DGAA*: A Dynamically Repaired Double-Strand Break Genetic Algorithm with Pheromone Guided A*
Disciplines
Robotics | Systems Architecture | Theory and Algorithms
Abstract (300 words maximum)
Path planning is an integral part of robotics that enhances their function: Mobile robots have an increased range of utility as compared to their stationary counterparts, but what good is their mobility if it is found to be inefficient? A* algorithm is a traversal method in which the immediate best position from the current spot is chosen without regard for the global configuration of the map, ensuring a fast path creation yet not the shortest path. Ant Colony Optimization (ACO) uses multiple agents, known as ants, to traverse the map to look for the best path, leaving pheromone trails whose influence depends on how many times said path is traveled. Genetic Algorithm (GA) is inspired from the biological processes of crossovers, mutations, and selection: an initial set of paths are produced, being subjected to mutations and combining with others to produce offspring with the hope the offspring is better than either of its parents. The best paths from the batch are then taken into the next generation, and this process repeats until either the maximum number of generations are met or the desired result is attained. This project introduces a method that combines the decisiveness of A*, the exploration factor of ACO, and the refinement of a more biologically accurate GA to procure a more structurally sound path planning method. DGAA*, the algorithm in question, is compared with the aforementioned three methods in environments of varying complexity to see if it is a viable alternative to other traversal algorithms. Results show that DGAA*, with the base of GA, both escapes the suboptimal routes A* gets trapped in and has a controlled global exploration factor, to reduce needless searches, guaranteeing that mobile robots using this technique travel efficiently.
Academic department under which the project should be listed
SPCEET - Robotics and Mechatronics Engineering
Primary Investigator (PI) Name
Muhammad Tanveer
Additional Faculty
Cary Chun, Intelligent Robotic Systems, cchun@students.kennesaw.edu
DGAA*: A Dynamically Repaired Double-Strand Break Genetic Algorithm with Pheromone Guided A*
Path planning is an integral part of robotics that enhances their function: Mobile robots have an increased range of utility as compared to their stationary counterparts, but what good is their mobility if it is found to be inefficient? A* algorithm is a traversal method in which the immediate best position from the current spot is chosen without regard for the global configuration of the map, ensuring a fast path creation yet not the shortest path. Ant Colony Optimization (ACO) uses multiple agents, known as ants, to traverse the map to look for the best path, leaving pheromone trails whose influence depends on how many times said path is traveled. Genetic Algorithm (GA) is inspired from the biological processes of crossovers, mutations, and selection: an initial set of paths are produced, being subjected to mutations and combining with others to produce offspring with the hope the offspring is better than either of its parents. The best paths from the batch are then taken into the next generation, and this process repeats until either the maximum number of generations are met or the desired result is attained. This project introduces a method that combines the decisiveness of A*, the exploration factor of ACO, and the refinement of a more biologically accurate GA to procure a more structurally sound path planning method. DGAA*, the algorithm in question, is compared with the aforementioned three methods in environments of varying complexity to see if it is a viable alternative to other traversal algorithms. Results show that DGAA*, with the base of GA, both escapes the suboptimal routes A* gets trapped in and has a controlled global exploration factor, to reduce needless searches, guaranteeing that mobile robots using this technique travel efficiently.