AirSim获取无人机物理参数

追光_逐梦

已于 2025-04-19 16:33:02 修改

阅读量788

点赞数 8

文章标签：学习游戏引擎 github 无人机

于 2024-12-25 15:29:05 首次发布

本文链接：https://blog.csdn.net/weixin_46398923/article/details/144719505

版权

一、AirSim没有直接获取无人机物理参数的API

AirSim 是微软开源的一个跨平台的建立在虚幻引擎(Unreal Engine)上的无人机以及其它自主移动设备的模拟器。

AirSim为开发者提供了一定数量的API(Application Programming Interface,应用程序编程接口)以供使用者迅速快捷进行仿真实验。

然而，AirSim中没有用于获取无人机物理参数，包括质量、尺寸、惯性矩阵等的API，考虑到在相当多的仿真情境下，使用者需要获取以及修改无人机物理参数，因此本文介绍一种获取无人机物理参数的方法，与大家交流，如有错误，欢迎大家批评指正。

二、Airsim源码中各参数的位置

为了获取无人机相关物理参数，需要查阅AirSim源码中的MultiRotorParameters.hpp文件，其路径为

AirLib/include/vehicles/multirotor/MultiRotorParams.hpp

该文件中，给出了包括四旋翼、六旋翼、八旋翼无人机以及轮式机器人的相关参数，我们以四旋翼无人机为例，相关代码为

void setupFrameGenericQuad(Params& params)
        {
            //set up arm lengths
            //dimensions are for F450 frame: http://artofcircuits.com/product/quadcopter-frame-hj450-with-power-distribution
            params.rotor_count = 4;
            std::vector<real_T> arm_lengths(params.rotor_count, 0.2275f);

            //set up mass
            //this has to be between max_thrust*rotor_count/10 (1.6kg using default parameters in RotorParams.hpp) and (idle throttle percentage)*max_thrust*rotor_count/10 (0.8kg using default parameters and SimpleFlight)
            //any value above the maximum would result in the motors not being able to lift the body even at max thrust,
            //and any value below the minimum would cause the drone to fly upwards on idling throttle (50% of the max throttle if using SimpleFlight)
            //Note that the default idle throttle percentage is 50% if you are using SimpleFlight
            params.mass = 1.0f;

            real_T motor_assembly_weight = 0.055f; //weight for MT2212 motor for F450 frame
            real_T box_mass = params.mass - params.rotor_count * motor_assembly_weight;

            // using rotor_param default, but if you want to change any of the rotor_params, call calculateMaxThrust() to recompute the max_thrust
            // given new thrust coefficients, motor max_rpm and propeller diameter.
            params.rotor_params.calculateMaxThrust();

            //set up dimensions of core body box or abdomen (not including arms).
            params.body_box.x() = 0.180f;
            params.body_box.y() = 0.11f;
            params.body_box.z() = 0.040f;
            real_T rotor_z = 2.5f / 100;

            //computer rotor poses
            initializeRotorQuadX(params.rotor_poses, params.rotor_count, arm_lengths.data(), rotor_z);

            //compute inertia matrix
            computeInertiaMatrix(params.inertia, params.body_box, params.rotor_poses, box_mass, motor_assembly_weight);
        }

1、无人机质量

根据代码，四旋翼无人机的质量为1kg，并且注释中提及，四旋翼无人机的质量需要介于0.8kg（使用默认参数时的怠速油门升力）和1.6kg（使用默认参数时的最大升力）之间

2、电机质量和机身质量

四旋翼无人机每个电机的质量为0.055kg，因此机身质量为1kg-4*0.055kg，即0.78kg

3、无人机臂长

四旋翼无人机臂长为0.2275m

4、机身尺寸

机身尺寸为x方向上0.18m，y方向上0.11m，z方向上0.04m，电机平面到无人机质心的距离为0.0025m

5、四旋翼无人机布局与电机旋转方向

代码中给出四旋翼无人机布局为“X”型，注释中也给出了电机旋转方向

 protected: //static utility functions for derived classes to use
        /// Initializes 4 rotors in the usual QuadX pattern:  http://ardupilot.org/copter/_images/MOTORS_QuadX_QuadPlus.jpg
        /// which shows that given an array of 4 motors, the first is placed top right and flies counter clockwise (CCW) and
        /// the second is placed bottom left, and also flies CCW.  The third in the array is placed top left and flies clockwise (CW)
        /// while the last is placed bottom right and also flies clockwise.  This is how the 4 items in the arm_lengths and
        /// arm_angles arrays will be used.  So arm_lengths is 4 numbers (in meters) where four arm lengths, 0 is top right,
        /// 1 is bottom left, 2 is top left and 3 is bottom right.  arm_angles is 4 numbers (in degrees)  relative to forward vector (0,1),
        /// provided in the same order where 0 is top right, 1 is bottom left, 2 is top left and 3 is bottom right, so for example,
        /// the angles for a regular symmetric X pattern would be 45, 225, 315, 135.  The rotor_z is the offset of each motor upwards
        /// relative to the center of mass (in meters).
        static void initializeRotorQuadX(vector<RotorPose>& rotor_poses /* the result we are building */,
                                         uint rotor_count /* must be 4 */,
                                         real_T arm_lengths[],
                                         real_T rotor_z /* z relative to center of gravity */)
        {
            Vector3r unit_z(0, 0, -1); //NED frame
            if (rotor_count == 4) {
                rotor_poses.clear();

                /* Note: rotor_poses are built in this order:
                 x-axis
            (2)  |   (0)
                 |
            -------------- y-axis
                 |
            (1)  |   (3)
            */
                // vectors below are rotated according to NED left hand rule (so the vectors are rotated counter clockwise).
                Quaternionr quadx_rot(AngleAxisr(M_PIf / 4, unit_z));
                rotor_poses.emplace_back(VectorMath::rotateVector(Vector3r(0, arm_lengths[0], rotor_z), quadx_rot, true),
                                         unit_z,
                                         RotorTurningDirection::RotorTurningDirectionCCW);
                rotor_poses.emplace_back(VectorMath::rotateVector(Vector3r(0, -arm_lengths[1], rotor_z), quadx_rot, true),
                                         unit_z,
                                         RotorTurningDirection::RotorTurningDirectionCCW);
                rotor_poses.emplace_back(VectorMath::rotateVector(Vector3r(arm_lengths[2], 0, rotor_z), quadx_rot, true),
                                         unit_z,
                                         RotorTurningDirection::RotorTurningDirectionCW);
                rotor_poses.emplace_back(VectorMath::rotateVector(Vector3r(-arm_lengths[3], 0, rotor_z), quadx_rot, true),
                                         unit_z,
                                         RotorTurningDirection::RotorTurningDirectionCW);
            }
            else
                throw std::invalid_argument("Rotor count other than 4 is not supported by this method!");
        }

第一个电机在无人机右前方（序号0），转动方向为逆时针；第二个电机在无人机左后方（序号1），转动方向也为逆时针；第三个电机在无人机左前方（序号2），转动方向为顺时针；第四个电机在无人机右后方（序号3），转动方向为顺时针。

代码中还描述了四个电机的当前位置，是由“十”字型的位置，即电机在机体坐标系中的坐标分别为(0, arm_lengths, rotor_z)、(0, -arm_lengths, rotor_z)、(arm_lengths, 0, rotor_z)和(-arm_lengths, 0, rotor_z)，绕z轴逆时针旋转45度得到的。

6、转动惯量矩阵

代码中给出的转动惯量矩阵计算方法为

{
            inertia = Matrix3x3r::Zero();

            //http://farside.ph.utexas.edu/teaching/336k/Newtonhtml/node64.html
            inertia(0, 0) = box_mass / 12.0f * (body_box.y() * body_box.y() + body_box.z() * body_box.z());
            inertia(1, 1) = box_mass / 12.0f * (body_box.x() * body_box.x() + body_box.z() * body_box.z());
            inertia(2, 2) = box_mass / 12.0f * (body_box.x() * body_box.x() + body_box.y() * body_box.y());

            for (size_t i = 0; i < rotor_poses.size(); ++i) {
                const auto& pos = rotor_poses.at(i).position;
                inertia(0, 0) += (pos.y() * pos.y() + pos.z() * pos.z()) * motor_assembly_weight;
                inertia(1, 1) += (pos.x() * pos.x() + pos.z() * pos.z()) * motor_assembly_weight;
                inertia(2, 2) += (pos.x() * pos.x() + pos.y() * pos.y()) * motor_assembly_weight;
            }
        }

其中，总转动惯量为机身转动惯量加四个电机的转动惯量

代入上面的数据，机身转动惯量为：

I_xx = 0.78/12 * (0.11^2 + 0.04^2) = 0.0008905

I_yy = 0.78/12 * (0.18^2 + 0.04^2) = 0.00221

I_yy = 0.78/12 * (0.18^2 + 0.11^2) = 0.0028925

电机产生的转动惯量计算公式中，pos.x()、pos.y()、pos.z()分别表示电机质心到无人机x轴、y轴、z轴的距离，因此电机产生的转动惯量为：

I_xx = 4 * ( 0.2275^2 / 2 + 0.0025^2 ) * 0.055 = 0.0058307

I_yy = 4 * ( 0.2275^2 / 2 + 0.0025^2 ) * 0.055 = 0.0058307

I_zz = 4 * (0.2275^2 / 2 * 2) * 0.055 = 0.0113864

最终，计算得到的转动惯量为机体和电机产生的转动惯量之和，因此默认参数情况下，四旋翼无人机转动惯量为

[I_xx I_yy I_zz] = [0.0067212 0.0080407 0.0142789]