PointPerception.hpp

Go to the documentation of this file.

// Copyright 2025 Intelligent Robotics Lab
//
// 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.

 
#ifndef EASYNAV_COMMON_TYPES__POINTPERCEPTIONS_HPP_
#define EASYNAV_COMMON_TYPES__POINTPERCEPTIONS_HPP_
 
#include <string>
#include <vector>
#include <optional>
 
#include "tf2/LinearMath/Transform.hpp"
#include "pcl/point_cloud.h"
#include "pcl/point_types.h"
#include "pcl/PointIndices.h"
 
#include "sensor_msgs/msg/laser_scan.hpp"
#include "sensor_msgs/msg/point_cloud2.hpp"
 
#include "rclcpp/time.hpp"
#include "rclcpp_lifecycle/lifecycle_node.hpp"
 
#include "easynav_common/types/Perceptions.hpp"
#include "easynav_common/CircularBuffer.hpp"
#include "easynav_common/RTTFBuffer.hpp"
 
namespace std
{

template<>
struct hash<std::tuple<int, int, int>>
{
  std::size_t operator()(const std::tuple<int, int, int> & key) const
  {
    std::size_t h1 = std::hash<int>()(std::get<0>(key));
    std::size_t h2 = std::hash<int>()(std::get<1>(key));
    std::size_t h3 = std::hash<int>()(std::get<2>(key));
    return h1 ^ (h2 << 1) ^ (h3 << 2);
  }
};
 
}  // namespace std
 
namespace easynav
{
 
struct PointPerceptionBufferType
{
  pcl::PointCloud<pcl::PointXYZ> data;
  std::string frame;
  rclcpp::Time stamp;
};

class PointPerception : public PerceptionBase
{
public:
  static constexpr std::string_view default_group_ = "points";

  static inline bool supports_msg_type(std::string_view t)
  {
    return t == "sensor_msgs/msg/LaserScan" ||
           t == "sensor_msgs/msg/PointCloud2";
  }

  pcl::PointCloud<pcl::PointXYZ> data;
 
  bool pending_available_{false};
  pcl::PointCloud<pcl::PointXYZ> pending_cloud_;
  std::string pending_frame_;
  rclcpp::Time pending_stamp_;
 

  void resize(std::size_t size)
  {
    data.points.resize(size);
  }

  const PointPerceptionBufferType & get_last_perception() const
  {
    return buffer.latest_ref();
  }
 
  void integrate_pending_perceptions()
  {
  // Access TF buffer singleton (already initialized somewhere with a clock)
    auto tf_buffer_ptr = RTTFBuffer::getInstance();
    auto & tf_buffer = *tf_buffer_ptr;
    const auto tf_info = tf_buffer.get_tf_info();
    const std::string & robot_frame = tf_info.robot_frame;
 
  // ------------------------------------------------------------------
  // 1. Push pending perception into the circular buffer exactly once.
  // ------------------------------------------------------------------
    if (pending_available_) {
      PointPerceptionBufferType pending_item;
      pending_item.data = std::move(pending_cloud_); // avoid deep copy
      pending_item.frame = pending_frame_;
      pending_item.stamp = pending_stamp_;
 
      buffer.push(std::move(pending_item));
      pending_available_ = false;
    }
 
    const std::size_t count = buffer.size();
    if (count == 0) {
    // No candidates at all: keep current visible state as is.
      return;
    }
 
  // ------------------------------------------------------------------
  // 2. Drain the circular buffer into a temporary vector so we can
  //    inspect all items and then rebuild the buffer.
  // ------------------------------------------------------------------
    std::vector<PointPerceptionBufferType> items;
    items.reserve(count);
 
    for (std::size_t i = 0; i < count; ++i) {
      PointPerceptionBufferType item;
      if (!buffer.pop(item)) {
        break; // Defensive guard if pop() fails unexpectedly.
      }
      items.push_back(std::move(item));
    }
 
    if (items.empty()) {
    // Nothing recovered from buffer: keep visible state untouched.
      return;
    }
 
  // ------------------------------------------------------------------
  // 3. Find indices:
  //    - newest_idx: newest perception by timestamp (regardless of TF),
  //    - newest_valid_idx: newest perception that has a valid TF.
  //
  //    IMPORTANT: store only indices to avoid copying point clouds
  //    during the scan.
  // ------------------------------------------------------------------
    std::optional<std::size_t> newest_idx;
    std::optional<std::size_t> newest_valid_idx;
 
    for (std::size_t i = 0; i < items.size(); ++i) {
      const auto & item = items[i];
 
    // Track newest item overall (used when no TF is valid).
      if (!newest_idx || item.stamp > items[*newest_idx].stamp) {
        newest_idx = i;
      }
 
      bool has_tf = false;
      try {
        has_tf = tf_buffer.canTransform(
        robot_frame,
        item.frame,
        tf2_ros::fromMsg(item.stamp),
        tf2::durationFromSec(0.0));
      } catch (...) {
      // Any TF exception is treated as "no valid TF" for this item.
        has_tf = false;
      }
 
      if (has_tf) {
        if (!newest_valid_idx || item.stamp > items[*newest_valid_idx].stamp) {
          newest_valid_idx = i;
        }
      }
    }
 
  // ------------------------------------------------------------------
  // 4. Update visible state BEFORE moving items back into the buffer.
  //    This guarantees that `data` corresponds to:
  //      - the newest TF-valid item if any exists, otherwise
  //      - the newest item overall.
  // ------------------------------------------------------------------
    if (newest_valid_idx) {
      const auto & sel = items[*newest_valid_idx];
      data = sel.data;        // single deep copy (intentional)
      frame_id = sel.frame;
      stamp = sel.stamp;
      valid = true;           // "valid" means "usable / not too old", not "TF ok"
      new_data = true;
    } else if (newest_idx) {
      const auto & sel = items[*newest_idx];
      data = sel.data;        // single deep copy (intentional)
      frame_id = sel.frame;
      stamp = sel.stamp;
      valid = true;
      new_data = true;
    } else {
    // Defensive: should not happen because items is non-empty.
      return;
    }
 
  // ------------------------------------------------------------------
  // 5. Rebuild the circular buffer from scratch (move-only, no copies).
  // ------------------------------------------------------------------
    buffer.clear();
 
    if (newest_valid_idx) {
    // Keep the newest TF-valid item and any newer items (even if TF is not yet available).
      const rclcpp::Time cutoff_stamp = items[*newest_valid_idx].stamp;
 
      for (auto & item : items) {
        if (item.stamp >= cutoff_stamp) {
          buffer.push(std::move(item));
        }
      }
    } else {
      // No TF-valid items: keep everything in case TF arrives later.
      for (auto & item : items) {
        buffer.push(std::move(item));
      }
    }
  }
 
protected:
  CircularBuffer<PointPerceptionBufferType> buffer{10};
};

class PointPerceptionHandler : public PerceptionHandler
{
public:
  std::string group() const override {return "points";}

  std::shared_ptr<PerceptionBase> create(const std::string &) override
  {
    return std::make_shared<PointPerception>();
  }

  rclcpp::SubscriptionBase::SharedPtr create_subscription(
    rclcpp_lifecycle::LifecycleNode & node,
    const std::string & topic,
    const std::string & type,
    std::shared_ptr<PerceptionBase> target,
    rclcpp::CallbackGroup::SharedPtr cb_group) override;
};

void convert(const sensor_msgs::msg::LaserScan & scan, pcl::PointCloud<pcl::PointXYZ> & pc);

sensor_msgs::msg::PointCloud2 perception_to_rosmsg(const PointPerception & perception);

sensor_msgs::msg::PointCloud2 points_to_rosmsg(const pcl::PointCloud<pcl::PointXYZ> & points);

using PointPerceptions =
  std::vector<std::shared_ptr<PointPerception>>;

PointPerceptions get_point_perceptions(std::vector<PerceptionPtr> & perceptionptr);

rclcpp::Time get_latest_point_perceptions_stamp(const PointPerceptions & perceptions);

class PointPerceptionsOpsView
{
public:
  struct VoxelKey
  {
    int x, y, z;

    bool operator==(const VoxelKey & other) const
    {
      return x == other.x && y == other.y && z == other.z;
    }
  };

  struct VoxelKeyHash
  {
    std::size_t operator()(const VoxelKey & key) const
    {
      std::size_t h1 = std::hash<int>{}(key.x);
      std::size_t h2 = std::hash<int>{}(key.y);
      std::size_t h3 = std::hash<int>{}(key.z);
      return h1 ^ (h2 << 1) ^ (h3 << 2);
    }
  };

  explicit PointPerceptionsOpsView(const PointPerceptions & perceptions);

  explicit PointPerceptionsOpsView(const PointPerception & perception);

  explicit PointPerceptionsOpsView(PointPerceptions && perceptions);
 
  PointPerceptionsOpsView(const PointPerceptionsOpsView &) = delete;
  PointPerceptionsOpsView & operator=(const PointPerceptionsOpsView &) = delete;
 
  PointPerceptionsOpsView(PointPerceptionsOpsView &&) = default;

  PointPerceptionsOpsView & filter(
    const std::vector<double> & min_bounds,
    const std::vector<double> & max_bounds,
    bool lazy_post_fuse = true);

  PointPerceptionsOpsView & downsample(double resolution);

  PointPerceptionsOpsView &
  collapse(const std::vector<double> & collapse_dims, bool lazy = true);

  pcl::PointCloud<pcl::PointXYZ> as_points() const;

  const pcl::PointCloud<pcl::PointXYZ> & as_points(int idx) const;

  PointPerceptionsOpsView & fuse(const std::string & target_frame, bool exact_time = true);

  PointPerceptionsOpsView & fuse(
    const std::string & target_frame,
    rclcpp::Time & stamp,
    bool exact_time = true);

  PointPerceptionsOpsView &
  add(
    const pcl::PointCloud<pcl::PointXYZ> points,
    const std::string & frame,
    rclcpp::Time stamp);

  const PointPerceptions & get_perceptions() const {return perceptions_;}

  rclcpp::Time get_latest_stamp() const;
 
private:
  std::optional<PointPerceptions> owned_;      
  const PointPerceptions & perceptions_;       
  std::vector<pcl::PointIndices> indices_;     
 
  // Lazy fusion state
  bool has_target_frame_ {false};              
  std::string target_frame_;                   
  std::vector<tf2::Transform> tf_transforms_;  
  std::vector<bool> tf_valid_;                 
 
  // Lazy collapse state
  bool collapse_x_ {false};                    
  bool collapse_y_ {false};                    
  bool collapse_z_ {false};                    
  float collapse_val_x_ {0.0f};                
  float collapse_val_y_ {0.0f};                
  float collapse_val_z_ {0.0f};                
 
  bool has_post_filter_ {false};
  double post_min_[3] {0.0, 0.0, 0.0};
  double post_max_[3] {0.0, 0.0, 0.0};
  bool use_post_min_[3] {false, false, false};
  bool use_post_max_[3] {false, false, false};
 
  // Temporary storage for as_points(int)
  mutable pcl::PointCloud<pcl::PointXYZ> tmp_single_cloud_;
};
 
}  // namespace easynav
 
#endif  // EASYNAV_COMMON_TYPES__POINTPERCEPTIONS_HPP_