Dual Control Reference Generation for Optimal Pick-and-Place Execution under Payload Uncertainty

TL;DR

This work tackles robot manipulation under unknown payload dynamics by formulating a dual-control reference-generation framework that jointly optimizes task execution and parameter learning. It introduces two approaches: robust-optimization-based trajectory design that accounts for closed-loop adaptation, and optimality-loss-based design that minimizes the expected degradation from parameter uncertainty, both leveraging Fisher information concepts. A simplified sensitivity-based variant without online adaptation extends the method to fixed-parameter controllers. Empirical validation on a 7-DoF manipulator performing pick-and-place under payload uncertainty demonstrates improved final accuracy and faster parameter identification, highlighting practical benefits for responsive, data-efficient manipulation with uncertain dynamics.

Abstract

This work addresses the problem of robot manipulation tasks under unknown dynamics, such as pick-and-place tasks under payload uncertainty, where active exploration and(/for) online parameter adaptation during task execution are essential to enable accurate model-based control. The problem is framed as dual control seeking a closed-loop optimal control problem that accounts for parameter uncertainty. We simplify the dual control problem by pre-defining the structure of the feedback policy to include an explicit adaptation mechanism. Then we propose two methods for reference trajectory generation. The first directly embeds parameter uncertainty in robust optimal control methods that minimize the expected task cost. The second method considers minimizing the so-called optimality loss, which measures the sensitivity of parameter-relevant information with respect to task performance. We observe that both approaches reason over the Fisher information as a natural side effect of their formulations, simultaneously pursuing optimal task execution. We demonstrate the effectiveness of our approaches for a pick-and-place manipulation task. We show that designing the reference trajectories whilst taking into account the control enables faster and more accurate task performance and system identification while ensuring stable and efficient control.

Figures (4)

• Figure 1: Comparison between tracking a nominal reference trajectory (top) and a dual-control trajectory generated by our framework (bottom), both shown in red. The task is to transport a payload (grey box) with unknown parameter $\theta$ from an initial position (green dot) to a target position (blue dot). The nominal trajectory's insufficient excitation of the system dynamics leads to poor parameter estimation and a consequent large deviation from the target. In contrast, the proposed dual-control trajectory actively excites task- and controller-specific inertial parameters during task execution, yielding improved parameter estimates, reduced uncertainty, thereby enabling precise tracking to the target position.
• Figure 2: Final pose error across controllers and trajectory generation methods. Robust Optimization (RO) and Optimality Loss (OL) achieve lower and more consistent errors than the nominal and FIM-based baselines.
• Figure 3: Comparison of three reference trajectories in joint space. (a) nominal trajectory lacks excitation, (b) FIM-based trajectory aggressively excites dynamics but is task-agnostic, and (c) robust optimization balances parameter excitation with precise task execution.
• Figure 4: Final pose errors comparing the sensitivity-based trajectory generated with the RO formulation against the nominal trajectory. The RO-based trajectory achieves significantly lower error by explicitly accounting for parameter uncertainty during planning.

Theorems & Definitions (2)

• Remark 1
• Remark 2