[0016] The implementation scenario of this embodiment is driving in formation: driving in a convoy, the first car is the lead vehicle and has the right to accelerate and decelerate, and the subsequent vehicles are slave cars without independent control rights, that is, they cannot perform acceleration and deceleration processing. The goal is to maintain a certain amount of synchronization between all vehicles to save vehicle fuel consumption.
[0017] Such as figure 1 As shown, the lead vehicle broadcasts the assigned data packet including the acceleration information of each subordinate vehicle through the wireless channel; the subordinate vehicle receives the assigned data packet through the wireless channel and parses out the corresponding acceleration information and controls the respective driving accordingly. At the same time, the slave car periodically updates the model parameters in the model information online, and sends the updated model information to the leader car in the same cycle for offline update of the slave car model.
[0018] When the error between the actual state information obtained from the state detection unit of the vehicle and the estimated state information estimated based on the model information is greater than the threshold, the slave vehicle sends the actual state information to the leader vehicle for analysis by the leader vehicle to obtain the corrected state information.
[0019] The state detection unit periodically detects whether the error is greater than a threshold and the period is less than the online update and sending period of the model parameters.
[0020] The lead vehicle updates the slave vehicle model offline and obtains the estimated state information of the slave vehicle by calculating the slave vehicle model until the corrected state information is received, that is, the lead vehicle replaces the estimated state information with the corrected state information as the actual state information of the slave vehicle .
[0021] The estimated state information is calculated and estimated from the current speed v and acceleration a and the distance d to the preceding vehicle, specifically: the estimated distance to the preceding vehicle at the next time is d Next , The estimated speed at the next moment is v Next , The state estimation formula is Among them: v, d are the distance and speed from the vehicle ahead at the current moment, a is the acceleration at the current moment, A and B are the parameters in the parallel model, expressed in the form of a matrix, a0~a3 are the value of each element in the 2x2 parameter matrix A, and b0~b1 are the value of each element in the 2x1 parameter matrix B.
[0022] The lead vehicle calculates the assigned acceleration based on actual state information or estimated state information, specifically: the assigned acceleration of each subordinate vehicle is weighted according to the state of the lead vehicle, the state of the preceding vehicle, and the state of the subordinate vehicle. Get, specifically: a i =w 1 ×(d Front -L)-w 2 ×(v Self -v Front )-w 3 ×(v Self -v Lead )+w×a Front +w 5 ×a Lead , Where: i is the serial number of the slave car, L is the target car distance, w 1 ~w 5 Is the weighting factor, d Front Is the distance between car i and the previous car, v Self Is the speed of the slave car i, v Front Is the speed of the preceding car from car i, v Lead Is the speed of the leader, a Front Is the acceleration of the preceding car from car i, a Lead It is the acceleration of the leading vehicle, and each weighting coefficient is set according to the actual driving situation.
[0023] The online update refers to: the slave car preferably uses the least square method to calculate the parallel model parameter matrix A and B based on the state information recorded by the slave car at each moment, and sends the parallel model parameter matrix A and B to the leader car periodically. The model parameters recover the model information of each slave car, and the estimated state information is obtained according to the model information and the current state information of the slave car. Specifically: the determination of the model parameters is an online update process, that is, the parameters will follow the system The operation is constantly updated. In the state information estimation formula , It is necessary to find the least square method to find the model parameters so that when the input state information and acceleration are determined, the error between the estimated next moment state information and the actual next moment state information is the smallest. The parameters of this model are the model parameters used. The model will be updated periodically (same as the period of sending model parameters in the next paragraph) to ensure that the model parameters can well describe the changes in state information. The specific mathematical representation of the model parameter matrix A and B calculated by the least square method is: Where v t0 …V tn Is the speed of the slave vehicle from time t0 to time tn, d t0 …D tn Is the distance between the vehicle and the preceding vehicle from time t0 to time tn, a t0 …A tn It is the acceleration of the host vehicle from time t0 to time tn. Assume Where v t1 …V tn+1 Is the speed of the slave vehicle from time t1 to time tn+1, d t1 …D tn+1 It is the distance between the own vehicle and the preceding vehicle from time t1 to time tn+1. Where A T And B T They are the updated transpose matrix of A and B, xData T-1 Is the pseudo-inverse matrix of the transposed matrix of xData, yData T Is the transposed matrix of yData.
[0024] This embodiment specifically uses two NI-USRP 2974s to run the LTE Sidelink Mode4 physical layer framework, and the fastest data transmission period is set to 10ms; each USRP runs the above method, one of the two is set as the leader, and the other is set For the slave car. At the same time, a vehicle motion simulator is constructed in two USRPs to generate data during the operation of the two vehicles. The speed of the vehicle and the distance to the vehicle ahead have a random error of 1% to simulate the sensor error.
[0025] The minimum period of the experimental setting to correct the data packet transmission is 10ms, and the period of the model parameter packet is 100ms. The error threshold is 0.1m+m/s.
[0026] The experimental results are as figure 2 As shown, the unit of abscissa of the three pictures is ms. The first ordinate is the error of the estimated state information, which is the sum of the estimated distance error and the estimated speed error, and the unit is m+m/s. The second and third chapters respectively indicate the sending time of the correction package and the status parameter update package. The current abscissa is sent at the moment, and the ordinate is 1.
[0027] First, pay attention to the high error of the first 500ms, which is caused by the two machines not being turned on at the same time, and one vehicle (USRP) is not participating in the system. Start the design process normally from 500ms.
[0028] From 500ms to 1200ms, the model parameters are being updated online, and the model cannot fit the status information well. At this time, the state information estimated by the model often has large errors. Correction packets are sent frequently to ensure the correct transmission of status information. From a numerical analysis, the maximum frequency of data packet transmission during this period is 100 packets per second, that is, an interval of 10ms. Compared with the traditional method, the data must be transmitted every 10ms interval. The number of data packets is less than or equal to the traditional method. The maximum error caused by the model has not been updated yet is within 0.32m+m/s, which is acceptable.
[0029] After 1200ms, the model was updated, and the number of correction packages was greatly reduced to almost none. Only a 100ms cycle model parameter update package is left. The transmission interval has been reduced by 1/10, and the channel occupancy has been reduced by nearly 90%.
[0030] If more radical parameter settings are used, the channel occupancy can be reduced by 99%. It may cause the error to become larger, but as long as the error is within the allowable range of the vehicle, it is feasible.
[0031] The above specific implementations can be locally adjusted by those skilled in the art in different ways without departing from the principle and purpose of the present invention. The protection scope of the present invention is subject to the claims and is not limited by the above specific implementations. All implementation schemes within the scope are bound by the present invention.