Biologically-inspired Techniques for Solving Constrained Multi-objective Optimization Problems

Description : Most real-world problems involve complex optimization with various conflicting specifications, which cannot be solved without advanced techniques. In addition, most of these problems are also commonly imbued by linear/nonlinear equality/inequality constraints, which only make them more difficult to solve, even by most renowned numerical optimization techniques. Hence, the major goal in solving this problem is to obtain set of feasible tradeoff solutions that satisfy the various optimization goals. The past decades have seen the development of powerful optimization techniques which draws inspiration from nature. Among those are a group of stochastic optimization algorithms inspired by Darwin's theory of evolution, collectively known as Evolutionary Algorithms (EAs). EAs have shown considerable success in locating the global optimum solution of optimization problems that may often be characterized by high dimensional, non-separable, multi-modal, constrained, and discontinuous/non-differentiable fitness landscape. In this project, we will examine and propose new constraint handling techniques that can be embedded into evolutionary algorithms. Particularly, a meta-Lamarckian approach will be explored to learn adaptively from the problem the best manner of constraint handling during an evolutionary optimization. Deliverables: Part 1: -Path planning for robots. i.e. Unmanned Vehicles. (Unmanned aerial vehicles, unmanned ground vehicles). - In 2D, using one of evolutionary algorithm concept. (Genetic algorithm). -Single-objective Optimization (Shortest Path). - Solve some constraints. E.g. Minimum collision Part 2: Extension of path one. -Path planning for robots. i.e. Unmanned Vehicles. (Unmanned aerial vehicles, unmanned ground vehicles). -In 3D, with the invention of new algorithm or concept.(hybrid of two types of evolutionary algorithm or compensation of drawbacks of any one concept. ) -Multi-Objective optimization. (More than one factor- speed, shortest path, etc) - Solve some constraints. (E.g. Minimum collision.) Platform : Python. Please give description of each class/methods as comments. If you are getting distance from user for different routes to choose the best one, or any other user input, please provide a ‘read me’ text on how to use the system.