CN111694019A - Intelligent driving education method based on laser radar and end-to-end control algorithm - Google Patents
Intelligent driving education method based on laser radar and end-to-end control algorithm Download PDFInfo
- Publication number
- CN111694019A CN111694019A CN202010402483.3A CN202010402483A CN111694019A CN 111694019 A CN111694019 A CN 111694019A CN 202010402483 A CN202010402483 A CN 202010402483A CN 111694019 A CN111694019 A CN 111694019A
- Authority
- CN
- China
- Prior art keywords
- data
- point cloud
- radar
- intelligent vehicle
- cloud data
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 31
- 238000012549 training Methods 0.000 claims abstract description 47
- 239000004576 sand Substances 0.000 claims abstract description 19
- 238000003062 neural network model Methods 0.000 claims abstract description 16
- 238000013507 mapping Methods 0.000 claims abstract description 7
- 238000012545 processing Methods 0.000 claims abstract description 4
- 230000006870 function Effects 0.000 claims description 30
- 238000004891 communication Methods 0.000 claims description 8
- 230000004913 activation Effects 0.000 claims description 7
- 238000013528 artificial neural network Methods 0.000 claims description 6
- 238000001514 detection method Methods 0.000 claims description 6
- 239000011159 matrix material Substances 0.000 claims description 5
- 230000001419 dependent effect Effects 0.000 claims description 3
- 238000004806 packaging method and process Methods 0.000 claims description 3
- 238000013135 deep learning Methods 0.000 abstract description 9
- 238000005516 engineering process Methods 0.000 description 8
- 238000013473 artificial intelligence Methods 0.000 description 7
- 238000011161 development Methods 0.000 description 7
- 230000008901 benefit Effects 0.000 description 5
- 230000000694 effects Effects 0.000 description 4
- 230000010354 integration Effects 0.000 description 4
- 240000007651 Rubus glaucus Species 0.000 description 2
- 235000011034 Rubus glaucus Nutrition 0.000 description 2
- 235000009122 Rubus idaeus Nutrition 0.000 description 2
- 238000013461 design Methods 0.000 description 2
- 230000003993 interaction Effects 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 238000012827 research and development Methods 0.000 description 2
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 description 1
- 230000009471 action Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 210000002569 neuron Anatomy 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 230000008569 process Effects 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 238000010408 sweeping Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
- G01S17/00—Systems using the reflection or reradiation of electromagnetic waves other than radio waves, e.g. lidar systems
- G01S17/88—Lidar systems specially adapted for specific applications
- G01S17/93—Lidar systems specially adapted for specific applications for anti-collision purposes
- G01S17/931—Lidar systems specially adapted for specific applications for anti-collision purposes of land vehicles
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Abstract
The invention relates to an intelligent driving education method based on a laser radar and an end-to-end control algorithm, which comprises the following steps: s1, controlling the miniature intelligent vehicle to run on the sand table track through the control terminal, and simultaneously starting the laser radar to scan the environment of the sand table track where the miniature intelligent vehicle is located to obtain control terminal operation instruction data and radar point cloud data; s2, performing data processing by adopting MATLAB according to the timestamp of the control terminal operation instruction data and the radar point cloud data to obtain the mapping of keyboard instruction data and point cloud data corresponding to the timestamp, and taking the keyboard instruction data and the point cloud data as training data; s3, inputting training data into a pre-constructed end-to-end control neural network model for training to obtain an end-to-end control model; the invention adopts an end-to-end control method to simplify deep learning knowledge, and realizes intelligent driving on a sand table through a simple laser radar and a miniature intelligent vehicle with high cost performance.
Description
Technical Field
The invention relates to the technical field of intelligent driving education, in particular to an intelligent driving education method based on a laser radar and an end-to-end control algorithm, which can show the basic principle of intelligent driving to students of all ages in a simple, easy and cost-effective mode to achieve the purpose of intelligent driving education.
Background
The integration of artificial intelligence technology and school education becomes a future trend and becomes an important driving force for educational development. "intelligent education" was written as a key task in "new generation artificial intelligence development planning" issued in 2017, and became an important component of the national strategy of artificial intelligence. However, there are many problems faced by intelligent education, the two most important of which are: how can talent reserve follow the needs of social science and technology development? In what manner is the integration of school education and intelligent technology achieved?
In the wave of artificial intelligence, unmanned driving is undoubtedly the technical field in which the most intense heat is studied. The thinking brought about by intelligent education is particularly in the unmanned level, and the problems are more obvious. First, the class of driverless education is significantly higher, and it is common for the student to be exposed to driverless research only during the student phase. However, in the high-level talent level, the development of unmanned education still faces many problems, and is limited by the cost of money and the like, and the problems and reality are often not effectively combined. As for the low age levels, driverless education is still in the vacuum phase. This undoubtedly presents a tremendous challenge to the unmanned talent reserve needs.
Therefore, the school rapidly combines the unmanned education and the artificial intelligence technology, the situation of solving the gap problem of the talents at the low age layer in the technical field of automatic driving becomes urgent, and how to design the unmanned display suitable for the school intelligent education becomes a more important ring in the problem. In order to achieve the unmanned technology education, a plurality of sensors are mounted on a vehicle to be tested. The system has the advantages that a plurality of sensors are provided, the research and development platform is precise, the system is complex, and students cannot understand the working principle and information interaction of each module; from the perspective of safety, students are inconvenient to get on the bus for experience; from the perspective of cost, the cost of real vehicles, including sensors with great value, as well as the field and oil consumption, is a problem which is difficult to face for schools. Therefore, conventional unmanned education is difficult to be popularized in school education.
Disclosure of Invention
Based on this, it is necessary to provide an intelligent driving education method based on a laser radar and an end-to-end control algorithm, aiming at the problems of complexity and high cost of the existing unmanned education system.
An intelligent driving education method based on a laser radar and an end-to-end control algorithm comprises the following steps:
s1, controlling the miniature intelligent vehicle to run on the sand table track through the control terminal, and simultaneously starting the laser radar to scan the environment of the sand table track where the miniature intelligent vehicle is located to obtain control terminal operation instruction data and radar point cloud data;
s2, performing data processing by adopting MATLAB according to the timestamp of the control terminal operation instruction data and the radar point cloud data to obtain the mapping of keyboard instruction data and point cloud data corresponding to the timestamp, and taking the keyboard instruction data and the point cloud data as training data;
s3, inputting training data into a pre-constructed end-to-end control neural network model for training to obtain optimal function fitting parameters and an end-to-end control model;
s4, writing the end-to-end control model into a launch file of a microcomputer of the miniature intelligent vehicle, and setting the miniature intelligent vehicle to be started automatically;
and S5, placing the miniature intelligent vehicle written into the end-to-end control model on the sand table runway, and starting a power supply to realize unmanned intelligent driving.
Preferably, step S1 includes: installing and compiling an arduino package and a dependent package of a control micro intelligent vehicle chassis, and setting parameters; writing a communication file and a keyboard control file in a microcomputer of the miniature intelligent vehicle; the method comprises the steps of placing the miniature intelligent vehicle on a sand table runway, establishing communication connection between a control terminal and the miniature intelligent vehicle, starting data recording and operating the miniature intelligent vehicle, wherein the data comprises control terminal operation instruction data and radar point cloud data.
Preferably, step S1 further includes: and respectively extracting the data in the data packet according to topics to obtain point cloud data of the radar and instruction data of the keyboard.
Preferably, step S2 includes: finding out a time period when the control terminal operation instruction data and the radar point cloud data are effective based on the timestamp labels, aligning the control terminal operation instruction data and the radar point cloud data one by one according to the timestamp, and packaging the data into a matrix; and marking the radar terminal operation instruction data and the radar point cloud data which are output by the MATLAB and correspond to the elapsed timestamp as training data.
Preferably, step S2 further includes: and in the point cloud data, point cloud data beyond the detection range of the laser radar is replaced by the radar detection distance maximum value, so that the finally obtained training data are readable numbers.
Preferably, step S3 includes: reference to the TensorFlow library; locally reading training data, assembling the training data into an array, wherein the training data are control terminal operation instruction data and radar point cloud data; setting parameters required by a fitting function in the end-to-end control neural network: weight W and offset b; defining an activation function of each layer network, and establishing connection between layers; defining cross entropy according to built-in functions of a reference library; selecting a built-in optimizer of the reference library to realize gradient descent; setting batch parameters and cycle times in the for cycle; and setting output parameters for controlling the storage path and accuracy of the neural network model end to end.
Preferably, the activation function is a tanh function, and the formula is:
preferably, the pre-constructed end-to-end control neural network model uses a TensorFlow library, including a 5-layer neural network, using an AdamaOptizer optimizer.
Preferably, in step S3, the training data is input into a pre-constructed end-to-end control neural network model for training, so as to obtain a function fitting parameter optimal based on the training data, an accuracy of the training data under the function fitting parameter, and the neural network model.
Preferably, the control terminal is a computer, and the microcomputer of the miniature intelligent vehicle is a raspberry type microcomputer.
Compared with the prior art, the invention has the beneficial effects that:
the invention simplifies the deep learning knowledge, realizes intelligent driving on a sand table by a simple laser radar and a miniature intelligent vehicle with high cost performance, and realizes the purpose of K12 deep learning education by obvious result comparison. The method comprises the following specific steps:
1. compared with other methods, the method directly establishes the mapping from input to action, and has a simple system structure and a clear result contrast.
2. The single-line laser radar is used as input, data are simple and efficient, interpretability and acceptability for low-age-level education are high, and interest of low-age-level students in intelligent driving and deep learning to artificial intelligence is stimulated.
3. The sand table, the miniature intelligent vehicle and the single-line laser radar are adopted, the cost is saved to the greatest extent, the operability and the variability of the sand table runway can be repeated, and the development course of the school can be favorably developed for teaching.
Drawings
Fig. 1 is a schematic flowchart of an intelligent driving education method based on a lidar and an end-to-end control algorithm according to an embodiment.
Fig. 2 is another schematic flowchart of an intelligent driving education method based on a lidar and an end-to-end control algorithm according to an embodiment.
Detailed Description
The scheme considers that the aim is to realize a set of unmanned technology education demonstration aiming at K12 school layer. In the technical education level, artificial intelligence is realized, and the interpretability of the realization process and principle is accepted by the understanding of a target science layer as much as possible; from the reality, as a set of education shows, the cost performance is considered, namely whether the education course is suitable for development and coverage. From the two points, the experimental object is firstly determined to be a miniature intelligent vehicle, namely the unmanned education trolley in the final display. Compared with a miniature intelligent vehicle, the real vehicle unmanned experiment has the advantages of more sensors, precise research and development platform, complex system and obvious advantages in the aspects of scientific research and social development macro. However, the problem of integration of education and intelligent technology of the eye-catching school is solved, and the defect of the miniature education trolley is changed into an advantage. From the perspective of safety, students are inconvenient to get on the bus for experience; in the aspect of learning, a real vehicle system is too complex, so that students cannot understand the working principle and information interaction of each module in a popular and easy way; from the perspective of cost, the cost of real vehicles, including sensors with great value, as well as the field and oil consumption, is a problem which is difficult to face for schools. The unmanned education dolly that this show was used, to its difference with the real car of unmanned, make good use of the strong points and keep away the weak point, can realize the high integration of school's education and intelligent technology.
In the aspect of sensor selection, a single-line laser radar is selected as a sensor, an end-to-end control algorithm is used as an unmanned technical method, and other technical tools are used as an auxiliary tool to realize unmanned education display. The method has the advantages that the single-line laser radar has high cost performance, does not make people feel strange when being used as a daily tool such as a sweeping robot and the like, and the generated point cloud data is simpler in principle explanation and has better effect on the realization of an end-to-end control algorithm; on the other hand, the mapping relation of the end-to-end control algorithm is simpler in principle explanation, and the inexplicability between the two ends of the mapping can just stimulate the interest of the target chemistry layer in deep learning. The method comprises the following specific steps:
referring to fig. 1-2, an intelligent driving education method based on a laser radar and an end-to-end control algorithm includes:
s1, controlling the miniature intelligent vehicle to run on the sand table track through the control terminal, and simultaneously starting the laser radar to scan the environment of the sand table track where the miniature intelligent vehicle is located to obtain control terminal operation instruction data and radar point cloud data; specifically, step S1 includes: installing and compiling an arduino package and a dependent package of a control micro intelligent vehicle chassis, and setting parameters; writing a communication file and a keyboard control file in a microcomputer of the miniature intelligent vehicle; the method comprises the steps of placing the miniature intelligent vehicle on a sand table runway, establishing communication connection between a control terminal and the miniature intelligent vehicle, starting data recording and operating the miniature intelligent vehicle, wherein the data comprises control terminal operation instruction data and radar point cloud data.
In this embodiment, the control terminal is a computer, and the microcomputer of the smart car is a raspberry-type microcomputer. The keyboard of the computer is in communication connection with the miniature intelligent vehicle through a wireless network, the wireless network is a launch start file written in a raspberry group in advance, the launch start file can ensure that the miniature intelligent vehicle is started along with a power supply to establish a wireless network hotspot, the computer establishes communication connection with the miniature intelligent vehicle through the connection hotspot, and then the running of the miniature intelligent vehicle is controlled by starting the keyboard control file. The steering mode of the miniature intelligent vehicle is differential rotation.
Wherein, step S1 further includes: and respectively extracting the data in the data packet according to topics to obtain point cloud data of the radar and instruction data of the keyboard.
S2, performing data processing by adopting MATLAB according to the timestamp of the control terminal operation instruction data and the radar point cloud data to obtain the mapping of keyboard instruction data and point cloud data corresponding to the timestamp, and taking the keyboard instruction data and the point cloud data as training data; specifically, step S2 includes: finding out a time period when the control terminal operation instruction data and the radar point cloud data are effective based on the timestamp labels, aligning the control terminal operation instruction data and the radar point cloud data one by one according to the timestamp, and packaging the data into a matrix; and marking the radar terminal operation instruction data and the radar point cloud data which are output by the MATLAB and correspond to the elapsed timestamp as training data.
Wherein, step S2 further includes: in the point cloud data, aiming at the point cloud data (embodied as inf in the data) which exceeds the detection range of the laser radar, inf is replaced by the radar detection distance maximum value, so that the finally obtained training data are readable numbers. And generating a vector with a label of 4 x 1 for a keyboard instruction of the computer, sequentially representing front, left, right and stop, wherein the label is adapted to a softmax function so as to ensure that the sum of the probabilities of all the classifications is 1.
S3, inputting training data into a pre-constructed end-to-end control neural network model for training to obtain optimal function fitting parameters and an end-to-end control model; specifically, step S3 includes:
reference to the TensorFlow library;
locally reading training data (two groups of data preprocessed by MATLAB), and assembling the training data into an array, wherein the training data is control terminal operation instruction data and radar point cloud data;
setting parameters required by a fitting function in the end-to-end control neural network: weight W and offset b; and carrying out initialization assignment on W and b by using tf.variable, wherein the matrix specification of each layer of W and b is obtained by the output of the previous layer, the initial W line number is the angle precision in the radar point cloud data, and the angle precision of the radar used in the invention is 360, namely, each degree corresponds to a distance value. The initial W column number is the next W row number, and is also the specification of the initial b matrix. The specification b of the last layer corresponds to an instruction label of the keyboard, and the noon is 4 in the invention, namely, the front, the left, the right and the stop of the keyboard instruction;
defining an activation function of each layer network, and establishing connection between layers; in order to ensure that the network finally obtains a nonlinear output which is not the linear output of input data but has the function of a neuron, a nonlinear activation function is used between layers. The activation function is a tanh function, and the formula is as follows:
defining cross entropy according to built-in functions of a reference library;
selecting a built-in optimizer of the reference library to realize gradient descent;
setting batch parameters and cycle times in the for cycle;
and setting output parameters for controlling the storage path and accuracy of the neural network model end to end. Namely, the fitting effect of the parameters in the training data is output once for a certain number of times in each cycle, and finally the accuracy of fitting of one parameter is output for all the training data.
In this embodiment, before the design cycle, setting and defining network parameters includes: defining a difference value between the target and the predicted value by using a TensorFlow built-in function, namely cross entropy; selecting a proper optimizer to realize gradient reduction of a loss function and ensure continuous optimization of the fitting effect of the fitting function, wherein an Adam optimizer is selected in the embodiment of the invention; the batch specification and the training cycle number in the for cycle are set, and the efficiency of training is reduced when the batch specification and the training cycle number are too large, and the precision of training is reduced when the batch specification and the training cycle number are too small, so that specific setting needs to be carried out according to the data volume of input data.
Further, some data points can be randomly discarded aiming at the packaged data so as to improve the robustness of the network, on the premise that the corresponding time stamps of the two groups of data are not influenced.
In step S3, the training data is input into a pre-constructed end-to-end control neural network model for training, and an optimal function fitting parameter based on the training data, an accuracy of the training data under the function fitting parameter, and the neural network model are obtained. The pre-constructed end-to-end control neural network model uses a TensorFlow library, comprises 5 layers of neural networks and adopts an AdamaOptizer optimizer. The TensorFlow library is an open source software library widely applied to machine learning and deep learning. In TensorFlow, a data container is regarded as a tensor, different data operations are regarded as a node, a connection line between nodes and tensors and between nodes is called edge, and the tensor walks through the corresponding node along the different edge to finally obtain a data operation result, namely the flow of tensor. In TensorFlow, aiming at training data with simple data format but huge quantity, the operation of end-to-end control can be efficiently realized, and a concise dataflow graph is output, so that the TensorFlow algorithm is more suitable for displaying and understanding the K12 education level compared with other algorithms.
S4, writing the end-to-end control model into a launch file of a microcomputer of the miniature intelligent vehicle, and setting the miniature intelligent vehicle to be started automatically; the input of the end-to-end control model is point cloud data acquired by the laser radar, the output is a corresponding operation instruction, and the end-to-end of the radar and the operation is established.
As shown in fig. 2, in step S4, the launch file should include preprocessed data and a trained network model, i.e., an end-to-end control model. Wherein the quality criterion of the network model can be determined by the accuracy of the output.
And S5, placing the miniature intelligent vehicle written into the end-to-end control model on the sand table runway, and starting a power supply to realize unmanned intelligent driving (automatic driving).
In conclusion, the invention discloses an intelligent driving education method based on a laser radar and an end-to-end control algorithm, which adopts the end-to-end control algorithm with obvious comparison effect and simple and understandable principle in deep learning to simplify deep learning knowledge, simultaneously adopts the laser radar with high cost performance and a miniature intelligent vehicle to realize intelligent driving on a sand table, and realizes the purpose of K12 (students at all ages) deep learning education through obvious result comparison.
The technical features of the embodiments described above may be arbitrarily combined, and for the sake of brevity, all possible combinations of the technical features in the embodiments described above are not described, but should be considered as being within the scope of the present specification as long as there is no contradiction between the combinations of the technical features.
The above-mentioned embodiments only express several embodiments of the present invention, and the description thereof is more specific and detailed, but not construed as limiting the scope of the invention. It should be noted that, for a person skilled in the art, several variations and modifications can be made without departing from the inventive concept, which falls within the scope of the present invention. Therefore, the protection scope of the present patent shall be subject to the appended claims.
Claims (10)
1. An intelligent driving education method based on a laser radar and an end-to-end control algorithm is characterized by comprising the following steps:
s1, controlling the miniature intelligent vehicle to run on the sand table track through the control terminal, and simultaneously starting the laser radar to scan the environment of the sand table track where the miniature intelligent vehicle is located to obtain control terminal operation instruction data and radar point cloud data;
s2, performing data processing by adopting MATLAB according to the timestamp of the control terminal operation instruction data and the radar point cloud data to obtain the mapping of keyboard instruction data and point cloud data corresponding to the timestamp, and taking the keyboard instruction data and the point cloud data as training data;
s3, inputting training data into a pre-constructed end-to-end control neural network model for training to obtain optimal function fitting parameters and an end-to-end control model;
s4, writing the end-to-end control model into a launch file of a microcomputer of the miniature intelligent vehicle, and setting the miniature intelligent vehicle to be started automatically;
and S5, placing the miniature intelligent vehicle written into the end-to-end control model on the sand table runway, and starting a power supply to realize unmanned intelligent driving.
2. The intelligent driving education method based on the lidar and the end-to-end control algorithm of claim 1, wherein the step S1 comprises:
installing and compiling an arduino package and a dependent package of a control micro intelligent vehicle chassis, and setting parameters;
writing a communication file and a keyboard control file in a microcomputer of the miniature intelligent vehicle;
the method comprises the steps of placing the miniature intelligent vehicle on a sand table runway, establishing communication connection between a control terminal and the miniature intelligent vehicle, starting data recording and operating the miniature intelligent vehicle, wherein the data comprises control terminal operation instruction data and radar point cloud data.
3. The intelligent driving education method based on lidar and end-to-end control algorithm of claim 2, wherein the step S1 further comprises:
and respectively extracting the data in the data packet according to topics to obtain point cloud data of the radar and instruction data of the keyboard.
4. The intelligent driving education method based on the lidar and the end-to-end control algorithm of claim 1, wherein the step S2 comprises:
finding out a time period when the control terminal operation instruction data and the radar point cloud data are effective based on the timestamp labels, aligning the control terminal operation instruction data and the radar point cloud data one by one according to the timestamp, and packaging the data into a matrix;
and marking the radar terminal operation instruction data and the radar point cloud data which are output by the MATLAB and correspond to the elapsed timestamp as training data.
5. The intelligent driving education method based on lidar and end-to-end control algorithm of claim 4, wherein the step S2 further comprises:
and in the point cloud data, point cloud data beyond the detection range of the laser radar is replaced by the radar detection distance maximum value, so that the finally obtained training data are readable numbers.
6. The intelligent driving education method based on the lidar and the end-to-end control algorithm of claim 5, wherein the step S3 comprises:
reference to the TensorFlow library;
locally reading training data, assembling the training data into an array, wherein the training data are control terminal operation instruction data and radar point cloud data;
setting parameters required by a fitting function in the end-to-end control neural network: weight W and offset b;
defining an activation function of each layer network, and establishing connection between layers;
defining cross entropy according to built-in functions of a reference library;
selecting a built-in optimizer of the reference library to realize gradient descent;
setting batch parameters and cycle times in the for cycle;
and setting output parameters for controlling the storage path and accuracy of the neural network model end to end.
8. the intelligent driving education method based on the lidar and the end-to-end control algorithm as claimed in claim 7, wherein the pre-constructed end-to-end control neural network model uses TensorFlow library, comprises 5 layers of neural network, and adopts AdamaOptimizer optimizer.
9. The intelligent driving education method based on lidar and end-to-end control algorithm of claim 8 wherein, in step S3, the training data is inputted into the pre-constructed end-to-end control neural network model for training, resulting in optimal function fitting parameters based on the training data, accuracy of the training data under the function fitting parameters, and the neural network model.
10. The intelligent driving education method based on the lidar and the end-to-end control algorithm as claimed in claim 1, wherein the control terminal is a computer, and the microcomputer of the miniature intelligent vehicle is a raspberry-type microcomputer.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010402483.3A CN111694019A (en) | 2020-05-13 | 2020-05-13 | Intelligent driving education method based on laser radar and end-to-end control algorithm |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010402483.3A CN111694019A (en) | 2020-05-13 | 2020-05-13 | Intelligent driving education method based on laser radar and end-to-end control algorithm |
Publications (1)
Publication Number | Publication Date |
---|---|
CN111694019A true CN111694019A (en) | 2020-09-22 |
Family
ID=72477744
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202010402483.3A Pending CN111694019A (en) | 2020-05-13 | 2020-05-13 | Intelligent driving education method based on laser radar and end-to-end control algorithm |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN111694019A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN117272685A (en) * | 2023-11-17 | 2023-12-22 | 西南交通大学 | Optimal sanding decision simulation method based on train operation parameters |
Citations (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104407612A (en) * | 2014-10-16 | 2015-03-11 | 重庆市合川区何茗机械加工厂 | Miniature intelligent vehicle and test method applying miniature intelligent vehicle |
CN108227707A (en) * | 2017-12-25 | 2018-06-29 | 清华大学苏州汽车研究院(吴江) | Automatic Pilot method based on laser radar and end-to-end deep learning method |
CN108595844A (en) * | 2018-04-26 | 2018-09-28 | 成都博士信智能科技发展有限公司 | Automatic Pilot control method and device |
CN109326195A (en) * | 2018-12-19 | 2019-02-12 | 天津卡达克数据有限公司 | A kind of sand table demo system for the advanced road traffic in simcity |
CN109461342A (en) * | 2018-12-19 | 2019-03-12 | 畅加风行(苏州)智能科技有限公司 | A kind of tutoring system and its teaching method for unmanned motor vehicle |
CN109508009A (en) * | 2018-12-13 | 2019-03-22 | 合肥工业大学 | A kind of miniature vehicle of intelligence with vehicle vehicle communication function |
US20190286153A1 (en) * | 2018-03-15 | 2019-09-19 | Nvidia Corporation | Determining drivable free-space for autonomous vehicles |
CN110346808A (en) * | 2019-07-15 | 2019-10-18 | 上海点积实业有限公司 | A kind of Processing Method of Point-clouds and system of laser radar |
-
2020
- 2020-05-13 CN CN202010402483.3A patent/CN111694019A/en active Pending
Patent Citations (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104407612A (en) * | 2014-10-16 | 2015-03-11 | 重庆市合川区何茗机械加工厂 | Miniature intelligent vehicle and test method applying miniature intelligent vehicle |
CN108227707A (en) * | 2017-12-25 | 2018-06-29 | 清华大学苏州汽车研究院(吴江) | Automatic Pilot method based on laser radar and end-to-end deep learning method |
US20190286153A1 (en) * | 2018-03-15 | 2019-09-19 | Nvidia Corporation | Determining drivable free-space for autonomous vehicles |
CN108595844A (en) * | 2018-04-26 | 2018-09-28 | 成都博士信智能科技发展有限公司 | Automatic Pilot control method and device |
CN109508009A (en) * | 2018-12-13 | 2019-03-22 | 合肥工业大学 | A kind of miniature vehicle of intelligence with vehicle vehicle communication function |
CN109326195A (en) * | 2018-12-19 | 2019-02-12 | 天津卡达克数据有限公司 | A kind of sand table demo system for the advanced road traffic in simcity |
CN109461342A (en) * | 2018-12-19 | 2019-03-12 | 畅加风行(苏州)智能科技有限公司 | A kind of tutoring system and its teaching method for unmanned motor vehicle |
CN110346808A (en) * | 2019-07-15 | 2019-10-18 | 上海点积实业有限公司 | A kind of Processing Method of Point-clouds and system of laser radar |
Non-Patent Citations (7)
Title |
---|
冀杰;黄岩军;李云伍;吴飞;: "基于有限状态机的车辆自动驾驶行为决策分析" * |
刘子俊;张应鹏;林汉毅;熊驰;郑倩;: "基于智能微缩车的车道线检测算法验证" * |
张应鹏: "智能网联环境下协同式自适应巡航算法研究及试验验证" * |
杨恺等: "基于深度学习的车辆检测方法", 《计算机与网络》 * |
杨梓谦等: "基于深度学习的自动驾驶小车系统设计", 《网络信息工程》 * |
苏致远;马育林;李建市;周晶晶;: "基于多缩微车的智能交通半实物仿真平台" * |
许琳昊;朱舜;王立群;何军;: "基于强化学习与Lidar-SLAM技术智能小车开发" * |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN117272685A (en) * | 2023-11-17 | 2023-12-22 | 西南交通大学 | Optimal sanding decision simulation method based on train operation parameters |
CN117272685B (en) * | 2023-11-17 | 2024-01-26 | 西南交通大学 | Optimal sanding decision simulation method based on train operation parameters |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Tanimoto | Evolutionary games with sociophysics | |
Mousavi et al. | Traffic light control using deep policy‐gradient and value‐function‐based reinforcement learning | |
Bhattacharyya et al. | Simulating emergent properties of human driving behavior using multi-agent reward augmented imitation learning | |
CN106096729A (en) | A kind of towards the depth-size strategy learning method of complex task in extensive environment | |
US20210165375A1 (en) | Inverse reinforcement learning with model predictive control | |
CN110534009A (en) | A kind of unmanned course teaching aid of artificial intelligence | |
CN108333959A (en) | A kind of energy saving method of operating of locomotive based on convolutional neural networks model | |
Zhang et al. | A systematic solution of human driving behavior modeling and simulation for automated vehicle studies | |
CN112264999A (en) | Method, device and storage medium for intelligent agent continuous space action planning | |
CN114282433A (en) | Automatic driving training method and system based on combination of simulation learning and reinforcement learning | |
CN111694019A (en) | Intelligent driving education method based on laser radar and end-to-end control algorithm | |
Duan et al. | THE INNOVATIVE MODEL OF ARTIFICIAL INTELLIGENCE COMPUTER EDUCATION UNDER THE BACKGROUND OF EDUCATIONAL INNOVATION | |
Hu et al. | A data-driven solution for energy management strategy of hybrid electric vehicles based on uncertainty-aware model-based offline reinforcement learning | |
Sebestyen et al. | Machine learning methods in energy simulations for architects and designers | |
Rabault et al. | 18 Deep Reinforcement Learning Applied to Active Flow Control | |
CN101647033B (en) | Computer implemented method and apparatus for modeling and simulating human behavior | |
Jebessa et al. | Analysis of reinforcement learning in autonomous vehicles | |
Contardo et al. | Learning states representations in pomdp | |
Zhang et al. | ORCANet: Differentiable multi‐parameter learning for crowd simulation | |
Gonzalez et al. | Automating the CGF model development and refinement process by observing expert behavior in a simulation | |
Zarrouki | Reinforcement learning of model predictive control parameters for autonomous vehicle guidance | |
Argall et al. | Learning mobile robot motion control from demonstrated primitives and human feedback | |
Guo et al. | Simulated CAVs driving and characteristics of the mixed traffic using reinforcement learning method | |
Kumar et al. | A Novel Algorithm for Optimal Trajectory Generation Using Q Learning | |
Hassab et al. | Generative Design Methodology for Double Curved Surfaces using AI |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20200922 |
|
RJ01 | Rejection of invention patent application after publication |