CN109709956B - Multi-objective optimized following algorithm for controlling speed of automatic driving vehicle - Google Patents
Multi-objective optimized following algorithm for controlling speed of automatic driving vehicle Download PDFInfo
- Publication number
- CN109709956B CN109709956B CN201811600366.7A CN201811600366A CN109709956B CN 109709956 B CN109709956 B CN 109709956B CN 201811600366 A CN201811600366 A CN 201811600366A CN 109709956 B CN109709956 B CN 109709956B
- Authority
- CN
- China
- Prior art keywords
- data
- model
- headway
- ttc
- following
- 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.)
- Active
Links
Images
Landscapes
- Traffic Control Systems (AREA)
- Control Of Driving Devices And Active Controlling Of Vehicle (AREA)
- Feedback Control In General (AREA)
Abstract
The invention develops a multi-objective optimized following algorithm for controlling the speed of an automatic driving vehicle. The algorithm provides a model for controlling the automobile following speed based on deep reinforcement learning, and the model not only simulates human driving, but also directly optimizes driving safety, efficiency and comfort. A reward function reflecting driving safety, efficiency and comfort is constructed by combining collision time, headway empirical distribution and acceleration, an actual driving data training model in a Next Generation Simulation (NGSIM) project is used, the following behavior simulated by the model is compared with the behavior observed in NGSIM empirical data, and the reinforced learning intelligent body learns the vehicle speed safely, comfortably and efficiently in a mode of maximizing accumulated reward through tests and trial and error in a Simulation environment. The results show that the proposed following speed control algorithm shows better safe, efficient and comfortable driving ability compared to a real-world human driver.
Description
Technical Field
The invention relates to the field of automatic driving following control, in particular to a following algorithm for multi-objective optimization of speed control of an automatic driving vehicle.
Background
The following control is an important component of automatic driving intelligent decision, and comprises speed selection in free driving, vehicle distance keeping in vehicle following and braking in emergency. Under the condition that automatic driving and human driving coexist, the automatic driving vehicle makes a follow-up control decision similar to a human driver (anthropomorphic for short), so that the comfort level and the trust degree of passengers are improved, and meanwhile, other traffic participants can understand and predict the behavior of the automatic driving vehicle better, so that the safety interaction between the automatic driving and the human driving is realized. However, the traditional following model has many limitations when being applied to automatic following control, such as limitation on flexibility and accuracy of the model, difficulty in popularization to driving scenes and drivers except for calibration data, and incapability of reflecting driving styles and driving scenes of actual drivers of vehicles when being applied to automatic driving.
Deep Learning (DRL) is widely used in the fields of industrial manufacturing, simulation, robot control, optimization and scheduling, game playing, etc., and its basic idea is to learn the optimal strategy to achieve the goal by maximizing the accumulated award value obtained from the environment by the intelligent agent. The DRL method focuses more on learning a problem solving strategy, and does not fit data, so that the generalization capability of the DRL method is stronger, and reference is provided for automatic driving vehicle following control.
Disclosure of Invention
The purpose of the invention is: a multi-objective optimized following algorithm for controlling the speed of an automatically driven vehicle. The algorithm proposes a model for vehicle following speed control that directly optimizes driving safety, efficiency and comfort. Combining the Time To Collision (TTC), the experience distribution of the time headway and the Jerk (Jerk), a reward function reflecting the driving safety, the efficiency and the comfort is constructed, an actual driving data training model in a Next Generation Simulation (NGSIM) project is used, the following behavior simulated by the model is compared with the behavior observed in the NGSIM experience data, and the reinforcement learning intelligent body learns the vehicle speed safely, comfortably and efficiently in a mode of maximizing the accumulated reward through tests and trial and error in a simulation environment. The results show that the proposed following speed control algorithm shows better safe, efficient and comfortable driving ability compared to a real-world human driver.
The technical scheme adopted by the invention is as follows:
a multi-objective optimized following algorithm for controlling the speed of an automatic driving vehicle comprises the following steps:
step 1: data is acquired. And (3) extracting the following events based on the criteria that the front vehicle and the rear vehicle stay on the same lane and the length of the vehicle following events is greater than 15 seconds by using the data in the NGSIM project, and taking one part as training data and the other part as test data based on the extracted following events.
Step 2: a reward function is constructed. Characteristic quantities reflecting the relevant objectives (safety, comfort, efficiency) of the following control of the vehicle are proposed.
Step 2.1: time To Collision (TTC) is used to reflect safety. TTC represents the amount of time remaining before a collision of two vehicles and is formulated asWhere Sn-1, n (t) is the inter-vehicle distance, Δ Vn-1, n (t) is the relative velocity. Determining the safety threshold value to be 7 seconds according to NGSIM empirical data, and performing TTC feature construction:if TTC is less than 7 seconds, the TTC characteristic index is a negative value, and as the TTC approaches zero, the TTC characteristic approaches negative infinity, and the most severe punishment is shown for the situation of approaching collision.
Step 2.2: the driving efficiency is measured by headway. From the analysis, the lognormal distribution is adapted to the distribution of the acquired training data with a probability density function ofx>0. From the extracted data, it can be estimated that the mean μ and the logarithmic standard deviation σ of the distribution variable x are 0.4226 and 0.4365, respectively. And constructing the headway characteristics into the probability density value of the estimated headway lognormal distribution: fheadway ═ flognormal (headway | μ ═ 0.4226, σ ═ 0.4365). According to the headway time characteristic, the headway time of about 1.3 seconds corresponds to a high characteristic value, and the headway time is too long or too short and corresponds to a low characteristic value, so that the characteristic value estimates the high-flow headway maintenance behavior, and meanwhile punishs the unsafe or too long headway maintenance behavior.
Step 2.3: the driving comfort is measured by adopting the change rate Jerk of the acceleration, and the characteristic is as follows:
step 2.4: and establishing a comprehensive reward function. R is w1FTTC + w2Fheadway + w3fjerk, where w1, w2, w3 are coefficients of the features, all set to 1.
And step 3: and (5) training the model. And during each training, sequentially simulating the following events in the data, repeating the training for multiple times, and selecting the model which obtains the maximum average reward on the test data as the final model.
And 4, step 4: and (6) evaluating the model. And (4) comparing and evaluating the following behaviors obtained by the NGSIM data and the DDPG model simulation by using indexes such as TTC, headway, jerk and the like.
The invention has the advantages that:
1. the developed autonomous vehicle-following control logic is applicable to autonomous vehicle development;
2. the algorithmic model does not mimic human driving, but directly optimizes driving safety, efficiency, and comfort.
Drawings
FIG. 1 is a flow chart of the present invention.
Fig. 2NGSIM data is compared to the DDPG model driving safety.
Fig. 3 comparison of driving comfort between NGSIM data and DDPG models.
Detailed Description
The algorithm proposes a model for automobile follow-up speed control based on deep reinforcement learning, and the model does not imitate human driving but directly optimizes driving safety, efficiency and comfort. Combining the Time To Collision (TTC), the experience distribution of the time headway and the Jerk (Jerk), a reward function reflecting the driving safety, the efficiency and the comfort is constructed, an actual driving data training model in a Next Generation Simulation (NGSIM) project is used, the following behavior simulated by the model is compared with the behavior observed in the NGSIM experience data, and the reinforcement learning intelligent body learns the vehicle speed safely, comfortably and efficiently in a mode of maximizing the accumulated reward through tests and trial and error in a simulation environment. The results show that the proposed following speed control algorithm shows better safe, efficient and comfortable driving ability compared to a real-world human driver. The results show that the proposed following speed control algorithm shows better safe, efficient and comfortable driving ability compared to a real-world human driver.
The invention is described in detail below with reference to the following figures and specific examples, the steps of which are as follows:
step 1: data is acquired. And (3) extracting a following event based on the criteria that a front vehicle and a rear vehicle stay on the same lane and the length of the vehicle following event is greater than 15 seconds and the like by using data in a Next Generation Simulation (NGSIM) project, and taking one part as training data and the other part as test data based on the extracted following event.
Step 2: a reward function is constructed. Characteristic quantities reflecting the relevant objectives (safety, comfort, efficiency) of the following control of the vehicle are proposed.
Step 2.1: time To Collision (TTC) is used to reflect safety. TTC represents the amount of time remaining before a collision of two vehicles and is formulated asWhere Sn-1, n (t) is the inter-vehicle distance, Δ Vn-1, n (t) is the relative velocity. Determining the safety threshold value to be 7 seconds according to NGSIM empirical data, and performing TTC feature construction:if TTC is less than 7 seconds, the TTC characteristic index is a negative value, and as the TTC approaches zero, the TTC characteristic approaches negative infinity, and the most severe punishment is shown for the situation of approaching collision.
Step 2.2: the driving efficiency is measured by headway. From the analysis, the lognormal distribution is adapted to the distribution of the acquired training data with a probability density function ofx>0. Based on the extracted dataIt can be estimated that the mean μ and the logarithmic standard deviation σ of the distribution variable x are 0.4226 and 0.4365, respectively. And constructing the headway characteristics into the probability density value of the estimated headway lognormal distribution: fheadway ═ flognormal (headway | μ ═ 0.4226, σ ═ 0.4365). According to the headway time characteristic, the headway time of about 1.3 seconds corresponds to a high characteristic value, and the headway time is too long or too short and corresponds to a low characteristic value, so that the characteristic value estimates the high-flow headway maintenance behavior, and meanwhile punishs the unsafe or too long headway maintenance behavior.
Step 2.3: the driving comfort is measured by adopting the change rate Jerk of the acceleration, and the characteristic is as follows:
step 2.4: and establishing a comprehensive reward function. Establishing r ═ w1FTTC + w2Fheadway + w3fjerk according to the above steps 2.1, 2.2, 2.3, where w1, w2, w3 are coefficients of the features, all set to 1.
And step 3: and (5) training the model. And during each training, sequentially simulating the following events in the data, repeating the training for multiple times, and selecting the model which obtains the maximum average reward on the test data as the final model.
And 4, step 4: and (6) evaluating the model. And (4) comparing and evaluating the following behaviors obtained by the NGSIM data and the DDPG model simulation by using indexes such as TTC, headway, jerk and the like.
Examples
By comparing the empirical NGSIM data with the driving following behavior simulated by the DDPG model, the model can be tested to safely, efficiently and comfortably follow the front vehicle.
Data is acquired. Using the data in the NGSIM project, the following event is extracted based on criteria such as the leading and trailing vehicles staying on the same lane and the length of the vehicle following event >15 seconds.
In terms of driving safety, a following event is randomly selected from the NGSIM data set. FIG. 2 shows the observed velocities, spacings, and accelerations, along with the corresponding index values generated by the DDPG model. The driver in the NGSIM data drives at a very small inter-vehicle distance after 10 seconds, while the DDPG model always maintains a following gap of about 10 meters.
In terms of driving comfort, a follow-up event is randomly selected in the NGSIM dataset. FIG. 3 shows the observed velocity, pitch, acceleration and Jerk values, along with the corresponding index values generated by the DDPG model. The driver in NGSIM data produces frequent acceleration changes and large Jerk values during driving, while the DDPG model can maintain near constant acceleration and produce low Jerk values.
Based on the above, the proposed following speed control algorithm shows better safe, efficient and comfortable driving ability compared to the human driver in the NGSIM.
Claims (1)
1. A multi-objective optimized follow-up algorithm for controlling the speed of an automatically driven vehicle is characterized by comprising the following steps:
step 1: acquiring data;
using data in an NGSIM project, extracting a following event based on the criterion that a front vehicle and a rear vehicle stay on the same lane and the length of the vehicle following event is greater than 15 seconds, and taking one part as training data and the other part as test data based on the extracted following event;
step 2: constructing a reward function;
providing characteristic quantities reflecting related targets of automobile follow-up control, wherein the characteristic quantities specifically comprise safety, comfort and efficiency;
step 2.1: adopting TTC to reflect safety;
TTC is the time to collision, representing the amount of time remaining before two vehicles collide, and is formulated asWherein Sn-1, n (t) distance between cars, Δ Vn-1, n (t) relative velocity; determining the safety threshold value to be 7 seconds according to NGSIM empirical data, and performing TTC feature construction:if TTC is less than 7 seconds, the TTC characteristic index is a negative value, and as the TTC approaches zero, the TTC characteristic approaches negative infinity, and for approaching collisionThe situation (2) represents the most severe penalty;
step 2.2: measuring the driving efficiency by adopting headway ();
headway is the Headway time; from the analysis, the lognormal distribution is adapted to the distribution of the acquired training data with a probability density function ofx>0; from the extracted data, it can be estimated that the mean μ and the logarithmic standard deviation σ of the distribution variable x are 0.4226 and 0.4365, respectively; and constructing the headway characteristics into the probability density value of the estimated headway lognormal distribution: fheadway ═ flognormal (headway | μ ═ 0.4226, σ ═ 0.4365); according to the train head time characteristic, a train head time distance of about 1.3 seconds corresponds to a high characteristic value, and an overlong train head time distance or an overlong train head time distance corresponds to a low characteristic value, so that the characteristic value estimates a high-flow train head distance keeping behavior, and punishs an unsafe or overlong train head distance keeping behavior;
step 2.3: the driving comfort is measured by adopting the change rate Jerk of the acceleration, and the characteristic is as follows:
step 2.4: establishing a comprehensive reward function;
establishing r ═ w1FTTC + w2Fheadway + w3fjerk according to the above steps, where w1, w2, w3 are coefficients of features, all set to 1;
and step 3: training a model;
in each training, sequentially simulating the following events in the data, repeating the training for many times, and selecting a model which obtains the maximum average reward on the test data as a final model;
and 4, step 4: evaluating the model;
and comparing and evaluating the NGSIM data and the following behaviors obtained by DDPG model simulation by using TTC, headway and jerk indexes.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811600366.7A CN109709956B (en) | 2018-12-26 | 2018-12-26 | Multi-objective optimized following algorithm for controlling speed of automatic driving vehicle |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811600366.7A CN109709956B (en) | 2018-12-26 | 2018-12-26 | Multi-objective optimized following algorithm for controlling speed of automatic driving vehicle |
Publications (2)
Publication Number | Publication Date |
---|---|
CN109709956A CN109709956A (en) | 2019-05-03 |
CN109709956B true CN109709956B (en) | 2021-06-08 |
Family
ID=66258357
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201811600366.7A Active CN109709956B (en) | 2018-12-26 | 2018-12-26 | Multi-objective optimized following algorithm for controlling speed of automatic driving vehicle |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109709956B (en) |
Families Citing this family (18)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110321605A (en) * | 2019-06-19 | 2019-10-11 | 中汽研(天津)汽车工程研究院有限公司 | A kind of human-computer interaction coordination control strategy based on Multiple Velocity Model PREDICTIVE CONTROL |
CN110347043B (en) * | 2019-07-15 | 2023-03-10 | 武汉天喻信息产业股份有限公司 | Intelligent driving control method and device |
CN110488802B (en) * | 2019-08-21 | 2020-05-12 | 清华大学 | Decision-making method for dynamic behaviors of automatic driving vehicle in internet environment |
CN110716562A (en) * | 2019-09-25 | 2020-01-21 | 南京航空航天大学 | Decision-making method for multi-lane driving of unmanned vehicle based on reinforcement learning |
CN110992676B (en) * | 2019-10-15 | 2021-06-04 | 同济大学 | Road traffic capacity and internet automatic driving vehicle equivalent coefficient estimation method |
JP6970156B2 (en) * | 2019-10-18 | 2021-11-24 | トヨタ自動車株式会社 | Data generation method used for vehicle control, vehicle control device, vehicle control system, in-vehicle device and vehicle learning device |
JP6744598B1 (en) * | 2019-10-18 | 2020-08-19 | トヨタ自動車株式会社 | Vehicle control system, vehicle control device, and vehicle learning device |
CN112698578B (en) * | 2019-10-22 | 2023-11-14 | 北京车和家信息技术有限公司 | Training method of automatic driving model and related equipment |
CN110843746B (en) * | 2019-11-28 | 2022-06-14 | 的卢技术有限公司 | Anti-lock brake control method and system based on reinforcement learning |
DE102020201931A1 (en) * | 2020-02-17 | 2021-08-19 | Psa Automobiles Sa | Method for training at least one algorithm for a control unit of a motor vehicle, method for optimizing a traffic flow in a region, computer program product and motor vehicle |
CN112201069B (en) * | 2020-09-25 | 2021-10-29 | 厦门大学 | Deep reinforcement learning-based method for constructing longitudinal following behavior model of driver |
CN112614344B (en) * | 2020-12-14 | 2022-03-29 | 中汽研汽车试验场股份有限公司 | Hybrid traffic system efficiency evaluation method for automatic driving automobile participation |
CN113353102B (en) * | 2021-07-08 | 2022-11-25 | 重庆大学 | Unprotected left-turn driving control method based on deep reinforcement learning |
CN113954865B (en) * | 2021-09-22 | 2023-11-10 | 吉林大学 | Following control method for automatic driving vehicle in ice and snow environment |
CN113901718A (en) * | 2021-10-11 | 2022-01-07 | 长安大学 | Deep reinforcement learning-based driving collision avoidance optimization method in following state |
CN113954874B (en) * | 2021-11-03 | 2023-04-28 | 同济大学 | Automatic driving control method based on improved intelligent driver model |
CN114056332B (en) * | 2022-01-14 | 2022-04-12 | 清华大学 | Intelligent automobile following decision and control method based on cognitive risk balance |
CN115123159A (en) * | 2022-06-27 | 2022-09-30 | 重庆邮电大学 | AEB control method and system based on DDPG deep reinforcement learning |
Citations (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101606794A (en) * | 2009-07-17 | 2009-12-23 | 梁秀芬 | A kind of dynamic cinema seat equipment |
CN102955884A (en) * | 2012-11-23 | 2013-03-06 | 同济大学 | Safety distance calibration method in full-speed areas during following operation of high-speed train |
CN103101559A (en) * | 2013-02-16 | 2013-05-15 | 同济大学 | Full-speed field train interval real-time control method based on car-following behavior quality evaluation |
CN103248545A (en) * | 2013-05-28 | 2013-08-14 | 北京和利时电机技术有限公司 | Ethernetcommunication method and system for special effect broadcast system of dynamic cinema |
CN105654779A (en) * | 2016-02-03 | 2016-06-08 | 北京工业大学 | Expressway construction area traffic flow coordination control method based on vehicle-road and vehicle-vehicle communication |
CN106926844A (en) * | 2017-03-27 | 2017-07-07 | 西南交通大学 | A kind of dynamic auto driving lane-change method for planning track based on real time environment information |
CN108313054A (en) * | 2018-01-05 | 2018-07-24 | 北京智行者科技有限公司 | The autonomous lane-change decision-making technique of automatic Pilot and device and automatic driving vehicle |
CN108387242A (en) * | 2018-02-07 | 2018-08-10 | 西南交通大学 | Automatic Pilot lane-change prepares and executes integrated method for planning track |
CN108492398A (en) * | 2018-02-08 | 2018-09-04 | 同济大学 | The method for early warning that drive automatically behavior based on accelerometer actively acquires |
CN108932840A (en) * | 2018-07-17 | 2018-12-04 | 北京理工大学 | Automatic driving vehicle urban intersection passing method based on intensified learning |
Family Cites Families (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20100185369A1 (en) * | 2009-01-19 | 2010-07-22 | Jung-Woong Choi | Automatic transmission |
-
2018
- 2018-12-26 CN CN201811600366.7A patent/CN109709956B/en active Active
Patent Citations (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101606794A (en) * | 2009-07-17 | 2009-12-23 | 梁秀芬 | A kind of dynamic cinema seat equipment |
CN102955884A (en) * | 2012-11-23 | 2013-03-06 | 同济大学 | Safety distance calibration method in full-speed areas during following operation of high-speed train |
CN103101559A (en) * | 2013-02-16 | 2013-05-15 | 同济大学 | Full-speed field train interval real-time control method based on car-following behavior quality evaluation |
CN103248545A (en) * | 2013-05-28 | 2013-08-14 | 北京和利时电机技术有限公司 | Ethernetcommunication method and system for special effect broadcast system of dynamic cinema |
CN105654779A (en) * | 2016-02-03 | 2016-06-08 | 北京工业大学 | Expressway construction area traffic flow coordination control method based on vehicle-road and vehicle-vehicle communication |
CN106926844A (en) * | 2017-03-27 | 2017-07-07 | 西南交通大学 | A kind of dynamic auto driving lane-change method for planning track based on real time environment information |
CN108313054A (en) * | 2018-01-05 | 2018-07-24 | 北京智行者科技有限公司 | The autonomous lane-change decision-making technique of automatic Pilot and device and automatic driving vehicle |
CN108387242A (en) * | 2018-02-07 | 2018-08-10 | 西南交通大学 | Automatic Pilot lane-change prepares and executes integrated method for planning track |
CN108492398A (en) * | 2018-02-08 | 2018-09-04 | 同济大学 | The method for early warning that drive automatically behavior based on accelerometer actively acquires |
CN108932840A (en) * | 2018-07-17 | 2018-12-04 | 北京理工大学 | Automatic driving vehicle urban intersection passing method based on intensified learning |
Non-Patent Citations (5)
Title |
---|
Drivers’ rear end collision avoidance behaviors under different levels of situational urgency;Xuesong Wang,等;《Transportation Research Part C》;20161223;第71卷;第419-433页 * |
Modeling car-following behavior on urban expressways in Shanghai: A naturalistic driving study;Meixin Zhu,等;《Transportation Research Part C》;20180831;第93卷;第425-445页 * |
中美两国道路交通事故信息采集技术比较研究;王雪松,等;《中国安全科学学报》;20121031;第22卷(第10期);第79-87页 * |
基于自然驾驶数据的避撞预警对跟车行为影响;王雪松,朱美新,邢祎伦;《同济大学学报(自然科学版)》;20160731;第44卷(第7期);第1045-1051页 * |
驾驶员前向避撞行为特征的降维及多元方差分析;王雪松,朱美新,陈铭;《同济大学学报(自然科学版)》;20161231;第44卷(第12期);第1858-1866页 * |
Also Published As
Publication number | Publication date |
---|---|
CN109709956A (en) | 2019-05-03 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109709956B (en) | Multi-objective optimized following algorithm for controlling speed of automatic driving vehicle | |
CN109733415B (en) | Anthropomorphic automatic driving and following model based on deep reinforcement learning | |
CN106874597B (en) | highway overtaking behavior decision method applied to automatic driving vehicle | |
CN108919795B (en) | Automatic driving automobile lane change decision method and device | |
CN107169567A (en) | The generation method and device of a kind of decision networks model for Vehicular automatic driving | |
CN107168303A (en) | A kind of automatic Pilot method and device of automobile | |
CN106407563A (en) | A car following model generating method based on driving types and preceding vehicle acceleration speed information | |
CN110686906B (en) | Automatic driving test method and device for vehicle | |
CN112703459A (en) | Iterative generation of confrontational scenarios | |
CN113044064B (en) | Vehicle self-adaptive automatic driving decision method and system based on meta reinforcement learning | |
Bolduc et al. | Multimodel approach to personalized autonomous adaptive cruise control | |
CN112784485B (en) | Automatic driving key scene generation method based on reinforcement learning | |
CN113010967A (en) | Intelligent automobile in-loop simulation test method based on mixed traffic flow model | |
CN111824169B (en) | Method for reducing exhaust emissions of a drive train of a vehicle having an internal combustion engine | |
Wang et al. | High-level decision making for automated highway driving via behavior cloning | |
CN114511999A (en) | Pedestrian behavior prediction method and device | |
CN111159832A (en) | Construction method and device of traffic information flow | |
CN117242438A (en) | Method for testing a driver assistance system of a vehicle | |
Koenig et al. | Bridging the gap between open loop tests and statistical validation for highly automated driving | |
Su et al. | A traffic simulation model with interactive drivers and high-fidelity car dynamics | |
CN114179830A (en) | Autonomous overtaking method and system for automatic driving vehicle | |
Wei et al. | A learning-based autonomous driver: emulate human driver's intelligence in low-speed car following | |
CN115176297A (en) | Method for training at least one algorithm for a control unit of a motor vehicle, computer program product and motor vehicle | |
CN116629114A (en) | Multi-agent model training method, system, computer equipment and storage medium | |
Cacciabue et al. | Unified Driver Model simulation and its application to the automotive, rail and maritime domains |
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 | ||
GR01 | Patent grant | ||
GR01 | Patent grant |