mobility_path_to_external_object.cpp

Go to the documentation of this file.

// Copyright 2019-2023 Leidos
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
//     http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
 
#include <tf2/LinearMath/Transform.h>
 
#include <string>
#include <utility>
 
#include <motion_computation/impl/mobility_path_to_external_object_helpers.hpp>
#include <rclcpp/logger.hpp>
#include <rclcpp/logging.hpp>
 
#include <carma_perception_msgs/msg/external_object.hpp>
#include <carma_v2x_msgs/msg/mobility_path.hpp>
 
namespace motion_computation
{
namespace conversion
{
void convert(
  const carma_v2x_msgs::msg::MobilityPath & in_msg,
  carma_perception_msgs::msg::ExternalObject & out_msg,
  const lanelet::projection::LocalFrameProjector & map_projector)
{
  constexpr double mobility_path_points_timestep_size =
    0.1;  // Mobility path timestep size per message spec
 
  out_msg.size.x = 2.5;  // TODO(carma) identify better approach for object size in mobility path
  out_msg.size.y = 2.25;
  out_msg.size.z = 2.0;
 
  // get reference origin in ECEF (convert from cm to m)
  double ecef_x = static_cast<double>(in_msg.trajectory.location.ecef_x) / 100.0;
  double ecef_y = static_cast<double>(in_msg.trajectory.location.ecef_y) / 100.0;
  double ecef_z = static_cast<double>(in_msg.trajectory.location.ecef_z) / 100.0;
 
  // Convert general information
  // clang-off
  out_msg.presence_vector |= carma_perception_msgs::msg::ExternalObject::ID_PRESENCE_VECTOR;
  out_msg.presence_vector |= carma_perception_msgs::msg::ExternalObject::POSE_PRESENCE_VECTOR;
  out_msg.presence_vector |= carma_perception_msgs::msg::ExternalObject::VELOCITY_PRESENCE_VECTOR;
  out_msg.presence_vector |=
    carma_perception_msgs::msg::ExternalObject::OBJECT_TYPE_PRESENCE_VECTOR;
  out_msg.presence_vector |= carma_perception_msgs::msg::ExternalObject::BSM_ID_PRESENCE_VECTOR;
  out_msg.presence_vector |= carma_perception_msgs::msg::ExternalObject::DYNAMIC_OBJ_PRESENCE;
  out_msg.presence_vector |= carma_perception_msgs::msg::ExternalObject::PREDICTION_PRESENCE_VECTOR;
  // clang-on
 
  out_msg.object_type = carma_perception_msgs::msg::ExternalObject::SMALL_VEHICLE;
  std::hash<std::string> hasher;
 
  // TODO(carma) hasher returns size_t, message accept uint32_t which we might lose info
  auto hashed = hasher(in_msg.m_header.sender_id);
  out_msg.id = static_cast<uint32_t>(hashed);
 
  // convert hex std::string to uint8_t array
  for (size_t i = 0; i < in_msg.m_header.sender_bsm_id.size(); i += 2) {
    unsigned int num = 0;
    sscanf(in_msg.m_header.sender_bsm_id.substr(i, i + 2).c_str(), "%x", &num);
    out_msg.bsm_id.push_back((uint8_t)num);
  }
 
  // first point's timestamp
  out_msg.header.stamp =
    rclcpp::Time((uint64_t)in_msg.m_header.timestamp * 1e6);  // ms to nanoseconds
 
  // If it is a static object, we finished processing
  if (in_msg.trajectory.offsets.size() < 2) {
    out_msg.dynamic_obj = false;
    return;
  }
  out_msg.dynamic_obj = true;
 
  // get planned trajectory points
  carma_perception_msgs::msg::PredictedState prev_state;
  tf2::Vector3 prev_pt_ecef{ecef_x, ecef_y, ecef_z};
 
  auto prev_pt_map = impl::transform_to_map_frame(prev_pt_ecef, map_projector);
  double prev_yaw = 0.0;
 
  double message_offset_x = 0.0;  // units cm
  double message_offset_y = 0.0;
  double message_offset_z = 0.0;
 
  rclcpp::Duration mobility_path_point_delta_t(mobility_path_points_timestep_size * 1e9);
 
  // Note the usage of current vs previous in this loop can be a bit confusing
  // The intended behavior is we our always storing our prev_point but using
  // curr_pt for computing velocity at prev_point
  for (size_t i = 0; i < in_msg.trajectory.offsets.size(); i++) {
    auto curr_pt_msg = in_msg.trajectory.offsets[i];
 
    message_offset_x = static_cast<double>(curr_pt_msg.offset_x) + message_offset_x;
    message_offset_y = static_cast<double>(curr_pt_msg.offset_y) + message_offset_y;
    message_offset_z = static_cast<double>(curr_pt_msg.offset_z) + message_offset_z;
 
    tf2::Vector3 curr_pt_ecef{
      ecef_x + message_offset_x / 100.0, ecef_y + message_offset_y / 100.0,
      ecef_z + message_offset_z /
                 100.0};  // ecef_x is in m while message_offset_x is in cm. Want m as final result
    auto curr_pt_map = impl::transform_to_map_frame(curr_pt_ecef, map_projector);
 
    carma_perception_msgs::msg::PredictedState curr_state;
 
    if (i == 0)  // First point's state should be stored outside "predictions"
    {
      rclcpp::Time prev_stamp_as_time = rclcpp::Time(out_msg.header.stamp);
      rclcpp::Time updated_time_step = prev_stamp_as_time + mobility_path_point_delta_t;
 
      auto res = impl::composePredictedState(
        curr_pt_map, prev_pt_map, prev_stamp_as_time, updated_time_step,
        prev_yaw);  // Position returned is that of prev_pt_map NOT curr_pt_map
      curr_state = std::get<0>(res);
      prev_yaw = std::get<1>(res);
      // Compute out_msg pose
      out_msg.pose.pose = curr_state.predicted_position;  // Orientation computed from first point
                                                          // in offsets with location
      out_msg.velocity.twist = curr_state.predicted_velocity;  // Velocity derived from first point
 
    } else {
      // NOTE: This is where the time increment happens but it feels incorrect.
      // since the corresponding data is actually from the original prev_state stamp
      rclcpp::Time prev_stamp_as_time =
        rclcpp::Time(prev_state.header.stamp) + mobility_path_point_delta_t;
      rclcpp::Time updated_time_step = prev_stamp_as_time + mobility_path_point_delta_t;
 
      auto res = impl::composePredictedState(
        curr_pt_map, prev_pt_map, prev_stamp_as_time, updated_time_step, prev_yaw);
      curr_state = std::get<0>(res);
      prev_yaw = std::get<1>(res);
      out_msg.predictions.push_back(curr_state);
    }
 
    if (
      i == in_msg.trajectory.offsets.size() -
             1)  // if last point, copy the prev_state velocity & orientation to the last point too
    {
      curr_state.predicted_position.position.x = curr_pt_map.x();
      curr_state.predicted_position.position.y = curr_pt_map.y();
      curr_state.predicted_position.position.z = curr_pt_map.z();
      out_msg.predictions.push_back(curr_state);
    }
 
    prev_state = curr_state;
    prev_pt_map = curr_pt_map;
  }
 
  impl::calculateAngVelocityOfPredictedStates(out_msg);
 
  return;
}
 
namespace impl
{
std::pair<carma_perception_msgs::msg::PredictedState, double> composePredictedState(
  const tf2::Vector3 & curr_pt, const tf2::Vector3 & prev_pt, const rclcpp::Time & prev_time_stamp,
  const rclcpp::Time & curr_time_stamp, double prev_yaw)
{
  carma_perception_msgs::msg::PredictedState output_state;
  // Set Position
  output_state.predicted_position.position.x = prev_pt.x();
  output_state.predicted_position.position.y = prev_pt.y();
  output_state.predicted_position.position.z = prev_pt.z();
 
  // Set Orientation
  Eigen::Vector2d vehicle_vector = {curr_pt.x() - prev_pt.x(), curr_pt.y() - prev_pt.y()};
  Eigen::Vector2d x_axis = {1, 0};
  double yaw = 0.0;
  if (
    vehicle_vector.norm() <
    0.000001) {  // If there is zero magnitude use previous yaw to avoid divide by 0
    yaw = prev_yaw;
    RCLCPP_DEBUG_STREAM(
      rclcpp::get_logger("motion_computation::conversion"),
      "Two identical points sent for predicting heading. Forced to use previous yaw or 0 if "
      "first "
      "point");
  } else {
    yaw = std::acos(vehicle_vector.dot(x_axis) / (vehicle_vector.norm() * x_axis.norm()));
  }
 
  tf2::Quaternion vehicle_orientation;
  vehicle_orientation.setRPY(0, 0, yaw);
  output_state.predicted_position.orientation.x = vehicle_orientation.getX();
  output_state.predicted_position.orientation.y = vehicle_orientation.getY();
  output_state.predicted_position.orientation.z = vehicle_orientation.getZ();
  output_state.predicted_position.orientation.w = vehicle_orientation.getW();
 
  // Set velocity
  output_state.predicted_velocity.linear.x =
 
    vehicle_vector.norm() / (curr_time_stamp - prev_time_stamp).seconds();
 
  // Set timestamp
  output_state.header.stamp = builtin_interfaces::msg::Time(prev_time_stamp);
 
  return std::make_pair(output_state, yaw);
}
 
void calculateAngVelocityOfPredictedStates(carma_perception_msgs::msg::ExternalObject & object)
{
  if (!object.dynamic_obj || object.predictions.size() == 0) {
    return;
  }
 
  // Object's current angular velocity
  double dt = (rclcpp::Time(object.header.stamp) - rclcpp::Time(object.predictions[0].header.stamp))
                .seconds();
  object.velocity.twist.angular.z =
    (impl::getYawFromQuaternionMsg(object.pose.pose.orientation) -
     impl::getYawFromQuaternionMsg(object.predictions[0].predicted_position.orientation)) /
    dt;
 
  // Predictions' angular velocities
  auto prev_orient = object.pose.pose.orientation;
  auto prev_time = rclcpp::Time(object.predictions[0].header.stamp);
  for (auto & pred : object.predictions) {
    dt = (rclcpp::Time(pred.header.stamp) - prev_time).seconds();
    pred.predicted_velocity.angular.z =
      (impl::getYawFromQuaternionMsg(prev_orient) -
       impl::getYawFromQuaternionMsg(pred.predicted_position.orientation)) /
      dt;
  }
}
 
double getYawFromQuaternionMsg(const geometry_msgs::msg::Quaternion & quaternion)
{
  tf2::Quaternion orientation;
  orientation.setX(quaternion.x);
  orientation.setY(quaternion.y);
  orientation.setZ(quaternion.z);
  orientation.setW(quaternion.w);
 
  double roll;
  double pitch;
  double yaw;
  tf2::Matrix3x3(orientation).getRPY(roll, pitch, yaw);
 
  return yaw;
}
 
tf2::Vector3 transform_to_map_frame(
  const tf2::Vector3 & ecef_point, const lanelet::projection::LocalFrameProjector & map_projector)
{
  lanelet::BasicPoint3d map_point = map_projector.projectECEF(
    {ecef_point.x(), ecef_point.y(), ecef_point.z()},
    -1);  // Input should already be converted to m
 
  return tf2::Vector3(map_point.x(), map_point.y(), map_point.z());
}
}  // namespace impl
 
}  // namespace conversion
 
}  // namespace motion_computation