πŸŽ“ Lesson 4 D3

RRT* Algorithm Walkthrough with Industrial Convergence Guarantees

RRT* is a smart robot path planner that keeps improving its route over time until it finds the best possible path around obstacles.

🎯 Learning Objectives

  • βœ“ Explain the theoretical conditions required for RRT* asymptotic optimality in constrained industrial environments
  • βœ“ Design an RRT* implementation with collision-checking and adaptive rewiring radius for a surface mining drill rig navigation scenario
  • βœ“ Analyze convergence behavior by calculating effective connection radius bounds for a 3D blast-face mapping task
  • βœ“ Apply Lyapunov-like stability arguments to justify RRT*’s almost-sure convergence in non-stationary mine site conditions

πŸ“– Why This Matters

In autonomous drilling and muck removal systems at open-pit mines, planners must generate safe, fuel-efficient, and geotechnically compliant trajectories β€” not just collision-free ones. RRT* delivers provably better paths than RRT over time, enabling fleet-wide energy savings and reduced wear on hydraulic actuators. Real-world deployments by Komatsu and Sandvik show 12–18% reduction in cycle time when RRT*-guided path optimization replaces greedy A* in dynamic haul road routing.

πŸ“˜ Core Principles

RRT* extends RRT by introducing two key mechanisms: (1) incremental rewiring of newly added nodes to improve parent-child cost-to-come, and (2) a connection radius r_n = Ξ³ (log n / n)^(1/d) that shrinks with tree size n to ensure asymptotic optimality in d-dimensional configuration space. For mining applications, the configuration space includes joint limits, slope constraints, GPS/IMU uncertainty envelopes, and blast-hole exclusion zones. Convergence requires the free space to be open, bounded, and have positive measure β€” assumptions validated in surveyed pit benches but violated near unstable highwalls or unmodelled boulders.

πŸ“ Optimal Connection Radius

The connection radius r_n governs how many existing nodes are considered for rewiring when adding node x_new. Choosing r_n too large causes excessive computation; too small delays convergence. The theoretically justified form balances exploration and optimality.

RRT* Connection Radius

r_n = Ξ³ \left( \frac{\ln n}{n} \right)^{1/d}

Defines the maximum distance within which a newly sampled node searches for potential parents and neighbors for rewiring.

Variables:
SymbolNameUnitDescription
r_n Connection radius m Maximum Euclidean distance in configuration space for neighbor search
Ξ³ Scaling constant dimensionless Tuned parameter balancing convergence speed and computational load
n Number of nodes in tree count Current size of the RRT* tree
d Configuration space dimensionality dimensionless Degrees of freedom (e.g., 6 for rigid-body SE(3) + joint limits)
Typical Ranges:
Surface drill rig (6-DOF): 0.35 – 0.80 m
Underground LHD (7-DOF with bucket pose): 0.25 – 0.60 m

πŸ’‘ Worked Example

Problem: Given: 6-DOF drill rig navigation in a 3D mapped pit (d = 6), current tree size n = 5,000 nodes, Ξ³ = 1.5 (empirically tuned for mining terrain), calculate r_n in meters.
1. Step 1: Compute log(n) = ln(5000) β‰ˆ 8.517
2. Step 2: Compute log(n)/n = 8.517 / 5000 β‰ˆ 0.001703
3. Step 3: Compute (log(n)/n)^(1/d) = (0.001703)^(1/6) β‰ˆ 0.402
4. Step 4: Multiply by Ξ³: r_n = 1.5 Γ— 0.402 β‰ˆ 0.603 m
Answer: The result is 0.60 m, which falls within the safe range of 0.4–0.8 m for articulated drill rigs operating on compacted gravel haul roads.

πŸ—οΈ Real-World Application

At BHP’s Jimblebar Iron Ore Mine (Pilbara, WA), RRT* was deployed onboard autonomous rotary drill rigs (Caterpillar CS3000) to plan trajectories between blast-hole patterns while respecting real-time LiDAR-detected overhangs and slope-limited no-go zones. By tuning Ξ³ to 1.3 and enforcing r_n β‰₯ 0.35 m (minimum sensor resolution + actuator latency buffer), planners achieved <0.15 m path deviation from optimal offline CHOMP solutions β€” meeting ISO 19847:2022 functional safety requirements for Level 3 autonomy in mining.

πŸ“‹ Case Connection

πŸ“‹ Palletizing Robot Path Optimization for High-Speed E-Commerce Fulfillment

Vibration-induced misalignment causing 8% pallet collapse rate at 120 cycles/hour

πŸ“‹ Welding Robot Cell Collision-Free Path Synthesis for Automotive Body-in-White

Interference between dual-arm robots and fixture-mounted part carriers during simultaneous weld passes

πŸ“‹ Vision-Guided Bin Picking Trajectory Generation for Aerospace Fasteners

Cluttered bin geometry and reflective titanium fasteners causing pose estimation drift β†’ failed grasps and dropped parts

πŸ“‹ Collaborative Robot Trajectory Certification for Pharma Packaging Line

Need for ISO/TS 15066-compliant motion profiles validated for human-robot proximity during carton loading

πŸ“š References