Proximal policy optimization based dynamic path planning algorithm for mobile robots

Abstract

For the scenario where the overall layout is known and the obstacle distribution information is unknown, a dynamic path planning algorithm combining the A* algorithm and the proximal policy optimization (PPO) algorithm is proposed. Simulation experiments show that in all six test environments, the proposed algorithm finds paths that are on average about 2.04% to 5.86% shorter compared to the state-of-the-art algorithms in the literature, and reduces the number of training epochs before stabilization from tens of thousands to about 4000.

A Review of Deep Reinforcement Learning Algorithms for Mobile Robot Path Planning

“…Examples of such applications are vast, spanning from cleaning and delivery tasks [2], assistance in shopping malls [3], patrolling endeavors [4], to more complex scenarios such as autonomous driving, assisting the elderly in their homes [5], and executing search and rescue missions [6,7]. Moreover, sectors such as mining, household services, agriculture, and automated production can greatly benefit from these techniques [8].…”

“…To deal with pedestrians, the reward function includes a penalty if the agent is too close to any obstacle. In [8], the authors combined PPO with a traditional A* search algorithm for global path planning. The model performed better than EPRQL, DBPQ, DDQNP, and DWAQ.…”

“…With the successful integration of robotic arms in manufacturing and assembling industries [1], there is a lot of research going on in the field of mobile robotics, especially autonomous mobile robots. Autonomous mobile robots find their use in a plethora of activities such as cleaning, delivering [2], shopping mall assisting [3], patrolling [4], selfdriving cars, elderly home helping robots [5], surveying, search and rescue [6,7], automatic production, mining, household services, agriculture, and other fields [8]. To carry out their intended tasks, these robots should be able to reach their target by following a collision-free path.…”

A novel fixed‐time ZNN model and application to robot trajectory tracking

“…Given information on redundant robots and actuator states, the kinematic model enables the determination of the endeffector's position. Many scholars have made great achievements in robot motion planning problems [13][14][15]. Trajectory tracking essentially poses as an inverse kinematic problem, solvable through diverse methodologies.…”