Reinforcement learning-based mobile robot navigation

Abstract

In recent decades, reinforcement learning (RL) has been widely used in different research fields ranging from psychology to computer science. The unfeasibility of sampling all possibilities for continuous-state problems and the absence of an explicit teacher make RL algorithms preferable for supervised learning in the machine learning area, as the optimal control problem has become a popular subject of research. In this study, a system is proposed to solve mobile robot navigation by opting for the most popular two RL algorithms, Sarsa(λ) and Q(λ). The proposed system, developed in MATLAB, uses state and action sets, defined in a novel way, to increase performance. The system can guide the mobile robot to a desired goal by avoiding obstacles with a high success rate in both simulated and real environments. Additionally, it is possible to observe the effects of the initial parameters used by the RL methods, e.g., λ , on learning, and also to make comparisons between the performances of Sarsa(λ) and Q(λ) algorithms.

Mobile Robot Path Optimization Technique Based on Reinforcement Learning Algorithm in Warehouse Environment

“…In the mobile robot framework structure in Figure 6, the agent is controlled by the supervisor controller, and the robot controller is executed by the agent, so that the robot operates in the warehouse environment and the necessary information is observed and collected [25][26][27].…”

Reinforcement learning based adaptive power pinch analysis for energy management of stand-alone hybrid energy storage systems considering uncertainty

“…Typically, an agent will evaluate a state, and will then undertake an action either in an exploitative or exploratory manner thereafter and finally will receive an instant reward, while transitioning to a new state. Q-learning has tremendous success in robotics, especially in mobile robot navigation and obstacle avoidance [60,61]. In [62] the Dyna AI architecture was proposed to integrate both learning, and experience, based on online planning, as well as reactive execution in a stochastic environment.…”

Learning Navigation Behaviors End-to-End With AutoRL

“…Both behaviors need to produce dynamically feasible trajectories that are robust to many kinds of noise and inaccuracies, rely only on observations coming from primitive sensors, and avoid unexpected obstacles. In this paper, we present a reliable method to implement these navigation behaviors by learning 1 Google AI, Mountain View, CA 94043, USA lewispro,faust,mfiser,centaur@google.com * Authors contributed equally. end to end polices that directly map sensors to controls, and we show that these policies transfer from simulation to physical robots and new environments while robustly avoiding obstacles.…”