Multi Objective Optimization of Hand Exoskeleton

Multi-Objective Optimization of Hand Exoskeleton with Evolutionary Algorithms

We propose to optimize the link lengths of an underactuated hand exoskeleton with 9 of twelve link lengths. We have three different objective (i) maximum force transmission on the finger joint, (ii) minimizes the difference between two forces that are acting on the sliders, (iii) minimizes the need for a longer actuator displacement by using Evolutionary Algorithms. In this project we use 2 different Evolutionary Algorithms (Genetic Algorithm and Big Bang-Big Crunch Algorithm) and 2 different Methods (Elitist Non-Dominated Sorting and Strength Pareto) of them. We implemented NSGA-II and SPEA2 we derived from these algorithm NSBBBC and SPBBBC.

Article on ArXiv: https://arxiv.org/abs/2408.07384

Follow these steps to run the algorithm:

1. Open the folder in MATLAB.
2. Arrange the finger that you want to optimize from RUN_OPT_HANDEXO file finger part
3. Add the file path that algorithm you will use
4. Choose the optimization method by changing algorithm_name variable. For Elitist Non-Dominated Sorting Genetic Algorithm write NSGA; For Elitist Non-Dominated Sorting Big Bang-Big Crunch Algorithm write NSBBBC; For Strength Pareto Genetic Algorithm write SPEA2; For Strength Pareto Big Bang-Big Crunch Algorithm write SPBBBC
5. Arrange the upperbounds and lowerbounds you can change it from RUN_OPT_HANDEXO, UB (Upper Bounds) and LB (Lower Bounds) arrays.
6. For settings (number of generation, number of individuals, number of variables, selection method, crossover method, crossover probability, mutation method, mutation probability, survival method, run number, number of objective functions, arranging objective functions are maximize or minimize, dominance level strong or weak) you can reach them in RUN_OPT_HANDEXO funciton

Authors:

• Barış Akbaş
• Aleyna Söylemez
• Hüseyin Taner Yüksel
• Fabio Stroppa
• Mine Sarac