risk level is medium (weight = 0.5). The risk level can include a risk score (i.e., weight) and a risk label (i.e., high, medium, low). It should be understood that "other vehicle" in the above rules refers to the second vehicle.

[0106] It should be noted that the rules in the three examples above use different types of information to define different risk levels. In some embodiments, the risk level can be defined based on the type of each piece of information in the reference information. For example, if the reference information includes three pieces of information, then the risk level is defined by all three pieces of information.

[0107] Scenario-based risk assessment matches the current work scenario with predefined risk scenarios. The risk level knowledge base contains typical risk scenarios and their corresponding risk levels for various work scenarios. Through pattern recognition and matching algorithms, the risk type of the current work scenario can be quickly identified. By matching various information from the reference information, the final risk level reflects the overall risk level of the current work scenario.

[0108] In some embodiments, the risk level knowledge base employs a rule-based expert system that uses fuzzy logic to handle fuzzy concepts such as overly aggressive or somewhat close concepts, making risk assessment more akin to human thought. In some embodiments, the risk assessment algorithm employs a Bayesian network to handle uncertainties and dependencies between features.

[0109] In some embodiments, environmental changes in the work area can be continuously monitored, and when changes in key parameters are detected, such as sudden acceleration of other vehicles or the activation of turn signals, the process can be repeated. Figure 1 The method shown allows for a reassessment of risk levels, ensuring that control instructions are based on the latest risk situation.

[0110] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus embodiments, since they largely correspond to the method embodiments, the description is relatively simple; relevant parts can be referred to the descriptions of the method embodiments.

[0111] This disclosure also provides a job control device, including a module configured to execute the job control method of any of the above embodiments.

[0112] Figure 4 This is a schematic diagram illustrating the structure of a work control device according to some embodiments of the present disclosure.

[0113] like Figure 4 As shown, the operation control device includes: an acquisition module 401, an identification module 402, an evaluation module 403, a prediction module 404, and a generation module 405.

[0114] The acquisition module 401 is configured to acquire environmental information of the work area where the first vehicle is located. The first vehicle is an unmanned vehicle.

[0115] The identification module 402 is configured to identify environmental information to obtain identification information, including the work scene where the first vehicle is located, and the driving style and driving intention of the second vehicle in the work area.

[0116] The assessment module 403 is configured to perform a risk assessment based on reference information to obtain the risk level of the first vehicle's current risk, including identification information.

[0117] The prediction module 404 is configured to predict the actions that the second vehicle will take based on the second vehicle's driving style and driving intentions.

[0118] The generation module 405 is configured to generate control commands for controlling the first vehicle based on the risk level of the first vehicle's current risk and the predicted actions that the second vehicle will take.

[0119] In some embodiments, the job control device may further include other modules to execute the job control method of any of the above embodiments.

[0120] Figure 5 This is a schematic diagram illustrating the structure of a work control device according to other embodiments of the present disclosure.

[0121] like Figure 5 As shown, the job control device 500 includes a memory 501 and a processor 502 coupled to the memory 501. The processor 502 is configured to execute the job control method of any of the above embodiments based on instructions stored in the memory 501.

[0122] The memory 501 may include, for example, system memory, fixed non-volatile storage media, etc. The system memory may store, for example, an operating system, application programs, a boot loader, and other programs.

[0123] In some embodiments, the job control device 500 may further include an input / output interface 503, a network interface 504, a storage interface 505, etc. These interfaces 503, 504, and 505, as well as the memory 501 and processor 502, can be connected, for example, via a bus 506. The input / output interface 503 provides a connection interface for input / output devices such as a monitor, mouse, keyboard, and touchscreen. The network interface 504 provides a connection interface for various networked devices. The storage interface 505 provides a connection interface for external storage devices such as SD cards and USB flash drives.

[0124] According to some other embodiments of this disclosure, the job control device may include a feature recognition and perception layer, a brain-like decision-making and game-playing layer, and a planning and control layer. Figure 1 Steps 102 and 104 shown can be performed by the feature recognition and perception layer. Figure 1 Steps 106 and 108 shown can be executed by the brain-like decision-making and game-playing layer. Figure 1 Step 110 shown can be performed by the planning and control layer.

[0125] As some implementations, the operation control device can execute the operation control method of any of the above embodiments separately for different first vehicles to control the operation of the first vehicle separately. For example, the operation control device can be set up in the cloud. In this way, unmanned vehicles in the operation area can be globally controlled using the same operation control device.

[0126] In some other implementations, the work control device can be mounted on the first vehicle. In this way, each first vehicle can execute the work control method of any of the above embodiments to control its own work. This makes the work control of the first vehicle more timely, further improving safety and work efficiency.

[0127] This disclosure also provides an unmanned vehicle, including: the operation control device of any of the above embodiments.

[0128] One or more embodiments disclosed herein can have the following effects:

[0129] 1. By identifying asymmetric risks, it can avoid potential major accidents that traditional collision algorithms cannot identify, thus shifting the accident prevention line from avoiding collisions to eliminating risks.

[0130] 2. Enhance the safety of mixed traffic, enabling unmanned vehicles to interact safely and smoothly with manned vehicles, reducing human-machine interaction conflicts and emergencies caused by the rigid behavior of unmanned vehicles, and providing key safety guarantees for the transition period of mixed traffic in mines.

[0131] 3. Collaborative game theory decision-making avoids frequent and conservative emergency stops, ensuring the continuity and smoothness of the transportation process and improving the overall operating efficiency of unmanned vehicles.

[0132] 4. The decision-making logic of autonomous vehicles mimics that of experienced human drivers, making their behavior easier for other human drivers to understand and predict, and thus easier to accept and trust.

[0133] This disclosure also provides a computer-readable storage medium including computer program instructions that, when executed by a processor, implement the job control method of any of the above embodiments.

[0134] This disclosure also provides a computer program product, including a computer program that, when executed by a processor, implements the job control method of any of the above embodiments.

[0135] The embodiments of this disclosure have now been described in detail. To avoid obscuring the concept of this disclosure, some details known in the art have not been described. Those skilled in the art can fully understand how to implement the technical solutions disclosed herein based on the above description.

[0136] Those skilled in the art will understand that embodiments of this disclosure can be provided as methods, systems, or computer program products. Therefore, this disclosure can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this disclosure can take the form of a computer program product embodied on one or more computer-usable non-transitory storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0137] This disclosure is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this disclosure. It should be understood that the functions specified in one or more flowchart illustrations and / or one or more block diagrams can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate functions for implementing the functions in the flowchart illustrations. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0138] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0139] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0140] While specific embodiments of this disclosure have been described in detail by way of examples, those skilled in the art should understand that the examples are for illustrative purposes only and not intended to limit the scope of this disclosure. Those skilled in the art should understand that modifications can be made to the above embodiments or equivalent substitutions can be made to some technical features without departing from the scope and spirit of this disclosure. The scope of this disclosure is defined by the appended claims.

Claims

1. A job control method, comprising: Obtain environmental information about the work area where the first vehicle is located; the first vehicle is an unmanned vehicle. The environmental information is identified to obtain identification information, which includes the work scenario where the first vehicle is located, and the driving style and driving intention of the second vehicle in the work area. A risk assessment is performed based on reference information to determine the risk level of the first vehicle's current situation; the reference information includes the identification information. Based on the driving style and driving intentions of the second vehicle, predict the actions that the second vehicle will take; Based on the risk level of the first vehicle's current situation and the predicted actions that the second vehicle will take, control commands are generated to control the first vehicle.

2. The operation control method according to claim 1, wherein, Based on the risk level of the first vehicle's current situation and the predicted actions the second vehicle will take, control commands for controlling the first vehicle are generated, including: Based on the risk level of the current risk of the first vehicle and the predicted actions that the second vehicle will take, a control strategy for controlling the first vehicle is determined, the control strategy including acceleration, deceleration, maintaining the current speed or steering. Based on the risk level of the first vehicle and the control strategy, control commands for controlling the first vehicle are generated.

3. The operation control method according to claim 2, wherein, Based on the risk level of the first vehicle's current situation and the predicted actions the second vehicle will take, the control strategy for controlling the first vehicle is determined as follows: The control strategy is determined by selecting the strategy combination that yields the highest return function among multiple strategy combinations. Each strategy combination consists of one candidate strategy from the candidate strategy set of the first vehicle and the predicted action to be taken by the second vehicle. The benefit function comprehensively quantifies safety benefits, efficiency benefits, and comfort benefits, and the higher the risk level of the first vehicle, the greater the weight of the safety benefits.

4. The operation control method according to claim 2, wherein, Based on the risk level of the first vehicle's current situation and the control strategy, control commands for controlling the first vehicle are generated, including: Based on the risk level of the first vehicle's current situation and the control strategy, plan the trajectory of the first vehicle; The motion trajectory is converted into a control quantity instruction, and the control quantity instruction is sent to the actuator of the first vehicle. The control quantity includes throttle control quantity, brake control quantity, and steering control quantity.

5. The operation control method according to claim 1, wherein, The identification information also includes the load status of the second vehicle.

6. The operation control method according to claim 1, wherein, The reference information also includes at least one of the relative speed and relative distance between the first vehicle and the second vehicle.

7. The operation control method according to any one of claims 1-6, wherein, A risk assessment is conducted based on reference information to determine the current risk level of the first vehicle, including: The risk level is determined from the risk level knowledge base that matches each item in the reference information, the risk level knowledge base being pre-built based on historical accidents; The matched risk level is determined as the risk level of the current situation of the first vehicle.

8. The operation control method according to any one of claims 1-6, wherein: The environmental information includes first information, which includes the speed, longitudinal acceleration, lateral acceleration, and lateral deviation from the desired lane of the second vehicle. The driving style is obtained by identifying the first piece of information.

9. The operation control method according to claim 8, wherein, The first information also includes the distance between the second vehicle and the vehicle in front.

10. The operation control method according to any one of claims 1-6, wherein: The environmental information includes second information, which includes the trajectory curvature, lateral acceleration, and lateral velocity of the second vehicle. The driving intention is obtained by recognizing the second information.

11. The operation control method according to claim 10, wherein, The second information also includes the turn signal status and wheel angle of the second vehicle.

12. The operation control method according to any one of claims 1-6, wherein: The environmental information also includes third information, which includes the location information and map information of the first vehicle; The work scenario is obtained by identifying the third information.

13. The operation control method according to any one of claims 1-6, wherein: The work area is a mining work area; The operational scenarios include road slope, road type, and operational area type.

14. A work control device, comprising: A module configured to perform the job control method according to any one of claims 1-13.

15. A work control device, comprising: Memory; as well as A processor coupled to the memory is configured to execute the job control method of any one of claims 1-13 based on instructions stored in the memory.

16. An unmanned vehicle, comprising: The operation control device as described in claim 14 or 15.

17. A computer-readable storage medium comprising computer program instructions, wherein, When the computer program instructions are executed by the processor, they implement the job control method according to any one of claims 1-13.

18. A computer program product comprising a computer program, wherein, When the computer program is executed by the processor, it implements the job control method according to any one of claims 1-13.

Images