Prediction: bug fixes and updates.

modules/prediction/common/feature_output.cc

@@ -17,6 +17,7 @@
 #include "modules/prediction/common/feature_output.h"
 
 #include <string>
+#include <vector>
 
 #include "modules/common/util/file.h"
 #include "modules/prediction/common/prediction_system_gflags.h"
@@ -25,7 +26,7 @@ namespace apollo {
 namespace prediction {
 
 Features FeatureOutput::features_;
-DataForLearning FeatureOutput::data_for_learning_;
+ListDataForLearning FeatureOutput::list_data_for_learning_;
 std::size_t FeatureOutput::index_ = 0;
 
 void FeatureOutput::Close() {
@@ -37,7 +38,7 @@ void FeatureOutput::Close() {
 void FeatureOutput::Clear() {
   index_ = 0;
   features_.Clear();
-  data_for_learning_.Clear();
+  list_data_for_learning_.Clear();
 }
 
 bool FeatureOutput::Ready() {
@@ -49,9 +50,15 @@ void FeatureOutput::Insert(const Feature& feature) {
   features_.add_feature()->CopyFrom(feature);
 }
 
-void FeatureOutput::InsertIntoLearningData(const Feature& feature) {
-  data_for_learning_.set_id(feature.id());
-  data_for_learning_.set_timestamp(feature.timestamp());
+void FeatureOutput::InsertDataForLearning(
+    const Feature& feature, const std::vector<double>& feature_values) {
+  DataForLearning* data_for_learning =
+      list_data_for_learning_.add_data_for_learning();
+  data_for_learning->set_id(feature.id());
+  data_for_learning->set_timestamp(feature.timestamp());
+  for (size_t i = 0; i < feature_values.size(); ++i) {
+    data_for_learning->add_features_for_learning(feature_values[i]);
+  }
 }
 
 void FeatureOutput::Write() {
@@ -66,14 +73,14 @@ void FeatureOutput::Write() {
     ++index_;
   }
 
-  if (!data_for_learning_.has_id()) {
+  if (!list_data_for_learning_.data_for_learning_size() <= 0) {
     ADEBUG << "Skip writing empty data_for_learning.";
   } else {
     const std::string file_name =
         FLAGS_prediction_data_dir + "/datalearn." +
         std::to_string(index_) + ".bin";
-    common::util::SetProtoToBinaryFile(data_for_learning_, file_name);
-    data_for_learning_.Clear();
+    common::util::SetProtoToBinaryFile(list_data_for_learning_, file_name);
+    list_data_for_learning_.Clear();
   }
 }
 

modules/prediction/common/feature_output.h

@@ -16,6 +16,8 @@
 
 #pragma once
 
+#include <vector>
+
 #include "modules/prediction/proto/offline_features.pb.h"
 
 namespace apollo {
@@ -54,7 +56,8 @@ class FeatureOutput {
     * @brief Insert a data_for_learning
     * @param A feature in proto
     */
-  static void InsertIntoLearningData(const Feature& feature);
+  static void InsertDataForLearning(
+      const Feature& feature, const std::vector<double>& feature_values);
 
   /**
    * @brief Write features to a file
@@ -69,7 +72,7 @@ class FeatureOutput {
 
  private:
   static Features features_;
-  static DataForLearning data_for_learning_;
+  static ListDataForLearning list_data_for_learning_;
   static std::size_t index_;
 };
 

modules/prediction/container/obstacles/obstacle.cc

@@ -1143,11 +1143,14 @@ void Obstacle::BuildLaneGraphFromLeftToRight() {
 
   // Treat the most probable lane_segment as the center, put its left
   // and right neighbors into a vector following the left-to-right order.
+  if (!feature->has_lane() || !feature->lane().has_lane_feature()) {
+    return;
+  }
   std::shared_ptr<const LaneInfo> center_lane_info =
       PredictionMap::LaneById(feature->lane().lane_feature().lane_id());
   std::vector<std::string> lane_ids_ordered;
   for (int i = center_lane_info->lane().left_neighbor_forward_lane_id().size()
-       - 1; i >= 0; i --) {
+       - 1; i >= 0; --i) {
     if (center_lane_info->lane().left_neighbor_forward_lane_id(i).has_id()) {
       lane_ids_ordered.push_back(
           center_lane_info->lane().left_neighbor_forward_lane_id(i).id());
@@ -1165,7 +1168,7 @@ void Obstacle::BuildLaneGraphFromLeftToRight() {
 
   // Build lane_graph for every lane_segment and update it into proto.
   int seq_id = 0;
-  for (size_t i = 0; i < lane_ids_ordered.size(); i ++) {
+  for (size_t i = 0; i < lane_ids_ordered.size(); ++i) {
     // Construct the local lane_graph based on the current lane_segment.
     bool vehicle_is_on_lane = (lane_ids_ordered[i] ==
                                center_lane_info->lane().id().id());

modules/prediction/container/obstacles/obstacles_container.cc

@@ -25,8 +25,6 @@
 #include "modules/prediction/common/prediction_system_gflags.h"
 #include "modules/prediction/container/obstacles/obstacle_clusters.h"
 
-#define ADEBUG AINFO
-
 namespace apollo {
 namespace prediction {
 

modules/prediction/evaluator/vehicle/lane_scanning_evaluator.cc

@@ -83,9 +83,9 @@ void LaneScanningEvaluator::Evaluate(Obstacle* obstacle_ptr) {
   //  - if in online mode, pass it through trained model to evaluate.
   std::vector<double> feature_values;
   ExtractFeatures(obstacle_ptr, lane_graph_ptr, &feature_values);
-  // TODO(jiacheng): sanity check of extracted feature_values:
   if (FLAGS_prediction_offline_mode) {
-    // TODO(jiacheng): save the extracted features locally for offline training.
+    FeatureOutput::InsertDataForLearning(*latest_feature_ptr, feature_values);
+    ADEBUG << "Save extracted features for learning locally.";
   } else {
     // TODO(jiacheng): once the model is trained, implement this online part.
   }
@@ -122,10 +122,16 @@ bool LaneScanningEvaluator::ExtractFeatures(
     ADEBUG << "Failed to extract static environmental features around obs_id = "
            << id;
   }
-  // TODO(jiacheng): sanity check of extracted static env features:
+  if (static_feature_values.size() %
+      (SINGLE_LANE_FEATURE_SIZE * LANE_POINTS_SIZE) != 0) {
+    ADEBUG << "Obstacle [" << id << "] has incorrect static env feature size: "
+           << static_feature_values.size() << ".";
+    return false;
+  }
   feature_values->insert(feature_values->end(),
                          static_feature_values.begin(),
                          static_feature_values.end());
+
   return true;
 }
 

modules/prediction/evaluator/vehicle/lane_scanning_evaluator.h

@@ -74,7 +74,7 @@ class LaneScanningEvaluator : public Evaluator {
   static const size_t OBSTACLE_FEATURE_SIZE = 5 * 9;
   static const size_t INTERACTION_FEATURE_SIZE = 8;
   static const size_t SINGLE_LANE_FEATURE_SIZE = 4;
-  static const size_t LANE_POINTS_SIZE = 20;
+  static const size_t LANE_POINTS_SIZE = 100;  // (100 * 0.2m = 20m)
 };
 
 }  // namespace prediction

modules/prediction/proto/offline_features.proto

@@ -29,4 +29,4 @@ message DataForLearning {
 
 message ListDataForLearning {
   repeated DataForLearning data_for_learning = 1;
-}
+}
\ No newline at end of file