An unmanned contactless logistics distribution automatic driving planning control strategy

Abstract

The application discloses an unmanned contactless logistics distribution automatic driving planning control strategy and belongs to the technical field of unmanned driving, which comprises a vehicle-mounted computer, a planning module installed in the interior of the vehicle-mounted computer and used for automatic driving planning, a control module installed in the interior of the vehicle-mounted computer and used for controlling the vehicle, and a decision module arranged in the interior of the vehicle-mounted computer and used for performing processing on the vehicle. Dynamic traffic information provided by a high-precision map is combined to adjust the driving path of the vehicle plan of the system in real time, and finally the best path to the destination is found. The decision technology of the automatic driving system is also based on deep learning. By processing data provided by two modules of high-precision map processing and prediction, the planning module can be continuously improved and perfected, so that the whole intelligent driving system is more reliable. Even in the real complex and changeable traffic, the system can make reasonable judgments and optimal planning paths.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of unmanned driving, and particularly relates to an unmanned contact logistics distribution automatic driving planning control strategy. BACKGROUND

[0002] At the end of 2019, 5 million large trucks were used for trunk transportation within a 500 km radius; 10 million trucks were used for regional transportation within a 50 km radius; and in the capillary logistics field within a 5 km radius, 30 million microcars, three-wheeled and two-wheeled vehicles were serving the e-commerce logistics and take-out markets, and the market prospect was broad. At present, many domestic Internet companies are developing research and development of park automatic driving solutions. In April 2020, the unmanned distribution vehicle of Jingdong Logistics played an important role in the logistics link during the epidemic period. The technical basis of the light automatic driving logistics vehicle product is based on automatic driving technology, intelligent logistics, vehicle networking and other emerging technologies.

[0003] However, the current logistics distribution only uses the related automatic driving, and the planning control in the automatic driving logistics distribution is not perfect, so as to easily cause congestion in logistics distribution. The application of the contactless distribution electric vehicle can effectively reduce the transportation cost of the logistics industry and improve the work efficiency, but the market still lacks benchmark products, and therefore the unmanned contact logistics distribution automatic driving planning control strategy is proposed. SUMMARY

[0004] The purpose of the present application is to provide an unmanned contact logistics distribution automatic driving planning control strategy to solve the problems in the background art.

[0005] To achieve the above purpose, the present application provides the following technical solution: an unmanned contact logistics distribution automatic driving planning control strategy, comprising a vehicle-mounted computer;

[0006] a planning module installed in the interior of the vehicle-mounted computer for automatic driving planning;

[0007] a control module installed in the interior of the vehicle-mounted computer for controlling the vehicle;

[0008] a decision module arranged in the interior of the vehicle-mounted computer for performing processing on the vehicle.

[0009] Further, the vehicle-mounted computer further comprises fastening bolts installed around the top of the vehicle-mounted computer, the top of the vehicle-mounted computer is provided with a heat dissipation net, and the surface of the vehicle-mounted computer is provided with a plurality of connection ports.

[0010] Further, the decision module further comprises a perception module and a prediction module for predicting and perceiving the trajectory of the vehicle.

[0011] Further, the vehicle-mounted computer further comprises a routing module, a positioning module and a high-precision map arranged in the vehicle-mounted computer, the positioning module and the high-precision map are capable of sending position information to the routing module and the control module, the control module comprises lateral control and longitudinal control, the planning module further comprises internal trajectory planning and speed planning.

[0012] Further, the planning module will calculate a comfortable and safe path for the control module by combining the navigation map, the route, the current position, the basic state information of the vehicle and the obstacle information.

[0013] Further, the following steps are included:

[0014] Step one: the planning module first initializes the planning module from the main function MAIN function entrance, obtains high-precision map information, and then executes the initialization state of the adapter manager object in the initialization function, and then checks the initialization state of the positioning adapter, the chassis adapter, the route search response adapter and the route search request adapter in the adapter manager, if the prediction is enabled, the initialization state of the prediction adapter is checked, if the traffic signal is enabled, the initialization state of the traffic signal adapter is checked;

[0015] Step two: next, the initialization of the candidate path is executed, and then the creation function of the RTKReplay Planner and the EMPlanner object is registered to the planner factory class object planner_factory_ by calling Planning:RegisterPlanners, in addition, the required planner object is dynamically generated by reading the planner type given in the configuration file using planner_factory_.CreateObject(config_.planner_type), the EMPlanner object is actually generated, the EM algorithm is used, and finally the initialization of the planner object is executed;

[0016] Step three: the function actually executing the planning is the RunOnce function to execute specific work. The function first calls the related functions of theAdapterManager class to obtain the positioning and chassis information required for planning; then based on the above information, the ComputeStitchingTrajectory function is called to calculate the stitching trajectory, if the following conditions are met, we will use the current position of the vehicle itself to plan:

[0017] 1. The automatic driving mode is turned off;

[0018] 2. Or we didn't get the trajectory from the last planning period;

[0019] 3. Or the position is too far from the actual and target;

[0020] Step four: these cases will call the ComputeReinitStitchingTrajectory function, and the vehicle's own x, y, z, angle, curvature, linear velocity, linear acceleration as the starting point of planning, but if the own position is used as the starting point, it will cause the trajectory to be discontinuous, and the own position is generally used as the projection point of the last trajectory as the starting point of planning;

[0021] Step five: then call the Planning: InitFrame function to initialize the planning frame; if everything is normal, call the Planning: Plan function to execute the actual planning, call the Plan function of the planner object, the current implementation is to call the plan function in the EM planning algorithm, and the decision obtains a set of available reference line information; then call the Frame: FindDriveReferenceLineInfo function to obtain the optimal reference line information: best_reference_line from the available reference line information set; the FindDriveReferenceLineInfo function finds the reference line information with the lowest value from the available reference line information list as the final driving path

[0022] Step six: the core algorithm of the whole function is to call the optimizer in the task class object array one by one for optimal estimation according to the current frame and reference line information, calculate the optimal path and speed of the current reference line;

[0023] Step seven: finally, call the Planning: PublishPlanningPb function to publish the calculated planning path to the control module, and the control module executes related acceleration, deceleration, steering and other actions.

[0024] Compared with the prior art, the beneficial effects of the present application are:

[0025] 1、The present application provides dynamic traffic information by combining high-precision maps, and adjusts the planned driving path of the vehicle in real time to finally find the best path to the destination. The decision-making technology of the automatic driving system is also based on deep learning, which processes the data provided by the high-precision map and prediction modules, and the planning module can be continuously improved and perfected, so that the whole intelligent driving system is more reliable, and even in the complex and variable traffic in reality, the system can make reasonable judgments and optimal planning paths;

[0026] 2, The application combines the navigation map, route, current position, basic state information of the vehicle (such as vehicle positioning, speed, acceleration, whole vehicle chassis CANBUS information), obstacle information, and prediction information of traffic participants, and a comfortable and safe path is calculated by the planning module to the control module, the trajectory issued by the planning module is a local trajectory, not a navigation path. First, the routing module needs to plan an optimal global navigation route according to the current vehicle information and traffic information, and then the planning module calculates a local planning route according to the navigation route;

[0027] 3, The planning module of the application can help route planning, provide selection of different routes, help the vehicle find suitable driving space, and also help identify the exact center line of the lane, so that the vehicle drives along the center line. BRIEF DESCRIPTION OF DRAWINGS

[0028] Figure 1 is the overall structure schematic diagram of the application;

[0029] Figure 2 is the internal structure schematic diagram of the vehicle-mounted computer of the application;

[0030] Figure 3 is the structure schematic diagram of the planning module of the application;

[0031] Figure 4 is the flow structure schematic diagram of the application;

[0032] Figure 5 is the planning control flow structure schematic diagram of the application.

[0033] In the figure: 1, vehicle-mounted computer; 2, fastening bolt; 3, connection port; 4, heat dissipation net; 5, planning module; 6, decision module; 7, routing module; 8, control module; 9, perception module; 10, prediction module; 11, positioning module; 12, high-precision map. DETAILED DESCRIPTION

[0034] The application will be further described below in combination with examples.

[0035] The following examples are used to illustrate the application, but cannot be used to limit the protection scope of the application. The conditions in the examples can be further adjusted according to specific conditions, and simple improvements of the method of the application under the concept of the application also belong to the protection scope of the application.

[0036] Please refer to Figures 1-5 , the application provides an unmanned contact logistics distribution automatic driving planning control strategy, including a vehicle-mounted computer 1;

[0037] A planning module 5 is installed in the vehicle computer 1 for automatic driving planning;

[0038] A control module 8 is installed in the vehicle computer 1 for vehicle control;

[0039] A decision module 6 is installed in the vehicle computer 1 for vehicle execution processing.

[0040] The vehicle computer 1 further comprises fastening bolts 2 installed around the top of the vehicle computer 1, so that the vehicle computer 1 can be disassembled and repaired. The top of the vehicle computer 1 is provided with a heat dissipation net 4, so that the heat generated by the vehicle computer 1 during operation can be dissipated through the heat dissipation net 4. The surface of the vehicle computer 1 is provided with a plurality of connection ports 3, so as to be connected with the vehicle computer 1 through the connection ports 3.

[0041] The decision module 6 further comprises a perception module 9 and a prediction module 10 for predicting and perceiving the trajectory of the vehicle.

[0042] The vehicle computer 1 further comprises a routing module 7, a positioning module 11 and a high-precision map 12 installed in the vehicle computer 1. The positioning module 11 and the high-precision map 12 can send position information to the routing module 7 and the control module 8. The control module 8 comprises lateral control and longitudinal control. The planning module 5 further comprises internal trajectory planning and speed planning.

[0043] The planning module 5 calculates a comfortable and safe path for the control module 8 by combining the navigation map, route, current position, basic state information of the vehicle and obstacle information. The path planning of the automatic driving system is based on the traditional static path planning, and combines the dynamic traffic information provided by the high-precision map to adjust the driving path planned by the system in real time, and finally finds the best path to the destination. The decision technology of the automatic driving system is also based on deep learning. By processing the data provided by the high-precision map and the prediction module, the planning module can continuously improve and perfect, so that the whole intelligent driving system is more reliable. Even in the real complex and changeable traffic, the system can make reasonable judgment and optimal planning path.

[0044] The method comprises the following steps:

[0045] Step one: the planning module first from the main function MAIN function entrance, first planning module initialization, get high-precision map information, in the initialization function inside, then execute the initialization state of the adapter manager object, then check the initialization state of the positioning adapter, chassis adapter, route finding response adapter, route finding request adapter in the adapter manager, if the prediction is enabled, continue to check the initialization state of the prediction adapter, if the traffic light is enabled, continue to check the initialization state of the traffic light adapter;

[0046] Step two: next, the initialization of the candidate path is executed, and then the creation functions of the RTKReplay Planner and EMPlanner objects are registered to the planner factory class object planner_factory_ by calling Planning:RegisterPlanners. In addition, by reading the planner type given in the configuration file, the required planner object is dynamically generated using planner_factory_.CreateObject(config_.planner_type), and the EMPlanner object is actually generated using the EM algorithm. Finally, the initialization of the planner object is executed.

[0047] Step three: the function that actually executes the planning is the RunOnce function, which performs specific work. This function first calls the relevant functions of theAdapterManager class to obtain the positioning and chassis information required for planning. Then, based on the above information, the ComputeStitchingTrajectory function is called to calculate the stitching trajectory. If the following conditions are met, we will use the current position of the vehicle itself to plan:

[0048] 1. The autonomous driving mode is off;

[0049] 2. Or we didn't get the trajectory from the last planning period;

[0050] 3. Or the position deviates too far from the actual and target;

[0051] Step four: these conditions will call the ComputeReinitStitchingTrajectory function to use the vehicle's own x, y, z, angle, curvature, linear velocity, and linear acceleration as the starting point for planning. However, if the own position is used as the starting point, it will cause the trajectory to be discontinuous. Generally, the own position is used as the starting point for planning at the projection point of the last trajectory.

[0052] Step five: After that, the Planning: InitFrame function is called to initialize the planning frame; if everything is normal, the Planning: Plan function is called to perform the actual planning. The Plan function of the planner object is called, and the current implementation is to call the plan function in the EM planning algorithm to obtain the available reference line information set (may be more than one); then the Frame: FindDriveReferenceLineInfo function is called to obtain the optimal reference line information: best_reference_line from the available reference line information set. The FindDriveReferenceLineInfo function finds the reference line information with the lowest cost value from the list of available reference line information as the final driving path;

[0053] Step six: The core algorithm of the entire function follows, which is to call the optimizers (traffic decision maker, dynamic path optimization, path decision maker, dynamic speed optimization, speed decision maker, and secondary optimization) in the task class object array one by one to perform optimal estimation for the current frame and reference line information, and to calculate the optimal path and speed of the current reference line.

[0054] Step seven: Finally, the Planning: PublishPlanningPb function is called to publish the calculated planning path to the control module. The control module performs related acceleration, deceleration, steering, and other actions.

[0055] Although embodiments of the present application have been shown and described, it is to be understood that various modifications, substitutions, replacements, and variations can be made to these embodiments without departing from the principles and spirit of the present application, and the scope of the present application is defined by the appended claims and their equivalents.

Claims

1. A planning and control strategy for unmanned contactless logistics delivery autonomous driving, characterized in that: Including the vehicle-mounted computer (1); The planning module (5) is installed inside the vehicle computer (1) for autonomous driving planning; The control module (8) is installed inside the vehicle computer (1) to control the vehicle; The decision module (6) is located inside the vehicle computer (1) and is used to perform vehicle processing. The vehicle computer (1) also includes fastening bolts (2) installed around the top of the vehicle computer (1), a heat dissipation mesh (4) is installed on the top of the vehicle computer (1), and multiple connection ports (3) are provided on the surface of the vehicle computer (1). The decision module (6) also includes a perception module (9) and a prediction module (10) for predicting and perceiving the trajectory of the vehicle; The vehicle computer (1) also includes a routing module (7), a positioning module (11), and a high-precision map (12) installed inside the vehicle computer (1). The positioning module (11) and the high-precision map (12) can send location information to the routing module (7) and the control module (8). The control module (8) includes lateral control and longitudinal control. The planning module (5) also includes internal trajectory planning and speed planning. The planning module (5) calculates a comfortable and safe path for the control module (8) by combining the navigation map, route, current location, basic vehicle status information and obstacle information.

2. The autonomous driving planning and control strategy for unmanned contactless logistics delivery according to claim 1, characterized in that: Includes the following steps: Step 1: The planning module starts from the main function MAIN and first initializes itself by obtaining high-precision map information. Inside the initialization function, it executes the initialization status of the adapter manager object. Then, it checks the initialization status of the positioning adapter, chassis adapter, routing response adapter, and routing request adapter in the adapter manager. If prediction is enabled, it continues to check the initialization status of the prediction adapter. If traffic lights are enabled, it continues to check the initialization status of the traffic light adapter. Step 2: Initialize the candidate paths. Then, call Planning:RegisterPlanners to register the creation functions of RTKReplayPlanner and EMPlanner objects to the planner factory class object planner_factory_. In addition, by reading the planner type given in the configuration file, use planner_factory_.CreateObject(config_.planner_type) to dynamically generate the required planner objects. The actual EM Planner object is generated using the EM algorithm. Finally, the planner object initialization is performed. Step 3: The actual function executing the planning is the RunOnce function, which performs the specific tasks. This function first calls the relevant functions of the AdapterManager class to obtain the positioning and chassis information required for planning; then, based on the above information, it calls the ComputeStitchingTrajectory function to calculate the bonding trajectory. We will use the vehicle's current position for planning in the following situations:

1. Autopilot mode is off; 2. Or the trajectory was not obtained from the previous planning cycle; 3. Or the location deviates too far from the actual target; Step 4: In these cases, the ComputeReinitStitchingTrajectory function will be called, using the vehicle's own x, y, z, angle, curvature, linear velocity, and linear acceleration as the starting point for planning. However, if the vehicle's own position is used as the starting point, it will cause the trajectory to be discontinuous. Generally, the projection point of the vehicle's own position on the previous trajectory is used as the starting point for planning. Step 5: Then call the Planning:InitFrame function to initialize the planning frame; If everything is normal, the `Planning:Plan` function is called to execute the actual planning. This calls the `Plan` function of the planner object, which in the current implementation calls the `plan` function in the EM planning algorithm. The decision is made to obtain a set of available reference lines, which is not unique. Then, the `Frame:FindDriveReferenceLineInfo` function is called to obtain the best reference line from the set of available reference lines: `best_reference_line`. The `FindDriveReferenceLineInfo` function finds the reference line with the lowest cost from the list of available reference lines and uses it as the final driving path. Step Six: The next step is the core algorithm of the entire function. For the current frame and reference line information, the optimizer in the task class object array is called one by one to perform optimal estimation and calculate the optimal path and speed of the current reference line. Step 7: Finally, call the `Planning:PublishPlanningPb` function to publish the calculated planned path to the control module. The control module then executes the relevant acceleration, deceleration, steering, and other actions.

Images