TL;DR
This paper introduces a novel Incremental Nonlinear Dynamic Inversion (INDI) control strategy for hydraulic hexapod flight simulator motion systems. By leveraging sensor feedback and requiring less precise model information, INDI significantly improves force tracking performance and robustness to uncertainties compared to traditional methods, leading to more accurate and reliable flight simulation.
What Problem Does This Paper Address?
Controlling hydraulic hexapod flight simulators presents a significant challenge due to the inherent nonlinear dynamics of hydraulic actuators and the complex interactions within the six-degree-of-freedom (DOF) Stewart platform. Traditional control methods often struggle with model uncertainties (like valve overlaps, oil leakage, and temperature variations) and the fact that hydraulic actuators are not direct force generators. This leads to reduced tracking accuracy and robustness. The paper aims to address this by developing a more robust and less model-dependent inner-loop force control strategy that can be integrated into a cascaded control system for the entire motion platform.
Methods
The core of the paper is the application of Incremental Nonlinear Dynamic Inversion (INDI) for the inner-loop force control of individual hydraulic actuators. Unlike traditional Nonlinear Dynamic Inversion (NDI) which requires a detailed system model to compute the total control input, INDI calculates the increment of the control input based on state derivative measurements. This is achieved by approximating the system dynamics using a first-order Taylor expansion and then designing a control law that linearizes the system's behavior.
Key aspects of the INDI method presented:
- Sensor-Based: It relies on state derivative measurements (specifically, pressure difference derivatives) rather than a precise model of the hydraulic dynamics.
- Incremental Control: The control input is updated incrementally, making it less sensitive to model and parameter uncertainties.
- Virtual Control: The INDI controller transforms the nonlinear system into a linear one with a "virtual control input" ($v$), allowing for simpler linear controller design.
- Filtering: A second-order low-pass filter is used to process pressure difference measurements, mitigating noise amplification from differentiation. A technique is employed to synchronize the filtered measurement with the control input memory.
The overall control system is cascaded, with an outer-loop motion controller (using feedforward based on inverse dynamics and PD feedback) generating reference forces, and the proposed INDI controller acting as the inner-loop force controller for each hydraulic actuator. A traditional NDI-based Cascade AP (CdP) controller is used as a benchmark for comparison.
Datasets
Not discussed in the paper. The simulations are performed on a well-validated, fully nonlinear simulation model of the SIMONA Research Simulator (SRS) at TU Delft.
Key Insights and Ideas
- INDI's Robustness to Model Uncertainty: The most significant insight is INDI's inherent robustness. By focusing on the change in control input and relying on sensor feedback, it bypasses the need for highly accurate hydraulic actuator models, which are notoriously difficult to obtain and maintain. This is crucial for systems with varying parameters like oil properties and valve characteristics.
- Sensor-Based Control Advantage: The reliance on pressure difference derivatives makes INDI particularly well-suited for hydraulic systems where pressure measurements are readily available and can be used to infer force.
- Decoupling Hydraulic Dynamics: The cascaded control structure, with INDI handling the inner-loop force control, effectively decouples the complex hydraulic dynamics from the overall motion control problem, simplifying the design and improving performance.
- Improved Tracking Performance: The INDI controller demonstrably leads to better force and position tracking compared to the traditional CdP controller, especially under nominal conditions.
- Trade-off with State Measurement Quality: While INDI reduces model dependency, it increases reliance on the accuracy of state measurements. However, the paper suggests that filtering techniques can effectively manage sensor noise.
Key Results
- Nominal Performance: INDI achieved significantly lower root-mean-squared (RMS) force tracking errors (e.g., 0.0034 vs. 0.016 for actuator 1) and position errors compared to the CdP controller.
- Robustness to Parameter Mismatch: When faced with a 50% and 20% mismatch in the nominal valve flow parameter ($\Phi_n$), the INDI controller's performance remained largely unaffected, while the CdP controller's performance deteriorated significantly.
- Robustness to Supply Pressure Decrease: A 20% decrease in supply pressure ($P_s$) degraded the CdP controller's performance, whereas the INDI controller's performance remained intact.
- Independence from Certain Parameters: INDI is inherently insensitive to parameters like leakage flow ($L_{im}$), which do not appear in its control law.
Limitations and Future Work
- Sensitivity to State Measurement Quality: The paper acknowledges that INDI's performance is sensitive to the quality of state measurements. While filtering is proposed, further investigation into optimal filtering strategies could be beneficial.
- Fast Actuator Dynamics Assumption: The INDI approach relies on the assumption that actuator dynamics are fast compared to system dynamics for the Taylor series approximation to hold.
- Future Work: The paper suggests that INDI is promising for other applications involving hydraulic driven robots and could lead to better performance in those areas. Further exploration of INDI's application in real-world systems beyond simulations is implied.
Summary updated