Speed control method and device of unmanned vehicle and unmanned vehicle

Abstract

The invention relates to the technical field of unmanned driving, automatic driving and unmanned vehicles, and discloses a speed control method and device of an unmanned vehicle and the unmanned vehicle. The method comprises the following steps: acquiring current position information of the unmanned vehicle, and acquiring a corresponding theoretical speed limit value according to the current position information; calculating an empirical speed value corresponding to the current position of the unmanned vehicle according to a predetermined empirical speed function; wherein the empirical speed function is determined based on driving data when the manned vehicle is driven to different positions; and generating a speed control strategy according to the theoretical speed limit value and the experience speed value, and controlling the driving speed of the unmanned vehicle based on the speed control strategy. According to the invention, by fusing static map speed limiting and driving experience data, more intelligent and efficient unmanned mine car speed control is realized.

Description

Technical Field

[0001] The present application relates to the fields of unmanned driving, automatic driving, and unmanned vehicles, and in particular to a speed control method and device for an unmanned vehicle and an unmanned vehicle. Background Art

[0002] At present, the speed control of unmanned driving in mines is usually based on the speed limit input of fixed sections of the map layer and the speed limit source input of the platform, which is input into the global planning, and then input into the decision-making planning module. In this strong rule-based speed control, the speed limit value mainly comes from the static lane-level speed limit layer stored in the map. In addition, when manned vehicles equipped with data acquisition systems are driving in mines, experienced mining truck drivers will drive more efficiently and fuel-efficiently than unmanned vehicles when performing mining, transportation, loading and unloading tasks. This experience data is stored as expert experience knowledge.

[0003] However, the above-mentioned speed control scheme based on strong rules has the following defects: when the upstream speed limit changes suddenly, the system will use emergency braking to achieve emergency deceleration, which will have a negative impact on safety and energy saving; when encountering complex working conditions (such as uphill and downhill, curves, potholes, different load conditions, etc.) and bad weather, the system will not adaptively adopt a suitable speed limit control strategy. This strong rule-based speed planning control cannot cope with the complex and changeable actual environment, which restricts the operating efficiency of unmanned mining vehicles. Summary of the invention

[0004] The present application provides a speed control method and device for an unmanned vehicle and an unmanned vehicle, which are used to achieve more intelligent and efficient speed control of an unmanned mining vehicle by integrating static map speed limits and driving experience data.

[0005] In a first aspect, the present application provides a speed control method for an unmanned vehicle, the speed control method for an unmanned vehicle comprising: obtaining current position information of the unmanned vehicle, and obtaining a corresponding theoretical speed limit value based on the current position information; calculating an empirical speed value corresponding to the current position of the unmanned vehicle based on a predetermined empirical speed function; wherein the empirical speed function is determined based on driving data of a manned vehicle when it travels to different positions; generating a speed control strategy based on the theoretical speed limit value and the empirical speed value, and controlling the driving speed of the unmanned vehicle based on the speed control strategy.

[0006] In a second aspect, the present application provides a speed control device for an unmanned vehicle, the speed control device for an unmanned vehicle comprising:

[0007] An acquisition module is used to obtain the current position information of the unmanned vehicle and obtain the corresponding theoretical speed limit value according to the current position information;

[0008] A calculation module, used to calculate an empirical speed value corresponding to the current position of the unmanned vehicle according to a predetermined empirical speed function; wherein the empirical speed function is determined based on driving data of a manned vehicle when driving to different positions;

[0009] The generation module is used to generate a speed control strategy according to a theoretical speed limit value and an empirical speed value, and control the driving speed of the unmanned vehicle based on the speed control strategy.

[0010] A third aspect of the present application provides an unmanned vehicle, comprising: a speed control device for an unmanned vehicle as described in the second aspect.

[0011] In the technical solution provided by the present application, by obtaining the current position information in real time and dynamically calculating the theoretical speed limit value, it is ensured that the driving speed of the unmanned vehicle is always within the safe range, and the risk of speeding is effectively prevented. Secondly, the introduction of the empirical speed function makes full use of the actual driving experience accumulated by manned vehicles, so that the speed control of the unmanned vehicle is closer to the actual driving habits and improves the transportation efficiency. Thirdly, the theoretical speed limit value and the empirical speed value are combined to generate the speed control strategy, which not only ensures driving safety but also takes into account the operating efficiency, so that the unmanned vehicle can make reasonable speed adjustments under different road conditions. In particular, in complex road conditions such as uphill and downhill, bends, etc., the method can give more appropriate speed control suggestions based on empirical data, avoiding the operational stiffness that may be caused by relying solely on theoretical speed limit. At the same time, the method realizes the transition from static speed limit to dynamic control, ensures the stability of vehicle operation through continuous speed adjustment, reduces unnecessary rapid acceleration and rapid deceleration, and realizes more intelligent and efficient unmanned mine car speed control. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0013] Figure 1 A schematic diagram of an embodiment of a speed control method for an unmanned vehicle in an embodiment of the present application;

[0014] Figure 2 A schematic diagram of a flow chart of speed control of an unmanned vehicle in an embodiment of the present application;

[0015] Figure 3 A schematic diagram of neural network training in an embodiment of the present application;

[0016] Figure 4 Schematic diagram of an embodiment of a speed control device for an unmanned vehicle in an embodiment of the present application. DETAILED DESCRIPTION

[0017] Embodiments of the present application provide a speed control method, device and unmanned vehicle for an unmanned vehicle. The terms "first", "second", "third", "fourth", etc. (if any) in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments described here can be implemented in an order other than that illustrated or described here. In addition, the terms "including" or "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0018] For ease of understanding, the specific process of the embodiment of the present application is described below. Figure 1 , an embodiment of the speed control method of the unmanned vehicle in the embodiment of the present application includes:

[0019] Step S101, obtaining the current position information of the unmanned vehicle, and obtaining the corresponding theoretical speed limit value according to the current position information;

[0020] Step S102, calculating an empirical speed value corresponding to the current position of the unmanned vehicle according to a predetermined empirical speed function; wherein the empirical speed function is determined based on driving data of a manned vehicle when driving to different positions;

[0021] Step S103: Generate a speed control strategy according to the theoretical speed limit value and the empirical speed value, and control the driving speed of the unmanned vehicle based on the speed control strategy.

[0022] It is understandable that the execution subject of the present application may be a speed control device for an unmanned vehicle, or a terminal or a server, which is not specifically limited here. The present application embodiment is described by taking a server as the execution subject as an example.

[0023] Optionally, after obtaining the current position information of the unmanned vehicle, the theoretical speed limit value is first calculated. The theoretical speed limit value calculation includes two parts: first, the static map speed limit value is matched from the vehicle-side static lane-level speed limit layer according to the current position, and then the dynamic speed limit adjustment value is matched from the cloud-based dynamic speed limit layer according to the real-time scene information of the current position (including load status, road conditions, environmental information, etc.). The specific calculation of the dynamic speed limit adjustment value is achieved by accumulating the speed limit adjustment amounts corresponding to each scenario factor, including: vehicle load status (no load 0km / h, heavy load -2km / h), road slipperiness (dry 0km / h, slightly slippery -2km / h, severely slippery -4km / h), road pothole degree (flat 0km / h, slightly pothole-2km / h, severely pothole-4km / h), traffic congestion degree (no congestion 0km / h, slightly congested -2km / h, severely congested -4km / h), weather conditions (sunny 0km / h, light rain -2km / h, heavy rain -4km / h; no snow 0km / h, light snow -2km / h, heavy snow -4km / h), lighting conditions (daytime 0km / h, night -2km / h).

[0024] For the calculation of the empirical speed function, a large amount of driving data of manned vehicles equipped with a data acquisition system when driving in the mine is collected, including the position and speed information in the vehicle operation data. The collected position information is used as input, and the corresponding speed information is used as a label. The convolutional neural network is used for training to obtain the empirical speed prediction model based on position. The final speed control strategy is calculated by the following formula:

[0025] V(s) = A·(V map speed limit(s)+V dynamic speed limit(s))+B·V experience speed(s)

[0026] Where A and B are the normalized coefficients of the theoretical speed limit and the empirical speed limit respectively, V(s) is the speed position function, s is the position parameter, Vmap speed limit (s) is the speed limit value corresponding to position s in the static map layer, Vdynamic speed limit (s) is the dynamic speed limit adjustment value based on the real-time scene, and Vexperiential speed (s) is the predicted value of the empirical speed function at position s.

[0027] Among them, see Figure 2In this application, the map information stored on the vehicle side is first called to obtain the lane speed limit information. The dynamic speed limit adjustment information stored in the cloud is called. This information is configured based on the current working conditions and road conditions, and will be updated and sent to the vehicle side only when changes occur. Integrate the map lane speed limit information and the dynamic speed limit adjustment information to obtain the theoretical speed limit value on the vehicle side based on the vehicle's position. The vehicle collects a large amount of vehicle operation data when it is driven by someone (experienced driver driving data) under different working conditions and road conditions. Then a large amount of driver experience driving data is input, and feature extraction and training are performed with the help of a deep learning model to output an experience speed function based on the road position. Based on the map lane speed limit, dynamic speed limit adjustment and the driver's experience speed function, a more efficient and energy-saving unmanned driving strategy is obtained. According to the specific requirements and actual conditions of the mine, turn on or off the experience speed setting / adjustment of related parameter configurations.

[0028] In the embodiment of the present application, by obtaining the current position information in real time and dynamically calculating the theoretical speed limit value, it is ensured that the driving speed of the unmanned vehicle is always within the safe range, and the risk of speeding is effectively prevented. Secondly, the introduction of the empirical speed function makes full use of the actual driving experience accumulated by manned vehicles, so that the speed control of the unmanned vehicle is closer to the actual driving habits and improves the transportation efficiency. Thirdly, the theoretical speed limit value and the empirical speed value are combined to generate the speed control strategy, which not only ensures driving safety but also takes into account the operating efficiency, so that the unmanned vehicle can make reasonable speed adjustments under different road conditions. In particular, in complex road conditions such as uphill and downhill, bends, etc., the method can give more appropriate speed control suggestions based on empirical data, avoiding the operational stiffness that may be caused by relying solely on theoretical speed limits. At the same time, the method realizes the transition from static speed limit to dynamic control, ensures the stability of vehicle operation through continuous speed adjustment, reduces unnecessary rapid acceleration and rapid deceleration, and realizes more intelligent and efficient unmanned mine car speed control.

[0029] In an optional embodiment, the process of executing step S101 may specifically include the following steps:

[0030] Get the corresponding theoretical speed limit value based on the current location information, including:

[0031] (2) According to the current location information, obtain the static map speed limit value corresponding to the current location;

[0032] (3) determining a real-time dynamic speed limit adjustment value based on real-time scene information of the current location; and

[0033] (4) The static map speed limit value is adjusted using the dynamic speed limit adjustment value to obtain the theoretical speed limit value.

[0034] Optionally, when the unmanned vehicle travels to a certain location, it first needs to obtain the static speed limit value corresponding to the location from the static lane-level speed limit layer on the vehicle side. Then, the real-time scene information of the current location is collected through the on-board sensor, including three types of data: load status, road conditions, and environmental information. For the judgment of the load status, the vehicle load data is collected in real time through the load sensor installed on the vehicle, and the actual load value collected is compared with the rated load of the vehicle. For example, for a mine car with a rated load of 50 tons, when the measured load exceeds 80% of the rated load (ie 40 tons), it is judged to be in a heavy load state, and when it is less than 20% (ie 10 tons), it is judged to be an unloaded state.

[0035] The road condition information is collected by the road condition sensor, which determines the road condition by detecting the friction coefficient and flatness of the road. For example, a friction coefficient of more than 0.7 indicates dry road surface, 0.4-0.7 indicates slightly slippery road surface, and less than 0.4 indicates severely slippery road surface. The road surface flatness index is calculated by the road height data scanned by the laser radar. An index greater than 95 indicates a flat road surface, 85-95 indicates slightly pothole-prone road surface, and less than 85 indicates severely pothole-prone road surface. The collection of environmental information involves multiple sensors: the traffic flow density is detected by the front millimeter-wave radar to determine the degree of congestion. A density less than 20% indicates no congestion, 20%-60% indicates light congestion, and greater than 60% indicates severe congestion; the weather conditions are determined by rain sensors and visibility sensors. A visibility greater than 500 meters and no precipitation indicates a sunny day, a visibility of 200-500 meters or a precipitation less than 2mm / h indicates light rain / light snow, and a visibility less than 200 meters or a precipitation greater than 2mm / h indicates heavy rain / heavy snow; the lighting conditions are determined by a light sensor. A light intensity greater than 1000lux indicates daytime, and less than 1000lux indicates night.

[0036] For example: when an unmanned mining car is driving, the load sensor detects that the current load is 45 tons, corresponding to the heavy load state; the road surface state sensor detects that the friction coefficient is 0.5, corresponding to slightly slippery; the rain sensor detects that the current precipitation is 1.5mm / h, and the visibility sensor detects that the visibility is 350 meters, corresponding to light rain weather. Based on the preset dynamic speed limit adjustment rules, the corresponding speed limit adjustment values ​​are -2km / h, -2km / h and -2km / h, respectively, and the total dynamic speed limit adjustment value is -6km / h after accumulation. If the static speed limit value at this location is 35km / h, the final theoretical speed limit value is 29km / h.

[0037] In an optional embodiment, obtaining a static map speed limit value corresponding to the current location according to the current location information includes:

[0038] (1) According to the current location information, obtain the static lane-level speed limit layer corresponding to the current location;

[0039] (2) Based on the current location information, the static map speed limit value is matched from the static lane-level speed limit layer;

[0040] Based on the real-time scene information of the current location, determine the real-time dynamic speed limit adjustment value, including:

[0041] (1) Based on the current location information, obtain the corresponding dynamic speed limit layer;

[0042] (2) According to the real-time scene information, the real-time dynamic speed limit adjustment value is obtained from the dynamic speed limit layer.

[0043] Optionally, when the unmanned vehicle travels to a certain location, it first needs to obtain the static speed limit value corresponding to the location from the static lane-level speed limit layer on the vehicle side. Then, the real-time scene information of the current location is collected through the on-board sensor, including three types of data: load status, road conditions, and environmental information. For the judgment of the load status, the vehicle load data is collected in real time through the load sensor installed on the vehicle, and the actual load value collected is compared with the rated load of the vehicle. For example, for a mine car with a rated load of 50 tons, when the measured load exceeds 80% of the rated load (ie 40 tons), it is judged to be in a heavy load state, and when it is less than 20% (ie 10 tons), it is judged to be an unloaded state.

[0044] The road condition information is collected by the road condition sensor, which determines the road condition by detecting the friction coefficient and flatness of the road. For example, a friction coefficient of more than 0.7 indicates dry road surface, 0.4-0.7 indicates slightly slippery road surface, and less than 0.4 indicates severely slippery road surface. The road surface flatness index is calculated by the road height data scanned by the laser radar. An index greater than 95 indicates a flat road surface, 85-95 indicates slightly pothole-prone road surface, and less than 85 indicates severely pothole-prone road surface. The collection of environmental information involves multiple sensors: the traffic flow density is detected by the front millimeter-wave radar to determine the degree of congestion. A density less than 20% indicates no congestion, 20%-60% indicates light congestion, and greater than 60% indicates severe congestion; the weather conditions are determined by rain sensors and visibility sensors. A visibility greater than 500 meters and no precipitation indicates a sunny day, a visibility of 200-500 meters or a precipitation less than 2mm / h indicates light rain / light snow, and a visibility less than 200 meters or a precipitation greater than 2mm / h indicates heavy rain / heavy snow; the lighting conditions are determined by a light sensor. A light intensity greater than 1000lux indicates daytime, and less than 1000lux indicates night.

[0045] For example: when an unmanned mining car is driving, the load sensor detects that the current load is 45 tons, corresponding to the heavy load state; the road surface state sensor detects that the friction coefficient is 0.5, corresponding to slightly slippery; the rain sensor detects that the current precipitation is 1.5mm / h, and the visibility sensor detects that the visibility is 350 meters, corresponding to light rain weather. Based on the preset dynamic speed limit adjustment rules, the corresponding speed limit adjustment values ​​are -2km / h, -2km / h and -2km / h, respectively, and the total dynamic speed limit adjustment value is -6km / h after accumulation. If the static speed limit value at this location is 35km / h, the theoretical speed limit value is 29km / h.

[0046] In an optional embodiment, according to the real-time scene information, the real-time dynamic speed limit adjustment value is matched from the dynamic speed limit layer, including:

[0047] (1) Extracting real-time scene information of the unmanned vehicle at its current location;

[0048] The real-time scene information includes at least one of the following: the current load status information of the unmanned vehicle, the load status information includes one of the following information: an empty state and a heavy state; the road condition information corresponding to the current position, the road condition information includes at least one of the road slipperiness information and the road pothole degree information; the environment information corresponding to the current position, the environment information includes at least one of the traffic congestion degree, weather conditions and lighting conditions;

[0049] (2) Determine a dynamic speed limit adjustment value according to real-time scene information; wherein the dynamic speed limit adjustment value mapping table is used to represent the dynamic speed limit adjustment values ​​corresponding to different scene information.

[0050] In an optional embodiment, determining the dynamic speed limit adjustment value according to the scene information includes:

[0051] (1) Determining a dynamic speed limit adjustment value from a preset dynamic speed limit adjustment value mapping table according to the scene information, including:

[0052] (2) determining a target scene state of the unmanned vehicle based on the scene information, wherein the target scene state is used to represent driving environment information of the unmanned vehicle;

[0053] (3) determining a dynamic speed limit adjustment value set from a dynamic speed limit adjustment value mapping table based on a target scenario state, wherein the dynamic speed limit adjustment value set includes a plurality of dynamic speed limit adjustment values;

[0054] (4) Determine a dynamic speed limit adjustment value according to the dynamic speed limit adjustment value set.

[0055] In an optional embodiment, determining the dynamic speed limit adjustment value according to the dynamic speed limit adjustment value set includes: summing a plurality of dynamic speed limit adjustment values ​​in the dynamic speed limit adjustment value set to obtain the dynamic speed limit adjustment value.

[0056] Optionally, the scene information of the current position is extracted in real time for dynamic speed limit adjustment. Scene information extraction first collects raw data through on-board sensors, including load sensors, road surface state sensors, laser radars, millimeter-wave radars, precipitation sensors, visibility sensors, and light sensors. The load sensor collects load data every 100 milliseconds, the road surface state sensor collects friction coefficient every 500 milliseconds, the laser radar scans the road height data at a frequency of 10Hz, the millimeter-wave radar updates the target detection result every 1 second, the precipitation sensor and the visibility sensor sample once every 1 second, and the light sensor collects light intensity at a frequency of 1Hz. Feature extraction and state determination are performed on the collected raw data to obtain standardized scene state information. The load state determination is determined by comparing the ratio of the measured load to the rated load. A load greater than 80% is heavy load, and a load less than 20% is unloaded. The degree of road slipperiness is divided according to the friction coefficient. A load greater than 0.7 is dry, 0.4-0.7 is light slippery, and less than 0.4 is severe slippery. The degree of potholes on the road is determined based on the standard deviation of the road surface height. If it is less than 5mm, it is flat; if it is 5-15mm, it is slightly pothole-prone; and if it is greater than 15mm, it is severely pothole-prone. The degree of traffic congestion takes into account the average vehicle speed and the distance between vehicles within 200 meters ahead. If the average speed is greater than 40km / h and the distance is greater than 50 meters, it is not congested; if it is 20-40km / h or 30-50 meters, it is slightly congested; and if it is less than 20km / h or less than 30 meters, it is severely congested. The weather conditions are determined based on precipitation and visibility. If the precipitation is less than 0.5mm / h and the visibility is greater than 500 meters, it is sunny; if it is 0.5-2mm / h or 200-500 meters, it is light rain / light snow; and if it is greater than 2mm / h or less than 200 meters, it is heavy rain / heavy snow. The lighting conditions are based on 1000lux, with greater than 1000lux being daytime and less than 1000lux being night.

[0057] After the scene information is standardized, the adjustment value corresponding to each scene state is determined through the preset dynamic speed limit adjustment value mapping table. The dynamic speed limit adjustment value mapping table is as follows:

[0058] Dynamic speed limit adjustment value mapping table

[0059]

[0060] The mapping table stipulates that: the load state adjustment value is 0km / h for no load and -2km / h for heavy load; the road slipperiness adjustment value is 0km / h for dry, -2km / h for slightly slippery, and -4km / h for severely slippery; the road pothole adjustment value is 0km / h for flat, -2km / h for slightly potholey, and -4km / h for severely potholey; the traffic congestion adjustment value is 0km / h for no congestion, -2km / h for slightly congested, and -4km / h for severely congested; the weather condition adjustment value is 0km / h for sunny day, -2km / h for light rain / light snow, and -4km / h for heavy rain / heavy snow; the lighting condition adjustment value is 0km / h for daytime and -2km / h for nighttime. The dynamic speed limit adjustment values ​​corresponding to all scene states are summed to obtain the final comprehensive adjustment value.

[0061] For example, when an unmanned mining truck is driving, the load sensor continuously collects load data (46.5 tons, 46.3 tons, and 46.4 tons) within a sampling period of 100ms, takes an average value of 46.4 tons, and compares it with the rated load of 50 tons to determine it as a heavy load state; the road surface state sensor collects friction coefficient data (0.52, 0.51, 0.53, 0.50, and 0.51) within 500ms, and the average value of 0.514 is determined to be slightly slippery; the laser radar scans 100 road height sampling points at 10Hz, and the calculated standard deviation is 11.5mm, which is determined to be slightly pothole-prone; the millimeter-wave radar detects 5 vehicles in front within 1s, with an average speed of 32km / h and a minimum distance of 35 meters, which is determined to be slightly congested; the precipitation sensor measures a precipitation of 1.6mm / h, and the visibility sensor measures a visibility of 320 meters, which is comprehensively determined to be light rain; the light sensor measures a light intensity of 1850lux, which is determined to be daytime. The corresponding adjustment values ​​obtained by looking up the table are -2km / h (heavy load), -2km / h (slightly slippery), -2km / h (mild potholes), -2km / h (mild congestion), -2km / h (light rain), and 0km / h (daytime). After summing up, the final dynamic speed limit adjustment value of -10km / h is obtained.

[0062] In an optional embodiment, determining the empirical speed function comprises the following steps:

[0063] (1) Collecting driving data of manned vehicles under different working and road conditions; the driving data includes: driving position information and driving speed information corresponding to the driving position;

[0064] (2) The driving position information is used as the input of the neural network to be trained, and the corresponding driving speed information is used as the true value of the neural network to be trained. The neural network to be trained is trained to obtain an empirical speed function with the driving position information as the independent variable and the driving speed information as the dependent variable.

[0065] Optionally, see Figure 3, is a schematic diagram of neural network training in this application. First, the driving data of manned vehicles under different working conditions and road conditions are collected. These data are obtained by a data acquisition system installed on a manned mining car, and the acquisition system includes a GPS positioning module, a vehicle speed sensor, a posture sensor, etc. The GPS positioning module records the latitude and longitude coordinates and altitude of the vehicle every 100 milliseconds to form the vehicle's driving trajectory; the vehicle speed sensor collects real-time speed data every 50 milliseconds; the posture sensor records the vehicle's pitch angle, roll angle and other status information every 200 milliseconds. The collected raw data needs to be preprocessed, including three steps: data cleaning, time alignment and data standardization. In the data cleaning stage, invalid data points such as GPS signal loss and abnormal speed jump are deleted, and they are screened by setting thresholds, such as removing abnormal points with a speed change rate exceeding 5m / s2. Time alignment unifies the data of different sensors to the same timestamp sequence, and uses linear interpolation to supplement the missing data. Data standardization converts the position coordinates into displacement distances relative to the starting point, and the speed data is normalized to the range of 0-1.

[0066] The preprocessed data is used to train the deep neural network. The network input layer receives the vehicle position information, including relative displacement, road curvature, slope and other features; the output layer predicts the recommended driving speed at the location. The network adopts a multi-layer convolution structure. The first layer uses 32 3×3 convolution kernels to extract local features, the second layer uses 64 3×3 convolution kernels for feature combination, and the third layer is a fully connected layer to achieve feature mapping. Finally, the normalized speed value is output through the Sigmoid activation function. The training adopts the stochastic gradient descent optimization algorithm. The learning rate is initially set to 0.01, which is halved every 1000 iterations, and the batch size is 128.

[0067] For example, a mining area collected driving data for a month, and selected experienced drivers to carry out mining and transportation operations at different times and under different road conditions. The data collection system recorded the following data in a complete transportation task: the entire journey from the mining site to the unloading point was 5 kilometers, which took 15 minutes, and the GPS module recorded 3,000 location points with an average interval of 1.67 meters. The original data was cleaned, 35 GPS drift points and 28 speed anomaly points were removed, and 2,937 valid data samples were retained after cleaning. After time alignment, all sensor data were unified to a sampling interval of 100 milliseconds. Subsequently, data standardization was performed, and the 5-kilometer journey was converted into a relative displacement of 0-5000 meters, and the actual speed of 0-40km / h was normalized to the range of 0-1. Using 80% of the data as the training set and 20% as the validation set, the network converged after 2,000 rounds of training, and the average speed prediction error on the validation set was reduced to within 1.5km / h.

[0068] In an optional embodiment, a speed control strategy is generated according to a theoretical speed limit value and an empirical speed value, and a driving speed of the unmanned vehicle is controlled based on the speed control strategy, including:

[0069] (1) Determine the normalization coefficient of the theoretical speed limit value and the empirical speed value;

[0070] (2) Calculate the target speed limit value based on the theoretical speed limit value and the empirical speed value, as well as the corresponding normalization coefficients;

[0071] (3) Generate a speed control strategy according to the target speed limit value, and control the driving speed of the unmanned vehicle according to the speed control strategy.

[0072] Optionally, the two speed values ​​are made comparable through normalization processing, and the determination of the normalization coefficient is based on the statistical analysis of actual road test data. In the specific process, a typical road section is selected to collect driving data for one week, and the degree of difference between the theoretical speed limit value and the empirical speed value is recorded. After a large amount of data analysis, it is determined that the normalization coefficient of the theoretical speed limit value is 0.6, and the normalization coefficient of the empirical speed value is 0.4. The sum of the two is maintained at 1. Such a ratio not only ensures the safety speed limit requirements, but also appropriately incorporates the empirical driving characteristics.

[0073] The target speed limit is calculated by weighted summation, and the theoretical speed limit and the empirical speed value are weighted and combined with the normalization coefficient as the weight. For example, if the theoretical speed limit at a certain location is 40km / h and the empirical speed value is 35km / h, then the target speed limit at this location is 40×0.6+35×0.4=38km / h. This weighted method ensures that the final target speed limit will not exceed the safe speed limit range, and can also reflect the speed selection preference of the experienced driver at this location. When generating the speed control strategy based on the target speed limit, it is necessary to consider the vehicle dynamics constraints and adopt a smooth transition speed planning method. First, calculate the difference between the current speed and the target speed, and combine the maximum acceleration and deceleration constraints (usually the maximum acceleration is limited to 0.8m / s2, and the maximum deceleration is limited to 1.2m / s2) to determine the required acceleration or deceleration distance. Then adjust the speed within this distance, use a larger acceleration or deceleration for the first 60% of the distance, and gradually reduce the acceleration or deceleration for the last 40% of the distance to achieve a smooth transition.

[0074] For example, an unmanned mining vehicle drives to a continuous curved road section. The theoretical speed limit of the first curve in this area is 35km / h (including the static map speed limit of 40km / h and the dynamic speed limit adjustment of -5km / h), and the empirical speed value is 30km / h. The calculated target speed limit is 35×0.6+30×0.4=33km / h; the second curve is more sharp, the theoretical speed limit is reduced to 30km / h, and the empirical speed value is 25km / h. The calculated target speed limit is 30×0.6+25×0.4=28km / h. The current vehicle speed is 40km / h, and deceleration control needs to be carried out in two sections. The first deceleration interval is set to 100 meters, with the first 60 meters decelerating at 1m / s2 to 35km / h, and the last 40 meters smoothly transitioning to 33km / h at a deceleration of 0.5m / s2; the second deceleration interval is set to 80 meters, with the first 50 meters decelerating at 0.8m / s2 to 30km / h, and the last 30 meters smoothly transitioning to 28km / h at a deceleration of 0.4m / s2. The whole process achieves continuous and smooth speed adjustment, which not only meets the safety speed limit requirements, but also avoids sudden speed changes and braking, ensuring transportation efficiency and driving safety. Through this speed control strategy that considers theoretical speed limits and experience driving, unmanned mining vehicles can achieve reasonable speed planning under various road conditions, improving transportation efficiency while ensuring driving safety. Especially under complex road conditions, speed control that integrates experience driving data shows better adaptability and stability.

[0075] The speed control method of the unmanned vehicle in the embodiment of the present application is described above. The speed control device of the unmanned vehicle in the embodiment of the present application is described below. Figure 4 In one embodiment of the present application, a speed control device for an unmanned vehicle includes:

[0076] The acquisition module 201 is used to acquire the current position information of the driverless vehicle and acquire the corresponding theoretical speed limit value according to the current position information;

[0077] The calculation module 202 is used to calculate the empirical speed value corresponding to the current position of the unmanned vehicle according to a predetermined empirical speed function; wherein the empirical speed function is determined based on the driving data of the manned vehicle when driving to different positions;

[0078] The generating module 203 is used to generate a speed control strategy according to the theoretical speed limit value and the empirical speed value, and control the driving speed of the unmanned vehicle based on the speed control strategy.

[0079] The acquisition module 101 includes:

[0080] The acquisition submodule is used to obtain the static map speed limit value corresponding to the current location based on the current location information;

[0081] A determination submodule is used to determine a real-time dynamic speed limit adjustment value based on real-time scene information of the current location; and

[0082] The adjustment submodule is used to adjust the static map speed limit value by using the dynamic speed limit adjustment value to obtain the theoretical speed limit value.

[0083] The acquisition submodule includes:

[0084] An acquisition unit, used to acquire a static lane-level speed limit layer corresponding to the current position according to the current position information;

[0085] A matching unit, configured to obtain a static map speed limit value from the static lane-level speed limit layer based on the current position information;

[0086] The determination submodule comprises:

[0087] An analysis unit, configured to obtain a corresponding dynamic speed limit layer based on the current location information;

[0088] The screening unit is used to match and obtain a real-time dynamic speed limit adjustment value from the dynamic speed limit layer according to the real-time scene information.

[0089] The screening unit comprises:

[0090] An extraction subunit, used to extract real-time scene information of the unmanned vehicle at a current position;

[0091] The real-time scene information includes at least one of the following: the current load state information of the unmanned vehicle, the load state information includes one of the following information: an empty state and a heavy state; the road condition information corresponding to the current position, the road condition information includes at least one of the road slippery degree information and the road pothole degree information; the environmental information corresponding to the current position, the environmental information includes at least one of the traffic congestion degree, weather conditions and lighting conditions;

[0092] A mapping subunit is used to determine the dynamic speed limit adjustment value according to the real-time scene information; wherein the dynamic speed limit adjustment value mapping table is used to characterize the dynamic speed limit adjustment value corresponding to different scene information.

[0093] The mapping subunit is used for:

[0094] Determining a target scene state of the unmanned vehicle according to the scene information, wherein the target scene state is used to characterize driving environment information of the unmanned vehicle;

[0095] Based on the target scenario state, determining a dynamic speed limit adjustment value set from the dynamic speed limit adjustment value mapping table, wherein the dynamic speed limit adjustment value set includes a plurality of dynamic speed limit adjustment values;

[0096] The dynamic speed limit adjustment value is obtained by summing a plurality of dynamic speed limit adjustment values ​​in the dynamic speed limit adjustment value set.

[0097] The calculation module 202 is used to:

[0098] Collecting driving data of manned vehicles under different working and road conditions; the driving data includes: driving position information, and driving speed information corresponding to the driving position;

[0099] The driving position information is used as the input of the neural network to be trained, and the corresponding driving speed information is used as the true value of the neural network to be trained. The neural network to be trained is trained to obtain an empirical speed function with the driving position information as the independent variable and the driving speed information as the dependent variable.

[0100] The generating module 203 is used for:

[0101] Determining a normalization coefficient of the theoretical speed limit value and the empirical speed value;

[0102] Calculating a target speed limit value according to the theoretical speed limit value and the empirical speed value, and the normalization coefficients corresponding to each other;

[0103] A speed control strategy is generated according to the target speed limit value, and the driving speed of the unmanned vehicle is controlled according to the speed control strategy.

[0104] Through the collaboration of the above components, by obtaining the current position information in real time and dynamically calculating the theoretical speed limit, the driving speed of the unmanned vehicle is always within the safe range, effectively preventing the risk of speeding. Secondly, the introduction of the empirical speed function makes full use of the actual driving experience accumulated by manned vehicles, making the speed control of the unmanned vehicle closer to the actual driving habits and improving the transportation efficiency. Thirdly, the method of combining the theoretical speed limit value and the empirical speed value to generate the speed control strategy not only ensures driving safety but also takes into account the operating efficiency, so that the unmanned vehicle can make reasonable speed adjustments under different road conditions. In particular, in complex road conditions such as uphill and downhill, bends, etc., this method can give more appropriate speed control suggestions based on empirical data, avoiding the operational problems that may be caused by relying solely on theoretical speed limits. At the same time, this method realizes the transition from static speed limit to dynamic control, ensures the stability of vehicle operation through continuous speed adjustment, reduces unnecessary rapid acceleration and deceleration, and realizes more intelligent and efficient unmanned mine car speed control.

[0105] As described above, the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A speed control method for an unmanned vehicle, characterized in that: include: Obtain the current location information of the driverless vehicle, and obtain the corresponding theoretical speed limit value according to the current location information; Calculating an empirical speed value corresponding to the current position of the unmanned vehicle according to a predetermined empirical speed function; wherein the empirical speed function is determined based on driving data of a manned vehicle when driving to different positions; A speed control strategy is generated according to the theoretical speed limit value and the empirical speed value, and the driving speed of the unmanned vehicle is controlled based on the speed control strategy.

2. The method according to claim 1, characterized in that The theoretical speed limit value includes a static map speed limit value and a dynamic speed limit adjustment value; Get the corresponding theoretical speed limit value based on the current location information, including: According to the current location information, obtain the static map speed limit value corresponding to the current location; Determine a real-time dynamic speed limit adjustment value based on the real-time scene information of the current location; and The static map speed limit value is adjusted using the dynamic speed limit adjustment value to obtain the theoretical speed limit value.

3. The method according to claim 2, characterized in that According to the current location information, obtain the static map speed limit value corresponding to the current location, including: According to the current location information, obtain the static lane-level speed limit layer corresponding to the current location; Based on the current location information, a static map speed limit value is obtained by matching from the static lane-level speed limit layer; Based on the real-time scene information of the current location, determine the real-time dynamic speed limit adjustment value, including: Based on the current location information, obtaining a corresponding dynamic speed limit layer; According to the real-time scene information, a real-time dynamic speed limit adjustment value is obtained by matching from the dynamic speed limit layer.

4. The method according to claim 3, characterized in that According to the real-time scene information, a real-time dynamic speed limit adjustment value is obtained by matching from the dynamic speed limit layer, including: Extracting real-time scene information of the unmanned vehicle at the current position; The real-time scene information includes at least one of the following: the current load state information of the unmanned vehicle, the load state information includes one of the following information: an empty state and a heavy state; the road condition information corresponding to the current position, the road condition information includes at least one of the road slippery degree information and the road pothole degree information; the environmental information corresponding to the current position, the environmental information includes at least one of the traffic congestion degree, weather conditions and lighting conditions; The dynamic speed limit adjustment value is determined according to the real-time scene information; wherein the dynamic speed limit adjustment value mapping table is used to characterize the dynamic speed limit adjustment values ​​corresponding to different scene information.

5. The method according to claim 4, characterized in that Determining the dynamic speed limit adjustment value according to the scenario information includes: Determining the dynamic speed limit adjustment value from a preset dynamic speed limit adjustment value mapping table according to the scenario information includes: Determining a target scene state of the unmanned vehicle according to the scene information, wherein the target scene state is used to characterize driving environment information of the unmanned vehicle; Based on the target scenario state, determining a dynamic speed limit adjustment value set from the dynamic speed limit adjustment value mapping table, wherein the dynamic speed limit adjustment value set includes a plurality of dynamic speed limit adjustment values; The dynamic speed limit adjustment value is determined according to the dynamic speed limit adjustment value set.

6. The method according to claim 5, characterized in that Determining the dynamic speed limit adjustment value according to the dynamic speed limit adjustment value set includes: The dynamic speed limit adjustment value is obtained by summing a plurality of dynamic speed limit adjustment values ​​in the dynamic speed limit adjustment value set.

7. The method according to claim 1, characterized in that The empirical speed function is determined in the following manner: Collecting driving data of manned vehicles under different working and road conditions; the driving data includes: driving position information, and driving speed information corresponding to the driving position; The driving position information is used as the input of the neural network to be trained, and the corresponding driving speed information is used as the true value of the neural network to be trained. The neural network to be trained is trained to obtain an empirical speed function with the driving position information as the independent variable and the driving speed information as the dependent variable.

8. The method according to claim 1, characterized in that Generate a speed control strategy based on the theoretical speed limit value and the empirical speed value, and control the driving speed of the unmanned vehicle based on the speed control strategy, including: Determining a normalization coefficient of the theoretical speed limit value and the empirical speed value; Calculating a target speed limit value according to the theoretical speed limit value and the empirical speed value, and the normalization coefficients corresponding to each other; A speed control strategy is generated according to the target speed limit value, and the driving speed of the unmanned vehicle is controlled according to the speed control strategy.

9. A speed control device for an unmanned vehicle, characterized in that: include: An acquisition module is used to obtain the current position information of the unmanned vehicle and obtain the corresponding theoretical speed limit value according to the current position information; A calculation module, used to calculate an empirical speed value corresponding to the current position of the unmanned vehicle according to a predetermined empirical speed function; wherein the empirical speed function is determined based on driving data of a manned vehicle when driving to different positions; The generation module is used to generate a speed control strategy according to a theoretical speed limit value and an empirical speed value, and control the driving speed of the unmanned vehicle based on the speed control strategy.

10. An unmanned vehicle, characterized in that: include: A control device for an unmanned vehicle on a bumpy road as claimed in claim 9.

Images