Skip to content

Plan4ARI

Research M3 Core IPC

Neural Motion Control

The proposed control research combines prescribed performance requirements with adaptive neural compensation for uncertain industrial robot dynamics.

Motivation

Conventional control is powerful but depends on accurate modelling, observers and careful tuning. Industrial robots may also face elasticity, disturbances and configuration changes.

Proposed concept

Plan4ARI investigates Prescribed Performance Control combined with adaptive RBF neural networks to compensate uncertain dynamics while keeping performance requirements explicit.

How it works

Methodology

  1. Dynamic learning model.
  2. Error transformation.
  3. Filtered errors and adaptive control.
  4. Convergence-oriented closed-loop representation.
  5. Storage and reuse of learned weights.

Control design

Performance envelopes

Prescribed performance defines transient and steady-state error envelopes so the controller design has explicit bounds for overshoot, convergence and final accuracy.

Learning-based compensation

Adaptive RBF networks are proposed to compensate uncertain dynamics, elasticity, disturbances, constraints and possible actuator faults.

Expected design advantages

  • Proposed compliance modelling.
  • Expected reduced dependence on full-state observers.
  • Proposed MIMO design without repeated matrix inversion.
  • Expected limited computation requirements.
  • Proposed knowledge reuse through stored learned weights.

ActivityA5

Activity A5 focuses on neural robust motion control, automatic tuning, learned compensation and stress testing under realistic computation limits.

Proposed targets

Project target 01

Automatic tuning.

Project target 02

Reduced tracking error.

Project target 03

Reduced overshoot.

Project target 04

Computation-time stress testing.

Project target 05

Reduced controller-deployment time.

Targets describe planned outcomes, not results already achieved.