On reduction of oscillations in target tracking by artificial potential field method

Abstract

Artificial potential field is a powerful method of navigation for mobile robots in a cluttered environment. Despite the advantages that it offers, this method is not free from local minima and oscillation problems. This paper addresses the oscillation problem near the obstacles and in narrow passages. A comparative study has been made between the traditional gradient descent technique and second order methods and it has been shown that the Levenberg-Marquardt algorithm improves upon the oscillation problem and generates smoother trajectories in fewer steps.

Switching formation strategy with the directed dynamic topology for collision avoidance of a multi‐robot system in uncertain environments

“…12 a ). Biswas and Kar [31] state that swings in the trajectories associated with APF have not yet been wholly cancelled out, especially in narrow passages between a cluster of obstacles. Additionally, the vehicles with APF slightly oscillate around their goal.…”

“…A potential field method was used in [16, 30] to achieve inter‐unmanned aerial vehicle (UAV) collision avoidance using a virtual repulsive force derived from the negative gradient of a potential function. The force calculating method may cause oscillations in the trajectory or even getting stuck due to local minima [31]. These observations have inspired us to conduct the research work presented in this paper.…”

“…Using this new technique, mobile robots are required not only to change their orders within the L–F architecture but also to shift their positions to generate a queuing pattern. Furthermore, the proposed control approach can be perfectly combined with our previous works in [32] to generate a comprehensive collision avoidance method in any circumstances while maintaining the robots' given formation and reaching the goal. • We propose a new way of determining the magnitude and direction of the repulsive force, using the region of intersection between virtual circles surrounding the robots to escape local minima and oscillation problems, which is a drawback of the artificial potential field (APF) control theory [31, 34]. • We conduct a stability analysis of the proposed control system based on the internal stability of SNI agents and SNI consensus controllers with their communication links. • We evaluate the efficacy of the proposed NI control system, not only in the form of numerical simulations but also in real robot experiments in the absence or presence of exogenous disturbances (e.g.…”

Robot Path Planning Agent for Evaluating Collaborative Machine Behavior

“…However, there are potential cases that the robot moves across the complex potential field area at a big pace of movement and positioning errors occur due to poor radio propagation. Relevant researchers have shown that when high motion speed and GPS positioning errors are taken into account, the trajectory of the robot is likely to shake violently [15][16]. As shown in Figure 1, when the robot moves between obstacles or in the aisle, the sharp change of direction problem is extremely serious.…”

“…In work of [16] the traditional gradient descent technique is compared with the second-order optimization method, and its results show that the Levenberg-Marquardt algorithm can improve the jitter problem and generate smoother trajectory with fewer steps.…”