How to deal with the complexity in robotic systems?

1. PATH PLANNING FOR ROBOTS

Considering that robots often perform tasks in workspaces in the presence of obstacles, it is required to plan collision-free motion based on the workspace geometry. This is mainly classified as offline planning or online planning in accordance with the workspace geometry information^[4]. This topic has attracted much attention in the robotics field, and the existing path planning methods are mainly divided into three categories: graph-based, sampling-based, and intelligent methods. For instance, the A* algorithm is a traditional graph-based search method based on known environmental information for the path planning method.

The paper entitled An improved A star algorithm for wheeledrobots path planning with jump points searchand pruning method^[5] proposes an effective method for global path planning of wheeled robots based on the A* algorithm. Within this graph-based search method, a jump point search method is adopted by adding only the nodes with special properties to the open list during the process of node expansion. Besides, simulation results are provided to verify the effectiveness of the proposed algorithm.

2. MOTION CONTROL IN ROBOTS

The motion control of robots is related to the movement of objects and can be defined on the basis of an integration of robot automation, robot dynamics, actuation mechanism, and embedded sensors in real time. Emerging technologies for robot motion control deal with the complexity of robot tasks, safety, and environments. Within the context of offshore robotics, the paper entitled Reinforcement learning-basedcontrol for offshore crane load-landing operations^[6] studies the motion control of the offshore crane system, such as the load-landing or -lifting operations in interconnection with a vessel under environmental conditions. During the crane motion, the impact between the loads and the vessels is to be minimized and controlled to avoid serious injuries and extensive damage.

Owing to both system and environment complexities, artificial intelligence and machine learning algorithms have been successfully applied for model learning in robotic control systems. Therefore, machine learning algorithms could play a crucial role in multiple offshore crane operations, including load-landing operations. Specifically, the authors^[6] attempted to utilize reinforcement learning (RL) algorithms for crane motion control. More specifically, they used the Q-learning algorithm to develop optimal control sequences for the offshore crane’s actuators to minimize the impact velocity between the crane’s load and the moving vessel.

Within the context of classical robot control methodologies, it is known that impedance control or admittance control is suitable for stiff or soft environments, respectively. Therefore, a hybrid impedance/admittance control mechanism can provide an adaptive behavior to the robot in order to interact with different environments. The paper entitled Improved impedance/admittanceswitching controller for the interaction with a variable stiffness environment^[7] aims to satisfy the continuity of the interaction force in the switching from impedance to admittance control when a feedforward velocity term is present. In addition, the switching parameters are adapted to improve the performance of the hybrid control framework to better exploit the properties of both impedance and admittance controllers.

In summary, according to the above-mentioned discussions, dealing with the complexity in robotic systems is highly related to the amount of information that is needed to describe robot dynamics in interaction with the environment or humans.