Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Revisiting Replanning from Scratch: Real-Time Incremental Planning with Fast Almost-Surely Asymptotically Optimal Planners

Mitchell E. C. Sabbadini, Andrew H. Liu, Joseph Ruan, Tyler S. Wilson, Zachary Kingston, Jonathan D. Gammell

TL;DR

Dynamic, obstacle-rich environments challenge traditional replanning by requiring both speed and path quality. The authors propose solving incremental planning as independent, fast ASAO planning problems, using planners like EIT* (and AORRTC for real-world tests) to rapidly generate initial, high-quality intermediate paths and converge toward optimal global solutions without explicit plan reuse, with the global path composed as $\pi = \sigma_{1,s_1} \oplus \cdots \oplus \sigma_{N,s_N}$ and cost $c(\pi)$. In simulated 2D worlds and a real seven-DOF Franka arm, independent ASAO replanning with EIT* achieves higher success rates and shorter median global path lengths than plan-reuse planners such as RRTX and other ASAO methods, even at tight budgets like 50 ms. This approach simplifies implementation by avoiding obstacle-change tracking, demonstrates real-time replanning capabilities in dynamic scenarios, and opens avenues for combining with obstacle motion predictions and richer sensing in future work.

Abstract

Robots operating in changing environments either predict obstacle changes and/or plan quickly enough to react to them. Predictive approaches require a strong prior about the position and motion of obstacles. Reactive approaches require no assumptions about their environment but must replan quickly and find high-quality paths to navigate effectively. Reactive approaches often reuse information between queries to reduce planning cost. These techniques are conceptually sound but updating dense planning graphs when information changes can be computationally prohibitive. It can also require significant effort to detect the changes in some applications. This paper revisits the long-held assumption that reactive replanning requires updating existing plans. It shows that the incremental planning problem can alternatively be solved more efficiently as a series of independent problems using fast almost-surely asymptotically optimal (ASAO) planning algorithms. These ASAO algorithms quickly find an initial solution and converge towards an optimal solution which allows them to find consistent global plans in the presence of changing obstacles without requiring explicit plan reuse. This is demonstrated with simulated experiments where Effort Informed Trees (EIT*) finds shorter median solution paths than the tested reactive planning algorithms and is further validated using Asymptotically Optimal RRT-Connect (AORRTC) on a real-world planning problem on a robot arm.

Revisiting Replanning from Scratch: Real-Time Incremental Planning with Fast Almost-Surely Asymptotically Optimal Planners

TL;DR

Dynamic, obstacle-rich environments challenge traditional replanning by requiring both speed and path quality. The authors propose solving incremental planning as independent, fast ASAO planning problems, using planners like EIT* (and AORRTC for real-world tests) to rapidly generate initial, high-quality intermediate paths and converge toward optimal global solutions without explicit plan reuse, with the global path composed as and cost . In simulated 2D worlds and a real seven-DOF Franka arm, independent ASAO replanning with EIT* achieves higher success rates and shorter median global path lengths than plan-reuse planners such as RRTX and other ASAO methods, even at tight budgets like 50 ms. This approach simplifies implementation by avoiding obstacle-change tracking, demonstrates real-time replanning capabilities in dynamic scenarios, and opens avenues for combining with obstacle motion predictions and richer sensing in future work.

Abstract

Robots operating in changing environments either predict obstacle changes and/or plan quickly enough to react to them. Predictive approaches require a strong prior about the position and motion of obstacles. Reactive approaches require no assumptions about their environment but must replan quickly and find high-quality paths to navigate effectively. Reactive approaches often reuse information between queries to reduce planning cost. These techniques are conceptually sound but updating dense planning graphs when information changes can be computationally prohibitive. It can also require significant effort to detect the changes in some applications. This paper revisits the long-held assumption that reactive replanning requires updating existing plans. It shows that the incremental planning problem can alternatively be solved more efficiently as a series of independent problems using fast almost-surely asymptotically optimal (ASAO) planning algorithms. These ASAO algorithms quickly find an initial solution and converge towards an optimal solution which allows them to find consistent global plans in the presence of changing obstacles without requiring explicit plan reuse. This is demonstrated with simulated experiments where Effort Informed Trees (EIT*) finds shorter median solution paths than the tested reactive planning algorithms and is further validated using Asymptotically Optimal RRT-Connect (AORRTC) on a real-world planning problem on a robot arm.
Paper Structure (22 sections, 2 equations, 5 figures, 1 table, 1 algorithm)

This paper contains 22 sections, 2 equations, 5 figures, 1 table, 1 algorithm.

Table of Contents

  1. Introduction
  2. Literature Review
  3. Predictive Approaches
  4. Reactive Approaches
  5. Independent Replanning
  6. Methodology
  7. Optimal Planning
  8. Incremental Planning
  9. Independent Incremental Planning Approach
  10. Experiments
  11. Simulated Experiments
  12. Random Rectangles
  13. Wall Gap
  14. Double Enclosure
  15. Results
  16. ...and 7 more sections

Figures (5)

  • Figure 1: A Franka Research 3 arm replanning around a dynamic obstacle by independently solving new planning problems each time the obstacle information changes. The VAMP thomason_motions_microseconds_2024 implementation of AORRTC wilson2025aorrtc plans fast enough in a CAPT Ramsey-RSS-24 obstacle map that it is able to replan without stopping the arm.
  • Figure 2: A visualization of the paths traveled on 50 attempts of the incremental planning problem done by EIT*, RRT*, RRT-Connect and RRTX with a planning budget of 0.1 s. The hue of each path is determined by the the order it was run in the experiment. The length of a planner's global solution is the overall length of the path traveled. EIT* found the smallest median global solution length and had a 100% success rate. RRT* followed similar paths to EIT* due to its ASAO properties, when it was successful, but only succeeded 6% of the time. RRT-Connect also maintained a 100% success rate but had a longer median path length than EIT*. RRTX had a 0% success rate due to the overhead of optimally rewiring its solution trees.
  • Figure 3: A visualization of the intermediate paths found by EIT* with a sensor range of 0.1 and a planning budget of 0.1s. The gray area is unsensed and assumed to be free space by the planner. The dark green dot represents the starting point of the query, and the red dot represents the goal. The green path represents the path traveled so far. The number of circles gives a representation of the number of queries of the planner.
  • Figure 4: The success rates of all planners across 100 runs on three of the simulated worlds. A visualization of the world as well as a solution found by EIT* is shown in (a), (d) and (g). Each of the planners were tested with a planning budget of $0.1s$ and $0.05s$. The success rate at the $n^\text{th}$ query is defined by the percentage of planners that are able to find a solution to the $n^\text{th}$ intermediate problem within the planning budget. Success rates are carried forward, i.e., if a planner fails on a query it is also considered to fail on all subsequent queries. The independent ASAO approach with EIT* maintained the highest success rate across every world and every planning budget while finding the shortest global solution (Table \ref{['tab:results']}).
  • Figure 5: The combined success rate of each planner on 100 different random worlds for planning budgets of 0.01 s, 0.05 s and 0.1 s. EIT* had the highest total success rate at each planning budget and RRTX was not able to consistently solve the problems in the available time.