Data-Driven Multi-step Nonlinear Model Predictive Control for Industrial Heavy Load Hydraulic Robot

Dexian Ma,Bo Zhou

2024-11-21

Abstract:Automating complex industrial robots requires precise nonlinear control and efficient energy management. This paper introduces a data-driven nonlinear model predictive control (NMPC) framework to optimize control under multiple objectives. To enhance the prediction accuracy of the dynamic model, we design a single-shot multi-step prediction (SSMP) model based on long short-term memory (LSTM) and multilayer perceptrons (MLP), which can directly obtain the predictive horizon without iterative repetition and reduce computational pressure. Moreover, we combine offline and online models to address disturbances stemming from environmental interactions, similar to the superposition of the robot's free and forced responses. The online model learns the system's variations from the prediction mismatches of the offline model and updates its weights in real time. The proposed hybrid predictive model simplifies the relationship between inputs and outputs into matrix multiplication, which can quickly obtain the derivative. Therefore, the solution for the control signal sequence employs a gradient descent method with an adaptive learning rate, allowing the NMPC cost function to be formulated as a convex function incorporating critical states. The learning rate is dynamically adjusted based on state errors to counteract the inherent prediction inaccuracies of neural networks. The controller outputs the average value of the control signal sequence instead of the first value. Simulations and experiments on a 22-ton hydraulic excavator have validated the effectiveness of our method, showing that the proposed NMPC approach can be widely applied to industrial systems, including nonlinear control and energy management.

Robotics

What problem does this paper attempt to address?

### What problems does this paper attempt to solve?

This paper aims to solve the problems of precise nonlinear control and efficient energy management of industrial heavy - duty hydraulic robots in complex environments. Specifically, the author proposes a data - driven multi - step nonlinear model predictive control (NMPC) framework to optimize the control performance under multiple objectives. The following are the main problems that this paper attempts to solve:

1. **Precise nonlinear control**:
- Industrial robots, especially heavy - duty hydraulic robots, have complex nonlinear dynamic characteristics, and the traditional PID control method is difficult to meet the requirements of high - precision control.
- Existing model predictive control (MPC) methods rely on accurate dynamic models, but there are challenges in dealing with complex nonlinear systems, such as factors that are difficult to measure directly, such as combustion, friction, and fluid mechanics.

2. **Efficient energy management**:
- In the context of automation and unmanned operation, industrial systems need to improve productivity and achieve significant economic benefits. Therefore, in addition to precise control, effective use of energy also needs to be considered.

3. **Cumulative error and computational burden**:
- Traditional prediction methods usually construct the prediction time range step by step through iteration, which will lead to error accumulation and increase the computational burden. To solve this problem, the paper proposes a single - shot multi - step prediction (SSMP) model, which can directly obtain the prediction time range without iterative repetition, thereby reducing the computational pressure and error accumulation.

4. **Impact of environmental interaction**:
- The interaction between the robot and the external environment (such as different load conditions or temperature changes) will affect its dynamic characteristics, resulting in a decrease in the prediction accuracy of the offline - trained model. For this reason, the paper combines offline and online models to adapt to environmental changes and update the model weights in real time.

5. **Real - time optimization**:
- The optimization of the control sequence of nonlinear systems is a complex nonlinear programming (NLP) problem. The paper proposes an adaptive gradient - descent optimization method and designs a cost function suitable for multi - objective optimization to ensure the feasibility and efficiency in practical applications.

### Overview of solutions

To address the above problems, the author proposes the following solutions:

- **Single - shot multi - step prediction (SSMP) model**: Based on long - short - term memory (LSTM) and multi - layer perceptron (MLP), this model can directly predict the state changes of multiple future time steps, avoiding the problem of error accumulation in traditional methods.
- **Combination of offline and online models**: The offline model is used to learn the static characteristics of the system, while the online model compensates for the prediction deviation of the offline model through real - time data updates, thereby improving the adaptability to environmental changes.
- **Adaptive gradient - descent optimization method**: By introducing a variable learning rate, the influence of the model fitting error on the control performance is reduced, and the average value of the control signal sequence is used instead of the first value to improve the control accuracy.

### Experimental verification

The paper verifies the effectiveness of the proposed NMPC method on a 22 - ton hydraulic excavator through simulation and experiment, demonstrating the wide applicability and economic efficiency of this method in complex nonlinear systems.

In summary, through innovative data - driven methods and optimization strategies, this paper solves the problems of precise control and energy management of industrial heavy - duty hydraulic robots in complex environments, providing new ideas and technical support for unmanned operations.