[mrpt-nav]

Overview

Autonomous navigation, path planning

Library mrpt-nav

This library is part of MRPT and can be installed in Debian-based systems with:

sudo apt install libmrpt-nav-dev

Overview

mrpt-nav provides reactive and planned navigation for mobile robots. The architecture is layered:

CAbstractNavigator               ← base state machine (IDLE / NAVIGATING / SUSPENDED / NAV_ERROR)
  └── CWaypointsNavigator        ← waypoint sequencing on top of single-goal navigation
        └── CAbstractPTGBasedReactive   ← TP-Space reactive core
              ├── CReactiveNavigationSystem    ← 2-D robots
              └── CReactiveNavigationSystem3D  ← multi-level 3-D robots

Reactive navigation

Navigator state machine

mrpt::nav::CAbstractNavigator drives a state machine with four states (mrpt::nav::CAbstractNavigator::TState):

State

Meaning

No active navigation goal Actively navigating toward target Navigation paused (call An unrecoverable error occurred; query

Code Meaning ==== =====================================

No error Robot stopped due to safety violation Timeout: robot not approaching target Unclassified exception

Class Algorithm ===== =========================================================================================

Virtual Force Fields — repulsive forces from obstacles + attractive force toward target Nearness Diagram — gap-based obstacle avoidance (Minguez & Montano, 2004) Full-evaluation scoring across all TP-Space directions

Value Meaning ===== ===============================

Straight free path to target Narrow gap selected Wide gap selected No traversable gap; robot stops

Class Robot kinematics ===== =========================================================

Differential drive — circular arc Differential drive — circular arc + straight Differential drive — two circular arcs (same direction) Differential drive — two arcs + straight Differential drive — trapezoidal steering Holonomic robot with velocity blending

(default) Allow reverse motions to escape near-collision Reject any motion when robot is already in near-collision

Class Algorithm ===== ==============================================

Simple 2-D A* on an occupancy grid RRT planner in SE(2) using PTG-space expansion