Gav Verma

Autonomous path finding while considering multiple objectives or criteria is a problem that has been tackled by many with varying results. Self-driving vehicles traversing through a desert or hostile environment need to consider not only avoiding uncrossable obstacles but also prioritize their path through different types and levels of terrain. Unmanned Aerial Vehicles (UAV) used for military applications (such as drones) need to consider both the shortest distance to conserve fuel as well as… Show more

Autonomous path finding while considering multiple objectives or criteria is a problem that has been tackled by many with varying results. Self-driving vehicles traversing through a desert or hostile environment need to consider not only avoiding uncrossable obstacles but also prioritize their path through different types and levels of terrain. Unmanned Aerial Vehicles (UAV) used for military applications (such as drones) need to consider both the shortest distance to conserve fuel as well as the maximal proximity from enemies in an adaptive and dynamic manner. Simultaneous optimization of conflicting objectives is a difficult yet very real problem, and the Genetic Algorithm (GA) is a metaheuristic that is well suited for such a class of problems. Simply using weights or a combination of single solutions to balance conflicting objectives is tedious and does not scale well, whereas a decision maker for a Pareto optimal solution with the use of an evolutionary approach such as GA is not only far more practical, but also scales well directly with real-life problems. Show less

This paper proposes a real-time algorithm for guiding an agent through a stochastic and adversarial environment without computing the entire path to the goal node in one step. We incorporate a proximity radius and the Voronoi Diagram coupled with a heuristic based on weighted nodes to avoid adversaries and reach the goal node. Our implementation for a dynamic environment rivals existing algorithms for static environments, with high success ratios found when running simulations against even a… Show more

This paper proposes a real-time algorithm for guiding an agent through a stochastic and adversarial environment without computing the entire path to the goal node in one step. We incorporate a proximity radius and the Voronoi Diagram coupled with a heuristic based on weighted nodes to avoid adversaries and reach the goal node. Our implementation for a dynamic environment rivals existing algorithms for static environments, with high success ratios found when running simulations against even a very high number of adversaries. Show less