Path Smoothing Strategy Based on Metaheuristic Algorithms for Probabilistic Foam Luís Bruno Pereira do Nascimento Department of Computer and Electrical Engineering Federal University of Rio Grande do Norte Natal, RN, Brazil lbruno@ufrn.edu.br Vitor Gaboardi dos Santos Department of Computer and Electrical Engineering Federal University of Rio Grande do Norte Natal, RN, Brazil vitorgaboardi@ufrn.edu.br Diego da Silva Pereira Center of Innovation in Computer Technologies Federal Institute of Rio Grande do Norte Parnamirim, RN, Brazil diego.pereira@ifrn.edu.br Daniel Henrique Silva Fernandes Department of Computer and Electrical Engineering Federal University of Rio Grande do Norte Natal, RN, Brazil eng.danielhsf@gmail.com Pablo Javier Alsina Department of Computer and Electrical Engineering Federal University of Rio Grande do Norte Natal, RN, Brazil pablo@dca.ufrn.br ABSTRACT The probabilistic Foam method (PFM) is a sampling-based path planning algorithm that ensures a feasible path bounded by a safe region. This method is ideal for assistive robotics applications, which demands a high level of safety, such as performing a motion by an active exoskeleton. However, PFM generates non-smoothed paths, which results in non- anthropomorphic movements. Thus, this paper presents some optimization strategies based on metaheuristics to smooth the paths generated by PFM. Simulated experiments were performed using the Harmony Search Algorithm, and Ge- netic Algorithm and they were applied to an exoskeleton to overcome an obstacle. Results show that our proposed ap- proach is capable of smoothing paths for this application, which resulted in more anthropomorphic motions. Keywords Assistive Robotics; Active Exoskeleton; Path Smoothing; Genetic Algorithm; Harmony Search Algorithm 1. INTRODUCTION Applications in autonomous robotics are in constant ex- pansion due to the fast technological advances. Moreover, Path planning is one of the most relevant issues related to this field of research, since a planner calculates a set of poses Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and the full citation on the first page. To copy otherwise, or republish, to post on servers or to redistribute to lists, requires prior specific permission and/or a fee. and orientations that a robot must perform to accomplish a specific task, such as moving from an initial to a goal con- figuration without colliding with obstacles along the path [11]. Among the several path planning methods, the sampling- based planners are the most promising path planning cate- gory since they can find feasible paths using random sam- ples from the free configuration space. Sampling-based Path planners such as Probabilistic Roadmaps (PRM) [9] and Rapidly-Exploring Random Tree (RRT) [12] are the most known. The sampling-based path planners usually use few computer resources [13] and can be applied for robots with many degrees of freedom [14, 1]. Among different types of robots, assistive robots are de- vices that perform actions that benefit people with some kind of disability. For instance, Ortholeg [15] is a project of a lower limb active exoskeleton that helps physically chal- lenged people in the walking experience. The Ortholeg ex- oskeleton was designed with the concept of transparency, which can be defined as the capability of the device to make the walking experience as natural as possible, both for the user and the people around him [15]. Figure 1 illustrates the exoskeleton used in this study. In [3], a path planning algorithm called Probabilistic Foam Method (PFM) [20] was applied to the Ortholeg to provide safe movements for the task of overcoming a single obstacle. The planner PFM guarantees a high clearance path from the obstacles for safe maneuverability. A problem with the approach presented in [3] is that paths generated by PFM are non-smooth, which implies that the motion performed by the exoskeleton has a non-anthropomorphic pattern. Path smoothing problems that involves many constraints can be solved with optimization techniques, as shown in [19]. In this way, some metaheuristics have been used as path smoothing optimization, as can be seen in the works [8] and [2], where Genetic Algorithm (GA) and Particle Swarm Revista de Sistemas e Computação, Salvador, v. 10, n. 1, p. 51-58, jan./abr. 2020 http://www.revistas.unifacs.br/index.php/rsc