Modified RRT Algorithm Boosts Robot Navigation Efficiency in Three-Dimensional Simulations

A modified path-planning algorithm operating within a three-dimensional simulation framework achieved navigation success rates of up to 98 percent while cutting navigation completion times by up to 17.6 percent relative to standard approaches, according to research conducted by Sangeeth Venu and Muralimohan Gurusamy at the Vellore Institute of Technology. The investigation assessed the performance of the Modified Rapidly-Exploring Random Tree, or Modified RRT, using a simulation-based robotic navigation architecture built on ROS 2 and Gazebo.

The study expanded upon previous evaluations that had tested the algorithm strictly across two-dimensional benchmark maps against alternative planners, including standard RRT, RRT*, A*, and Informed RRT*. The study extended the evaluation to robotic navigation scenarios in three-dimensional simulated environments where the robot, obstacles, and terrain were represented.

Across the simulation framework, the team designed and tested five distinct operational scenarios. These experimental configurations comprised the three-dimensional reconstruction of selected benchmark maps, real-time dynamic obstacle avoidance, dynamic goal adaptation, navigation across simulated warehouse and campus environments, and the traversal of sloped terrain.

To capture environmental observations during navigation runs, the simulation architecture integrated synchronized RGB and depth sensing paired with point-cloud generation. While the ground robot, surrounding obstacles, and ground terrain were rendered fully in three dimensions, the global path planner functioned within the horizontal x-y plane, allowing researchers to evaluate planar planning execution alongside simulated multimodal perception.

All planners were analyzed under a shared evaluation protocol that enforced an identical costmap, collision checker, stopping budget, and parameter-tuning methodology. Under these standardized conditions, Modified RRT generated paths that were 10 to 15 percent shorter than those produced by standard RRT and 2 to 10 percent shorter than those from the next-best planner, Informed RRT*. These trajectory improvements, however, came with a tradeoff, as the modified algorithm required higher planning time than standard RRT in most evaluated scenarios.

Navigation safety metrics also showed substantial gains under ideal sensing conditions. When evaluating near-collision rates, Modified RRT lowered incidents by up to 74 percent in structured simulation settings and by 63 percent in unstructured settings compared to standard RRT, demonstrating the effectiveness of its sampling mechanisms during trajectory generation.

The research demonstrated, however, that simulated perception degradation notably diminished these safety gains. When the framework introduced real-world sensor imperfections—including depth noise, measurement dropouts, communication latency, and registration errors—near-collision rate reductions dropped to between 45 and 61 percent in structured environments and between 40 and 52 percent in unstructured spaces, underscoring the direct impact of sensor fidelity on autonomous robot navigation.

Published as an original research article in the Field Robotics section of Frontiers in Robotics and AI, the peer-reviewed study was formally received in August 2026 and accepted in October 2026. Sangeeth Venu and corresponding author Muralimohan Gurusamy emphasize that their framework delivers an integrated platform for jointly analyzing planar global path planning, robot navigation execution, and multimodal sensor simulation in three-dimensional environments.

Subject of Research: Simulation-based path planning and navigation evaluation for mobile ground robots

Article Title: Simulation-based evaluation of a modified RRT algorithm for ground-robot navigation in 3D environments

Article References: Venu, S., & Gurusamy, M. (2026). Simulation-based evaluation of a modified RRT algorithm for ground-robot navigation in 3D environments. Frontiers in Robotics and AI, 13, Article 1981973. https://doi.org/10.3389/frobt.2026.1981973

Image Credits: AI Generated

DOI: 10.3389/frobt.2026.1981973

Keywords: Modified RRT, Rapidly-Exploring Random Tree, ROS 2, Gazebo, robot navigation, path planning, point cloud, RGB-D perception, dynamic obstacle avoidance, dynamic goal adaptation, slope traversal, warehouse navigation, campus navigation, Informed RRT*, three-dimensional simulation, near-collision rate, depth sensing, planar global path planning