Multi-lane track optimization method for intelligent networked automobile
Through the multi-lane trajectory optimization method of intelligent connected vehicles, combined with Pontriajin maximum value and reinforcement learning algorithm, the traffic efficiency and fuel economy problems at urban intersections are solved, and the traffic efficiency and environmental protection and economicality of intersections are improved. Improvement.
Patent Information
- Application Number
- CN202410997960.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-24
- Publication Date
- 2024-11-29
AI Technical Summary
Traffic at urban intersections is complex, vehicle conflicts and delays are serious, and it is difficult for the existing technology to effectively optimize vehicle traffic efficiency and fuel economy.
Design a multi-lane trajectory optimization method for intelligent connected vehicles, define the state and constraints of intelligent connected vehicles through multi-lane spatio-temporal position relationship, build a composite optimization model for traffic efficiency and fuel economy, and use Pontriajin maximum value algorithm and Reinforcement learning Q-learning algorithm to obtain the optimal lane change strategy.
While ensuring vehicle safety, it can quickly improve the traffic efficiency of intersections, improve environmental protection and economy, and shorten the waiting time of vehicles.
Smart Images

Figure CN119049272A_ABST
Abstract
Description
Technical Field
[0001] The invention relates to a multi-lane trajectory optimization method for an intelligent network-connected vehicle, and belongs to the technical field of intelligent network-connected vehicles. Background Art
[0002] As the bottleneck of the urban transportation system, intersections have complex traffic conditions, and vehicle conflicts, delays, and fuel consumption are becoming more and more serious. Therefore, more and more scholars are conducting different optimization control studies on intersections. New technologies and new methods are gradually integrated with traffic problems, which has helped the birth of new solutions to urban traffic problems.
[0003] In recent years, with the continuous development and improvement of technologies such as 5G, Internet of Vehicles, and big data, intelligent connected vehicles are gradually entering people's lives and will gradually become the main mode of travel in the future.
[0004] Intelligent Connected Vehicle (ICV) is equipped with advanced sensors, controllers, actuators and other devices, and integrates wireless communication technology, which can not only realize communication between vehicle units, but also realize communication between vehicle units and road test units.
[0005] At present, vehicles can perform functions such as interaction with the environment and autonomous driving under designed conditions. Vehicles can interact with roadside units to share information such as their own position and speed, and realize information sharing with intersection lights, which makes it possible to guide and control vehicles with reasonable strategies and optimize their running trajectories. Summary of the invention
[0006] The present invention designs and develops a multi-lane trajectory optimization method for intelligent connected vehicles. The state and constraints of the intelligent connected vehicles are defined through the spatiotemporal position relationship of multiple lanes, a spatiotemporal trajectory composite optimization model for traffic economic efficiency and fuel economy is constructed and solved, collaborative lane changing rules are set, and the optimal lane changing strategy is obtained, thereby improving the traffic efficiency at intersections and improving environmental protection and economy.
[0007] The technical solution provided by the present invention is:
[0008] A multi-lane trajectory optimization method for an intelligent connected vehicle, comprising:
[0009] Step 1: Establish the state and constraint conditions of intelligent connected vehicles based on the spatiotemporal position relationship of multiple lanes;
[0010] Step 2: Based on traffic efficiency and fuel economy, a spatiotemporal trajectory composite optimization model is constructed and solved using the Pontryagin maximum algorithm;
[0011] Step 3: Set the coordinated lane-changing rules for multiple lanes, and obtain the optimal spatiotemporal trajectory of the intelligent connected vehicle in a multi-lane environment based on the reinforcement learning Q-learning algorithm.
[0012] Preferably, the intelligent connected vehicle system state expression is:
[0013]
[0014] In the formula, the number of the nth vehicle entering the mth lane from the outside to the inside is C. mn , the state is x mn (t), x mn (t) = [x mn (t),y mn (t),v mn (t),v mn '(t)] T ;
[0015] In the formula, x mn (t), y mn (t) is vehicle C mn The coordinates relative to the intersection at time t, v mn (t) is vehicle C mn The driving speed at time t, v mn '(t) is vehicle C mn The lateral velocity at time t;
[0016] The input of spatiotemporal trajectory generation is vehicle C mn The acceleration u at time t mn (t):
[0017] u mn (t)=[u mn (t),u mn '(t)] T ;
[0018] In the formula, u mn (t) is vehicle C mn The longitudinal acceleration at time t, u mn '(t) is vehicle C mn The lateral acceleration at time t;
[0019] Set the vehicle target lane signal timing information And the traffic flow information I of adjacent lanes m+1 and m-1 m+1 (t), I m-1 (t);
[0020]
[0021] In the formula, is the phase state at the current moment, for Phase remaining time, R m is the red light duration corresponding to lane m, G m is the green light duration corresponding to lane m, v m-1 (t) are the average flow speeds of lanes m+1 and m-1, respectively, and k m+1 (t), k m-1 (t) are the average traffic flow densities of lanes m+1 and m-1 respectively.
[0022] Preferably, the constraints in step 1 include: vehicle spacing constraints, speed and acceleration constraints, lateral and longitudinal impact constraints, red light signal timing constraints and communication constraints.
[0023] Preferably, the step 2 comprises:
[0024] The objective function is constructed based on the dynamic cost, and the objective function is:
[0025]
[0026] In the formula, For vehicle C mn When exiting the road, It is vehicle C mn When entering the road section, is the fixed cost of the lane-changing process, is the variable cost of the lane changing process.
[0027] Preferably, the fixed cost function is expressed as:
[0028]
[0029] In the formula, For vehicle C mn The expected speed at the exit of the road segment, is the target lane, w1 is the weight coefficient of the road section travel time, w2 is the weight coefficient of the road section travel length, w3 is the weight coefficient of the expected speed when exiting the downstream exit of the road section, w4 is the weight coefficient of exiting from the target lane, L is the road section length, d is the lane width,
[0030] The variable cost function expression is:
[0031]
[0032] Where w5 is the weight coefficient of energy change caused by vehicle longitudinal acceleration, w6 is the weight coefficient of energy change caused by vehicle lateral acceleration, and w7 is the weight coefficient of fuel consumption. + , χ(u mn (t)) is the Heaviside function of the longitudinal acceleration of the connected vehicle, is the average fuel consumption rate of the road section.
[0033] Preferably, the optimal spatiotemporal trajectory target process includes:
[0034] Combining the known position and speed information of the lane change initiating vehicle with the target lane change environment;
[0035] The coordinated lane-changing strategy is obtained based on reinforcement learning, with the vehicle system state vector as input and the vehicle acceleration set as the output of reinforcement learning. The process satisfies the Markov property.
[0036] The five-tuple (S, A, P, R, γ) is set to describe the multi-lane spatiotemporal trajectory optimization process. The lane-changing strategy with the highest current benefit is calculated by the Q-learning algorithm. The optimal lane-changing strategy expression is:
[0037]
[0038] Where S is the set of state vectors of the intelligent connected vehicle system, A is the set of lateral and longitudinal accelerations of the vehicle's optimized output, P is the set of state transfer matrices, R is the set of reward functions for the collaborative lane changing process, and γ is the discount factor for the technical value function.
[0039] The beneficial effects of the present invention are as follows: the multi-lane trajectory optimization method for intelligent connected vehicles provided by the present invention is oriented to urban roads, combined with the data reported by intelligent connected vehicles, and fully considering the time-varying characteristics of urban road networks. It can quickly and effectively realize the multi-lane spatiotemporal trajectory optimization of intelligent connected vehicles on urban roads, while ensuring vehicle safety, effectively shortening the waiting time of vehicles on the road, improving the traffic efficiency at intersections and improving environmental protection and economy. BRIEF DESCRIPTION OF THE DRAWINGS
[0040] Figure 1 This is a flow chart of the multi-lane trajectory optimization method for an intelligent connected vehicle according to the present invention. DETAILED DESCRIPTION
[0041] The present invention is further described in detail below in conjunction with the accompanying drawings so that those skilled in the art can implement the invention with reference to the description.
[0042] like Figure 1As shown, the present invention provides a multi-lane trajectory optimization method for an intelligent connected vehicle, defines the state and constraints of the intelligent connected vehicle in combination with the spatiotemporal position relationship of multiple lanes, constructs a spatiotemporal trajectory composite optimization model and solves it by comprehensively considering the traffic economic efficiency and fuel economy, sets cooperative lane changing rules, obtains the optimal lane changing strategy, improves the traffic efficiency at the intersection and improves environmental protection and economy, including:
[0043] Step 1: Establish the state and constraint conditions of intelligent connected vehicles based on the spatiotemporal position relationship of multiple lanes;
[0044] Step 2: Based on traffic efficiency and fuel economy, a spatiotemporal trajectory composite optimization model is constructed and solved using the Pontryagin maximum algorithm;
[0045] Step 3: Set the coordinated lane-changing rules for multiple lanes, and obtain the optimal spatiotemporal trajectory of the intelligent connected vehicle in a multi-lane environment based on the reinforcement learning Q-learning algorithm.
[0046] In the present invention, as a preference, a 3×3 nine-square grid road network is selected to construct an application scenario of the trajectory optimization method.
[0047] Specifically include:
[0048] Set the nth vehicle number of the mth lane entering the road section from the outside to the inside as C mn , the state is x mn (t), then x mn (t) = [x mn (t),y mn (t),v mn (t),v mn '(t)] T , its initial state is:
[0049]
[0050] In the formula, x mn (t), y mn (t) is vehicle C mn The coordinates relative to the intersection at time t, v mn (t) is vehicle C mn The driving speed at time t, v mn '(t) is vehicle C mn The lateral velocity at time t;
[0051] The input of spatiotemporal trajectory generation is vehicle C mn The acceleration u at time t is mn (t),
[0052] u mn (t)=[u mn(t),u mn '(t)] T ;
[0053] In the formula, u mn (t) is vehicle C mn The longitudinal acceleration at time t, u mn '(t) is vehicle C mn The lateral acceleration at time t;
[0054] Set the vehicle target lane signal timing information And the traffic flow information I of adjacent lanes m+1 and m-1 m+1 (t), I m-1 (t);
[0055]
[0056] In the formula, is the phase state at the current moment, for Phase remaining time, R m is the red light duration corresponding to lane m, G m is the green light duration corresponding to lane m, v m-1 (t) are the average flow speeds of lanes m+1 and m-1, respectively, and k m+1 (t), k m-1 (t) are the average traffic flow densities of lanes m+1 and m-1 respectively.
[0057] Set the intelligent connected vehicle system state expression as:
[0058]
[0059] In order to ensure the stability and high passability of traffic flow at each intersection, five types of constraints are set: vehicle spacing constraints, speed and acceleration constraints, lateral and longitudinal impact constraints, red light signal timing constraints and communication constraints.
[0060] Among them, for vehicle C in lane m mn should be in line with the vehicle in front of it C m(n-1) A certain amount of spatial displacement and temporal displacement is maintained. At the same time, in order to ensure driving safety, the lateral lane-changing distance when the vehicle changes lanes shall not be greater than the width of a lane. In the present invention, as a preferred embodiment, the lane width is set to 3.5m.
[0061] In order to ensure the safety of vehicles driving on the road section, speed constraints are imposed on vehicles on the road section. For longitudinal speed constraints, that is, traffic direction constraints, the maximum speed is the maximum speed limit on the current road section, and the minimum speed is set to 0.
[0062] The objective function is constructed by considering fixed costs and variable costs, and the current vehicle system state vector is solved based on the Pontryagin maximum principle, thereby obtaining the motion state of each intelligent connected vehicle under the current road network.
[0063] In order to ensure that the traffic flow at a single intersection has higher passability and improve the fuel consumption efficiency of vehicles, the objective function is defined as follows, considering fixed costs and variable costs:
[0064]
[0065] In the formula, For vehicle C mn When exiting the road, It is vehicle C mn When entering the road section, is the fixed cost of the lane-changing process, is the variable cost of the lane changing process.
[0066] The expression of the fixed cost function is:
[0067]
[0068] In the formula, For vehicle C mn The expected speed at the exit of the road segment, is the target lane, w1 is the weight coefficient of the road section travel time, w2 is the weight coefficient of the road section travel length, w3 is the weight coefficient of the expected speed when exiting the downstream exit of the road section, w4 is the weight coefficient of exiting from the target lane, L is the road section length, d is the lane width,
[0069] In the present invention, as a preferred embodiment, variable costs are established based on the VT-Micro model to reduce fuel consumption:
[0070] The variable cost function expression is:
[0071]
[0072] Where w5 is the weight coefficient of energy change caused by vehicle longitudinal acceleration, w6 is the weight coefficient of energy change caused by vehicle lateral acceleration, and w7 is the weight coefficient of fuel consumption. + , χ(u mn (t)) is the Heaviside function of the longitudinal acceleration of the connected vehicle, Average fuel consumption rate of the road section.
[0073] The multi-lane trajectory problem is solved by the maximum principle, and the Hamiltonian function expression of the cost function J is:
[0074] H(x,u,λ,t)=λ T f(x,u,t)+h(x,u,t);
[0075] Where λ is defined as the co-state vector of vector x, which indicates that a small change in vector x The additional cost of the change in J caused;
[0076] Within the admissible set U, the minimum value of the cost function optimizes the input u * Then it must satisfy that the Hamiltonian function is in the minimum state, and the expression is:
[0077] H(x * ,u * ,λ * ,t * )≤H(x * ,u,λ * ,t);
[0078] The Hamiltonian function for the multi-lane trajectory problem is as follows:
[0079] H mn =λ1·v mn (t)+λ2·u mn (t)+w5·(u mn (t) 2 +2u mn (t)·v mn (t))·χ(u mn (t))
[0080] The optimal spatiotemporal trajectory target process includes:
[0081] Combining the known position and speed information of the lane change initiating vehicle with the target lane change environment;
[0082] The coordinated lane-changing strategy is obtained based on reinforcement learning, with the vehicle system state vector as input and the vehicle acceleration set as the output of reinforcement learning. The process satisfies the Markov property.
[0083] The five-tuple (S, A, P, R, γ) is set to describe the multi-lane spatiotemporal trajectory optimization process. The lane-changing strategy with the highest current benefit is calculated by the Q-learning algorithm. The optimal lane-changing strategy expression is:
[0084]
[0085] Where S is the set of state vectors of the intelligent connected vehicle system, A is the set of lateral and longitudinal accelerations of the vehicle's optimized output, P is the set of state transfer matrices, R is the set of reward functions for the collaborative lane changing process, and γ is the discount factor for the technical value function.
[0086] This method fully considers the time-varying characteristics of urban road networks, proposes a method for generating spatiotemporal trajectories for intelligent connected vehicles in urban road networks, and a multi-lane spatiotemporal trajectory optimization method based on this method, which quickly and effectively realizes the multi-lane spatiotemporal trajectory optimization of intelligent connected vehicles on urban roads.
[0087] The proposed trajectory method includes the establishment of relevant intelligent connected vehicle states, constraints and the construction of objective functions based on fixed and variable costs, and a fast solution method based on the Pontryagin maximum principle. While ensuring vehicle safety, it effectively shortens vehicle loss time, improves intersection traffic efficiency and improves environmental protection and economy.
[0088] Although the embodiments of the present invention have been disclosed as above, they are not limited to the applications listed in the specification and the implementation modes, and they can be fully applied to various fields suitable for the present invention. For those familiar with the art, additional modifications can be easily implemented. Therefore, without departing from the general concept defined by the claims and the scope of equivalents, the present invention is not limited to the specific details and the illustrations shown and described herein.
Claims
1. A multi-lane trajectory optimization method for intelligent connected vehicles, It is characterized in that include: Step 1: Establish the state and constraint conditions of intelligent connected vehicles based on the spatiotemporal position relationship of multiple lanes; Step 2: construct a spatiotemporal trajectory composite optimization model based on traffic efficiency and fuel economy, and solve it using a maximum value algorithm; Step 3: Set the coordinated lane-changing rules for multiple lanes, and obtain the optimal spatiotemporal trajectory of the intelligent connected vehicle in a multi-lane environment based on the reinforcement learning algorithm.
2. The multi-lane trajectory optimization method for an intelligent connected vehicle according to claim 1, It is characterized in that The state expression of the intelligent connected vehicle system is: In the formula, the number of the nth vehicle entering the mth lane from the outside to the inside is C. mn , the state is x mn (t), x mn (t) = [x mn (t),y mn (t),v mn (t),v mn '(t)] T ; Where xmn(t) and ymn(t) are the vehicle C mn At time t, vmn(t) is the coordinate of vehicle C relative to the intersection. mn At time t, vmn'(t) is the driving speed of vehicle C. mn The lateral velocity at time t; The input of spatiotemporal trajectory generation is vehicle C mn The acceleration umn(t) at time t is: u mn (t)=[u mn (t),u mn '(t)] T ; Where umn(t) is the vehicle C mn At time t, umn'(t) is the longitudinal acceleration of vehicle C mn The lateral acceleration at time t; Set the vehicle target lane signal timing information And the traffic flow information I of adjacent lanes m+1 and m-1 m+1 (t), I m-1 (t); In the formula, is the phase state at the current moment, for Phase remaining time, R m is the red light duration corresponding to lane m, G m is the green light duration corresponding to lane m, v m-1 (t) are the average flow speeds of lanes m+1 and m-1, respectively, and k m+1 (t), k m-1 (t) are the average traffic flow densities of lanes m+1 and m-1 respectively.
3. The multi-lane trajectory optimization method for an intelligent connected vehicle according to claim 2, It is characterized in that The constraints in step 1 include: vehicle spacing constraints, speed and acceleration constraints, lateral and longitudinal impact constraints, red light signal timing constraints, and communication constraints.
4. The multi-lane trajectory optimization method for an intelligent connected vehicle according to claim 3, It is characterized in that The second step comprises: The objective function is constructed based on the dynamic cost, and the objective function is: In the formula, For vehicle C mn When exiting the road, It is vehicle C mn When entering the road section, is the fixed cost of the lane-changing process, is the variable cost of the lane changing process.
5. The multi-lane trajectory optimization method for an intelligent connected vehicle according to claim 4, It is characterized in that The expression of the fixed cost function is: In the formula, For vehicle C mn The expected speed at the exit of the road segment, is the target lane, w 1 is the weight coefficient of road section travel time, w 2 is the weight coefficient of the road section travel length, w 3 is the weight coefficient of the expected speed when exiting the downstream exit of the road section, w 4 is the weight sparseness of driving out of the target lane, L is the length of the road section, d is the lane width, The variable cost function expression is: In the formula, w 5 is the weight coefficient of energy change caused by vehicle longitudinal acceleration, w 6 is the weight coefficient of energy change caused by vehicle lateral acceleration, w 7 is the sparse weight of fuel consumption, where w 5 、w 6 、w 7 ∈R + , χ(u mn (t)) is the Heaviside function of the longitudinal acceleration of the connected vehicle, is the average fuel consumption rate of the road section.
6. The multi-lane trajectory optimization method for an intelligent connected vehicle according to claim 5, It is characterized in that The optimal spatiotemporal trajectory target process includes: Combining the known position and speed information of the lane change initiating vehicle with the target lane change environment; The coordinated lane-changing strategy is obtained based on reinforcement learning, with the vehicle system state vector as input and the vehicle acceleration set as the output of reinforcement learning. The process satisfies the Markov property. The five-tuple (S, A, P, R, γ) is set to describe the multi-lane spatiotemporal trajectory optimization process. The lane-changing strategy with the highest current benefit is calculated by the Q-learning algorithm. The optimal lane-changing strategy expression is: Where S is the set of state vectors of the intelligent connected vehicle system, A is the set of lateral and longitudinal accelerations of the vehicle's optimized output, P is the set of state transfer matrices, R is the set of reward functions for the collaborative lane changing process, and γ is the discount factor for the technical value function.
