autoware_minimum_rule_based_planner

Overview

A minimum rule-based trajectory planner that generates safe and feasible trajectories for autonomous driving. It follows the planned route by constructing trajectories from lanelet centerline and applies multi-stage optimization for geometric smoothness, velocity profiles, and obstacle avoidance.

Features

• Centerline-based path planning: Generates paths along lanelet centerline from the HD map, extending backward and forward from the ego vehicle’s position
• Smooth goal connection: Refines the path near the goal pose for smooth stopping
• Path shifting: Shifts the centerline path to start from the ego vehicle’s current pose, using curvature-aware shift distance calculation based on ego velocity and lateral acceleration limits
• Trajectory smoothing: Applies an Elastic Band smoother for geometric smoothing via a plugin interface
• Trajectory modification: Applies modifier plugins (e.g., obstacle stop) for safety modifications
• Velocity optimization: Computes a jerk-filtered velocity profile respecting constraints on acceleration, jerk, and lateral acceleration
• Test mode: Supports bypassing path planning by directly receiving a PathWithLaneId topic

Inputs / Outputs

Inputs

Outputs

Parameters

{{ json_to_markdown(“planning/autoware_minimum_rule_based_planner/schema/minimum_rule_based_planner.schema.json”) }}

Parameters can be set via YAML configuration files in the config/ directory.

Jerk-filtered smoother parameters are defined in config/velocity_smoother/jerk_filtered_smoother.param.yaml. EB smoother parameters are defined in config/trajectory_optimizer_plugins/elastic_band_smoother.param.yaml.

Changelog for package autoware_minimum_rule_based_planner

0.53.0 (2026-09-29)

• Merge remote-tracking branch 'origin/main' into prepare-0.53.0-changelog

• chore(autoware_trajectory_processor)!: rename to autoware_trajectory_modifier (#13389) rename processor -> modifier

• feat(trajectory_modifier, minimum_rule_based_planner): integrate semseg pointcloud into modifier and backup planner modules (#13265)

  • fix get_nearest_object_collision function to return distance to projected collision point instead of didistance to initial object state
  • modify get_object_polygon lambda to simply polygon expansion for shape types other than POLYGON
  • fix format
  • refactor get_predicted_obj_pose_at_time() to use highest confidence non empty predicted path
  • support new perception pointcloud interface using point type PointXYZCPE, and update pointcloud processing code in modifier and backup planner
  • Switch obstacle-stop point type from pcl::PointXYZ to PointXYZCPE
  • Filter input points by configurable target class labels (class_id) and axis-aligned range/height bounds
  • Remove voxel-grid downsampling, Euclidean clustering, and convex-hull extraction from the obstacle-stop PCD pipeline
  • Replace PCL CropBox / transform_pointcloud usage with manual x/y/z filtering and transform (PointXYZCPE has no PCL .data member)
  • Add PointCloudClassification -> ObjectType mapping for semantic labels
  • Replace voxel_grid_filter / clustering params with pointcloud.target_types in config, parameter structs, and schemas for both packages
  • Default target_types to hazard, structure, and vegetation
  • Add autoware_point_types dependency to trajectory_modifier
  • Update trajectory_modifier obstacle-stop integration test params for the simplified filter path
  • Share the simplified PointCloudFilter path with minimum_rule_based_planner obstacle_stop
  • align PCD crop and clean up cluster debug leftovers
  • Match MRBP obstacle-stop crop AABB to trajectory_modifier using std::minmax and a 1.0 m buffer so inverted corners after yaw do not empty the crop box
  • Remove dead cluster_points debug path after clustering removal
  • Publish filtered_points from MRBP obstacle_stop and update debug text to filtered -> target in both packages
  • preserve PointXYZCPE fields in obstacle tracker
  • Store full PointXYZCPE in PersistentPoint instead of xyz-only geometry_msgs positions
  • Keep class_id, probability, and entropy when emitting active tracked points
  • clean up code
  • filter surround obstacle pointclouds by type and range
  • Parse PointXYZCPE clouds in surround_obstacle_stop and keep only configured semantic labels before proximity checks
  • Crop to the ego footprint expanded by front/side/back thresholds and hysteresis, then convert to PointXYZ for ProximityChecker
  • Add target_objects.pointcloud params (default hazard/structure/vegetation) to config, parameter structs, and schemas in both packages
  • Allow unknown labels in trajectory_modifier surround integration tests so XYZ-only fixtures still exercise the PCD path

  * Update planning/autoware_minimum_rule_based_planner/config/minimum_rule_based_planner.param.yaml Co-authored-by: Kotaro Uetake <<60615504+ktro2828@users.noreply.github.com>>

  * Update planning/autoware_minimum_rule_based_planner/config/minimum_rule_based_planner.param.yaml Co-authored-by: Kotaro Uetake <<60615504+ktro2828@users.noreply.github.com>>

  * Update planning/autoware_minimum_rule_based_planner/param/minimum_rule_based_planner_parameters.yaml Co-authored-by: Kotaro Uetake <<60615504+ktro2828@users.noreply.github.com>>

  * Update planning/autoware_minimum_rule_based_planner/param/minimum_rule_based_planner_parameters.yaml Co-authored-by: Kotaro Uetake <<60615504+ktro2828@users.noreply.github.com>>

  * Update planning/autoware_trajectory_modifier/config/trajectory_modifier.param.yaml Co-authored-by: Kotaro Uetake <<60615504+ktro2828@users.noreply.github.com>>

  * Update planning/autoware_trajectory_modifier/src/trajectory_modifier_plugins/surround_obstacle_stop.cpp Co-authored-by: Kotaro Uetake <<60615504+ktro2828@users.noreply.github.com>>

File truncated at 100 lines see the full file