Optimization of Rocker-Bogie Mechanism using Heuristic Approaches

TL;DR

This paper compares various heuristic algorithms for optimizing the suspension mechanism of planetary rovers, finding Simulated Annealing to be the most effective in terms of computational efficiency and overall fitness.

Contribution

It introduces a comprehensive evaluation of multiple heuristic algorithms for Rocker-Bogie suspension optimization, highlighting Simulated Annealing as the most effective method.

Findings

Simulated Annealing achieved the highest fitness score of 760.

Heuristic algorithms vary in efficiency and effectiveness.

Hyperparameter tuning impacts algorithm performance.

Abstract

Optimal locomotion and efficient traversal of extraterrestrial rovers in dynamic terrains and environments is an important problem statement in the field of planetary science and geophysical systems. Designing a superlative and efficient architecture for the suspension mechanism of planetary rovers is a crucial step towards robust rovers. This paper focuses on the Rocker Bogie mechanism, a standard suspension methodology associated with foreign terrains. After scrutinizing the available previous literature and by leveraging various optimization and global minimization algorithms, this paper offers a novel study on mechanical design optimization of a rovers suspension mechanism. This paper presents extensive tests on Simulated Annealing, Genetic Algorithms, Swarm Intelligence techniques, Basin Hoping and Differential Evolution, while thoroughly assessing every related hyper parameter, to find utility driven solutions. We also assess Dual Annealing and subsidiary algorithms for the aforementioned task while maintaining an unbiased testing standpoint for ethical research. Computational efficiency and overall fitness are considered key valedictory parameters for assessing the related algorithms, emphasis is also given to variable input seeds to find the most suitable utility driven strategy. Simulated Annealing was obtained empirically to be the top performing heuristic strategy, with a fitness of 760, which was considerably superior to other algorithms and provided consistent performance across various input seeds and individual performance indicators.