Airsim仿真双目相机时间戳不同步的解决办法

settings.json

bash 复制代码
{
  "SeeDocsAt": "https://github.com/Microsoft/AirSim/blob/main/docs/settings.md",
  "SettingsVersion": 1.2,
  "SimMode": "Multirotor",
  "ViewMode": "NoDisplay",
  "ClockSpeed": 1.0,
  "LocalHostIp": "192.168.35.220",
  "ApiServerPort": 41451,
  "Vehicles": {
    "drone_1": {
      "VehicleType": "SimpleFlight",
      "DefaultVehicleState": "Armed",
      "EnableCollisionPassthrogh": false,
      "EnableCollisions": true,
      "AllowAPIAlways": true,
      "Cameras": {
        "front_left_custom": {
          "CaptureSettings": [
            {
              "PublishToRos": 1,
              "ImageType": 0,
              "Width": 672,
              "Height": 376,
              "FOV_Degrees": 120,
              "TargetGamma": 1.5
            }
          ],
          "X": 0.50, "Y": -0.06, "Z": 0.10,
          "Pitch": 0.0, "Roll": 0.0, "Yaw": 0.0
        },
        "front_right_custom": {
          "CaptureSettings": [
            {
              "PublishToRos": 1,
              "ImageType": 0,
              "Width": 672,
              "Height": 376,
              "FOV_Degrees": 120,
              "TargetGamma": 1.5
            }
          ],
          "X": 0.50, "Y": 0.16, "Z": 0.10,
          "Pitch": 0.0, "Roll": 0.0, "Yaw": 0.0
        }
      },
      "X": 0,
      "Y": 0,
      "Z": 0,
      "Pitch": 0,
      "Roll": 0,
      "Yaw": 0
    }
  }
}

airsim_ros_pkgs/src/airsim_ros_wrapper.cpp

cpp 复制代码
void AirsimROSWrapper::img_response_timer_cb(const ros::TimerEvent& event)
{
    try {
        int image_response_idx = 0;

        uint64_t kk;

        for (const auto& airsim_img_request_vehicle_name_pair : airsim_img_request_vehicle_name_pair_vec_) {
            const std::vector<ImageResponse>& img_response = airsim_client_images_.simGetImages(airsim_img_request_vehicle_name_pair.first, airsim_img_request_vehicle_name_pair.second);
            
            // 创建 img_response 的副本,修改副本的时间戳
            std::vector<ImageResponse> img_response_copy = img_response;

            if (image_response_idx == 0){
                kk = img_response_copy[0].time_stamp;  // 获取第一个元素的时间戳
            }
            else {
                img_response_copy[0].time_stamp = kk;  // 使用保存的时间戳
            }

            if (img_response_copy.size() == airsim_img_request_vehicle_name_pair.first.size()) {
                process_and_publish_img_response(img_response_copy, image_response_idx, airsim_img_request_vehicle_name_pair.second);
                image_response_idx += img_response_copy.size();
            }

            // if (img_response.size() == airsim_img_request_vehicle_name_pair.first.size()) {
            //     process_and_publish_img_response(img_response, image_response_idx, airsim_img_request_vehicle_name_pair.second);
            //     image_response_idx += img_response.size();
            // }
        }
    }

    catch (rpc::rpc_error& e) {
        std::string msg = e.get_error().as<std::string>();
        std::cout << "Exception raised by the API, didn't get image response." << std::endl
                  << msg << std::endl;
    }
}
相关推荐
CoovallyAIHub1 小时前
OpenClaw一脚踩碎传统CV?机器终于不再只是看世界
深度学习·算法·计算机视觉
CoovallyAIHub2 小时前
仅凭单目相机实现3D锥桶定位?UNet-RKNet破解自动驾驶锥桶检测难题
深度学习·算法·计算机视觉
CoovallyAIHub5 天前
仿生学突破:SILD模型如何让无人机在电力线迷宫中发现“隐形威胁”
深度学习·算法·计算机视觉
CoovallyAIHub5 天前
从春晚机器人到零样本革命:YOLO26-Pose姿态估计实战指南
深度学习·算法·计算机视觉
CoovallyAIHub5 天前
Le-DETR:省80%预训练数据,这个实时检测Transformer刷新SOTA|Georgia Tech & 北交大
深度学习·算法·计算机视觉
CoovallyAIHub5 天前
强化学习凭什么比监督学习更聪明?RL的“聪明”并非来自算法,而是因为它学会了“挑食”
深度学习·算法·计算机视觉
CoovallyAIHub5 天前
YOLO-IOD深度解析:打破实时增量目标检测的三重知识冲突
深度学习·算法·计算机视觉
这张生成的图像能检测吗7 天前
(论文速读)XLNet:语言理解的广义自回归预训练
人工智能·计算机视觉·nlp·注意力机制
十铭忘7 天前
自主认知-行动1——架构
人工智能·计算机视觉
yuzhuanhei7 天前
YOLO26实操记录(自用)
人工智能·计算机视觉·目标跟踪