Title :
A comparative study on the path length performance of maze-searching and robot motion planning algorithms
Author :
Lumelsky, Vladimir J.
Author_Institution :
Dept. of Electr. Eng., Yale Univ., New Haven, CT, USA
fDate :
2/1/1991 12:00:00 AM
Abstract :
A number of existing maze-searching and robot motion planning algorithms are studied from the standpoint of a single performance criterion. The main motivation is to build a framework for selecting basic planning algorithms for autonomous vehicles and robot arm manipulators that operate in an environment filled with unknown obstacles of arbitrary shapes. In choosing an appropriate criterion, it is noted that besides convergence, minimizing the length of generated paths is a single major consideration in planning algorithms. In addition, since no complete information is ever available, optimal solutions are ruled out. Accordingly, the performance criterion is defined in terms of the upper bound on the length of generated paths as a function of the maze perimeter. The comparison shows that the special structure of graphs that correspond to planar environments with obstacles actually makes it possible to exceed the efficiency of general maze-searching algorithms
Keywords :
graph theory; mobile robots; optimisation; planning (artificial intelligence); search problems; artificial intelligence; autonomous vehicles; maze-searching; motion planning algorithms; optimisation; path length performance; path planning; robot; Automata; Computational complexity; Convergence; Graph theory; Motion planning; Path planning; Robot motion; Robotics and automation; Shape; Vehicles;
Journal_Title :
Robotics and Automation, IEEE Transactions on