AgentSkillsCN

placement_solver

阶段 2：利用 PSO 算法为所有实体计算最优的 (x, y, theta) 位置。适用场景：(1) 需要在工作空间中对多个实体进行定位；(2) 优化布局以提升可达性并避免碰撞；(3) 满足区域规划者提出的空间约束条件。系统会基于逆运动学、碰撞检测与区域约束，不断迭代优化成本函数。

SKILL.md
--- frontmatter
name: placement_solver
description: "Stage 2: Calculates optimal (x, y, theta) positions for all entities using PSO algorithms. Use when: (1) Need to position multiple entities in workspace, (2) Optimize layout for reachability and collision avoidance, (3) Satisfy spatial constraints from region proposer. Iteratively minimizes cost function based on IK, collisions, and zone constraints."
metadata:
  stage: "2"

Placement Solver

Overview

Optimizes spatial positioning of all workcell entities through PSO (Particle Swarm Optimization). Balances competing objectives: reachability, collision avoidance, and zone compliance.

When to Use This Skill

  • •After Stage 1 is complete (all requirements gathered)
  • •After region_proposer has defined spatial zones
  • •Ready to compute optimal layout positions

Workflow Instructions for Agent

Step 1: Prepare Layout Data Create a JSON file with:

  • •List of entities to place (robot, table, objects)
  • •Defined zones from region_proposer
  • •Any positioning constraints

Step 2: Execute PSO Script Run the optimization:

bash
python skills/placement_solver/scripts/solve_placement.py --scene_json "layout_data.json" --algorithm "pso" --iterations 100

Step 3: Parse Results Script outputs:

json
{
  "optimized_layout": {
    "robot": {"position": [0.0, 0.0, 0.0], "orientation": [0, 0, 0]},
    "table": {"position": [0.5, 0.3, 0.0], "orientation": [0, 0, 0]}
  },
  "optimization_metrics": {
    "final_cost": 8.5,
    "iterations": 100,
    "convergence": "good"
  }
}

Step 4: Interpret Quality

  • •Excellent: cost < 10 (ready for simulation)
  • •Good: 10 ≤ cost < 50 (acceptable, minor issues)
  • •Poor: cost ≥ 50 (consider adjusting constraints)

Step 5: Present Results to User Show:

  • •Optimized positions for each entity
  • •Cost metrics and quality assessment
  • •CRITICAL: ASK FOR USER CONFIRMATION before proceeding to Stage 3

Step 6: Wait for Confirmation DO NOT proceed to Stage 3 until user explicitly confirms the layout.

Step 7: Next Action Once confirmed → Load genesis_scene_builder skill

Usage

bash
python scripts/solve_placement.py --scene_json "scene.json" --algorithm "pso" --iterations 100

Algorithm Selection

  • •PSO (default): Fast convergence, good for moderate complexity (30 particles, social+cognitive forces)
  • •Diffusion: More thorough exploration, better for complex layouts (gradient descent with noise)

Cost Function

  1. •IK Reachability (weight: 5.0): Penalize unreachable positions
  2. •Collision Avoidance (weight: 100.0): Heavy penalty for interpenetration
  3. •Zone Compliance (weight: 50.0): Penalize zone boundary violations

Output Quality

  • •Excellent: cost < 10
  • •Good: 10 ≤ cost < 50
  • •Acceptable: cost ≥ 50

Input Schema

{ "entities_to_place": "array", "defined_zones": "array (Output from Region Proposer)", "algorithm": "string ('pso' | 'diffusion')", "iterations": 100 }

Output Schema

{ "status": "success", "placements": "array (List of entity IDs and their final x,y,z,quat)", "final_cost_score": "float (lower is better)" }