abstract
- This work presents two computationally low-cost optimization algorithms to determine parameters for the Dual-SLIP model producing stable walking gaits. While previous approaches use exhaustive or random searches and techniques requiring long simulations typically running until a fall is detected or an end condition is met, the approaches presented here only require the simulation of a single step thus reducing the computation time. The proposed optimization algorithms are based on Genetic Algorithms (GA) and Particle Swarm Optimization (PSO). Also, an Artificial Neural Network (ANN) and a Deep Neural Network (DNN) are presented to predict the stability and the number of steps a given set of parameters will be able to produce. After training, these algorithms do not require any simulation of the walking model to yield their result. Simulation results show the efficacy of the proposed algo-rithms.