Distributed centralized autopilot system and method

Distributed centralized autopilot system and method

  • CN 109,855,646 B
  • Filed: 04/30/2019
  • Issued: 02/28/2020
  • Est. Priority Date: 04/30/2019
  • Status: Active Grant
First Claim
Patent Images

1. A distributed centralized automatic driving system is provided, which clearly divides a data domain and a control flow, carries out modular design on the realization of each functional system, and deploys each module to a control unit of a corresponding data domain according to the load of a computing platform, and the realization comprises the following steps:

  • step 1, defining a functional module according to the type of a system bus;

    the description of the functional module adopts an extensible markup language format, and for each functional module, the definition description comprises an interface, a performance index required to be achieved, a completion function, a communication interface, a data source and the safety and priority level of the function;

    the functional module comprises a sensor processing module, a perception positioning module, a perception target detection module, a decision-making module, a planning module and a vehicle control module;

    the sensor processing module comprises laser radar processing, visual processing, an ultrasonic radar, an inertial unit, a GPS navigation and a milemeter;

    step 2, defining a data domain, highlighting the data source type and channel for supporting the completion of the appointed function module, and outputting the data of the function module for providing for other function modules in the system;

    the data flow exchange among all the functional modules is within the corresponding data domain capacity, and the physical bearing representation form of the data domain is a physical bus interface;

    the laser radar processing is used for outputting point clouds;

    the perception target detection module is used for scene detection, target detection, road detection and traffic sign detection;

    the perception target detection module is used for inputting point clouds, camera videos and vehicle global positions and outputting target positions, road rules and road guides;

    the perception positioning module is used for inputting point cloud, GPS longitude and latitude, IMU inertial unit, odometer and high-precision map and outputting the global position of the vehicle;

    the decision module is used for inputting road guide, vehicle global position, dynamic map and static map and outputting operation decision, lane rule and the position of a target in the dynamic map address;

    the planning module comprises available space, lane planning, traffic planning, path planning and path following;

    the planning module is used for inputting the global position of the vehicle, operation decision, lane rule, address position of the target in the dynamic map and road guidance and outputting a path which can be followed by the vehicle;

    the vehicle control module is used for inputting a path which can be followed by a vehicle and the current control sensor state of the vehicle and outputting a vehicle actuator control command;

    step 3, defining control and message flow, and adopting a client/server mode or a message publishing/subscribing mode;

    step 4, calculating unit mapping, referring to data domain division, mapping the functional module to a selected specified system platform through evaluation of computing capacity requirements and performance indexes, and making a mapping decision according to the whole system resources which can be used at present;

    step 5, realizing a functional module, and dividing the functional module into a hardware dependence layer, a hardware abstraction layer and an independent software layer according to the static definition and the data domain definition of the module;

    the physical implementation form of the functional module is represented by one to a plurality of tasks which are independently or coordinately operated in an independent software layer container;

    step 6, designing a scheduling system, and completing the distribution of a plurality of computing tasks in an independent software layer of the computing unit by the scheduling system;

    the task allocation and coordination among the computing units are completed by the negotiation of the scheduling systems distributed in the computing units;

    a plurality of main scheduling units and computing units are arranged in the system, the main scheduling units master resource information of all the computing units of the system, and the main scheduling units negotiate with scheduling middleware on each computing unit according to computing unit resources and computing task requirements to complete computing task allocation;

    the computing units are distributed on different chips and different hardware systems and are coordinated by distributed computing scheduling;

    the calculation task requirements comprise data requirements, performance requirements and resource requirements, and are used for completing all sensor cooperation of the automatic driving system application, sensor capability and calculation capability of system actual configuration, and self-adaptive configuration under the condition that multiple calculation platforms coexist.

View all claims

    Thank you for your feedback