Tracking control of redundant mobile manipulator: An RNN based metaheuristic approach
In this paper, we propose a topology of Recurrent Neural Network (RNN) based on a metaheuristic optimization algorithm for the tracking control of mobile-manipulator while enforcing nonholonomic constraints. Traditional approaches for tracking control of mobile robots usually require the computation...
Gespeichert in:
Veröffentlicht in: | Neurocomputing (Amsterdam) 2020-08, Vol.400, p.272-284 |
---|---|
Hauptverfasser: | , , , |
Format: | Artikel |
Sprache: | eng |
Schlagworte: | |
Online-Zugang: | Volltext |
Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
Zusammenfassung: | In this paper, we propose a topology of Recurrent Neural Network (RNN) based on a metaheuristic optimization algorithm for the tracking control of mobile-manipulator while enforcing nonholonomic constraints. Traditional approaches for tracking control of mobile robots usually require the computation of Jacobian-inverse or linearization of its mathematical model. The proposed algorithm uses a nature-inspired optimization approach to directly solve the nonlinear optimization problem without any further transformation. First, we formulate the tracking control as a constrained optimization problem. The optimization problem is formulated on position-level to avoid the computationally expensive Jacobian-inversion. The nonholonomic limitation is ensured by adding equality constraints to the formulated optimization problem. We then present the Beetle Antennae Olfactory Recurrent Neural Network (BAORNN) algorithm to solve the optimization problem efficiently using very few mathematical operations. We present a theoretical analysis of the proposed algorithm and show that its computational cost is linear with respect to the degree of freedoms (DOFs), i.e., O(m). Additionally, we also prove its stability and convergence. Extensive simulation results are prepared using a simulated model of IIWA14, a 7-DOF industrial-manipulator, mounted on a differentially driven cart. Comparison results with particle swarm optimization (PSO) algorithm are also presented to prove the accuracy and numerical efficiency of the proposed controller. The results demonstrate that the proposed algorithm is several times (around 75 in the worst case) faster in execution as compared to PSO, and suitable for real-time implementation. The tracking results for three different trajectories; circular, rectangular, and rhodonea paths are presented. |
---|---|
ISSN: | 0925-2312 1872-8286 |
DOI: | 10.1016/j.neucom.2020.02.109 |