Abstract Bioinspired optimization algorithms, derived from biological processes such as bacterial foraging and swarm behavior, have shown increasing potential in addressing high-dimensional, nonlinear, and time-varying problems in engineering. Their integration into robotic control architectures enables the development of adaptive, model-flexible schemes that are robust to uncertainty and real-time constraints. Anthropomorphic manipulator robots, widely used in manufacturing and medical applications, require high-performance motion control under structural uncertainty, dynamic perturbations, and limited sensing. This paper proposes a unified and robust control scheme that integrates three key components: (i) a bacterial foraging optimization algorithm for offline initialization of controller weights, (ii) B-spline artificial neural networks for online adjustment of adaptive control gains, and (iii) a robust motion control law based on integral reconstruction theory, which eliminates the need for velocity measurement or full dynamic models and avoids high-gain compensation. This architecture overcomes several limitations of classical model-based, PID, or adaptive-only approaches by combining learning, compensation, and optimization within a scalable framework. The proposed method is validated through multiple simulation studies involving anthropomorphic manipulators with documented physical parameters and subjected to varying disturbances. Comparative analysis demonstrates superior tracking precision, reduced control effort, and faster convergence dynamics. These results confirm the practical viability of the proposed framework for motion control in dynamically uncertain robotic platforms.
Galvan-Perez et al. (Tue,) studied this question.