Abstract:In this paper, a novel, dual-mode model predictive control framework is introduced that combines the dynamic window approach to navigation with reference tracking controllers. This adds a deliberative component to the obstacle avoidance guarantees present in the dynamic window approach as well as allow for the inclusion of complex robot models. The proposed algorithm allows for guaranteed convergence to a goal location while navigating through an unknown environment at relatively high speeds. The framework is applied in both simulation and hardware implementation to demonstrate the computational feasibility and the ability to cope with dynamic constraints and stability concerns.
What problem does this paper attempt to address?
The problem that this paper attempts to solve is how to ensure stability and optimality when a robot navigates at a relatively high speed in an unknown environment. Specifically, the paper aims to propose a dual - mode Model Predictive Control (MPC) framework by combining the Dynamic Window Approach (DWA) and a reference tracking controller, so as to ensure that the robot can converge to the target position and avoid obstacles when navigating rapidly in an unknown environment.
### Main problems:
1. **Stability in a highly dynamic environment**: When the robot operates at a high speed, dynamic constraints (such as acceleration and motion limitations) make it difficult to ensure stability and optimality.
2. **Real - time planning in an unknown environment**: In an unknown environment, the robot needs to constantly recalculate the path to deal with new environmental data, which increases the computational complexity.
3. **Local minimum problem**: The traditional DWA method is prone to getting stuck in local minima, resulting in ineffective convergence to the target position.
### Solutions:
The paper proposes a dual - mode MPC framework, which combines DWA and a reference tracking controller to solve the above problems. Specifically:
- **Dual - mode MPC framework**: By introducing a global planner to generate paths and using the MPC framework to consider the complete dynamic model of the system, it provides a guarantee of convergence to the target position.
- **Arc - based controller**: By using arc - based motion primitives (i.e., constant linear velocity \(v\) and angular velocity \(\omega\)), obstacle avoidance can be ensured without adding too much computational burden.
- **Reference tracking controller**: Through the reference tracking controller, it is ensured that the robot can move along the planned path, thus avoiding the local minimum problem.
### Mathematical expressions:
- Robot motion model (unicycle model):
\[
\begin{bmatrix}
\dot{x}_1 \\
\dot{x}_2 \\
\dot{\psi}_3
\end{bmatrix}
=
\begin{bmatrix}
v\cos(\psi) \\
v\sin(\psi) \\
\omega
\end{bmatrix}
\]
where \((x_1, x_2)\) is the two - dimensional position of the robot, \(\psi\) is the heading angle, and \(v\) and \(\omega\) are the linear velocity and angular velocity respectively.
- Optimization problem form:
\[
\min_{u(\cdot; t_0)} \int_{t_0}^{t_0+\Delta} L(x(t; t_0), u(t; t_0)) \, dt+\Psi(x(t_0+\Delta; t_0))
\]
Subject to the constraints:
\[
\dot{x}(t; t_0) = f(x(t; t_0), u(t; t_0)), \quad x(t_0+\Delta; t_0)\in X_f
\]
Through these methods, the paper successfully demonstrates the effectiveness of the algorithm in simulation and actual hardware implementation, proving its advantages in handling complex dynamics and maintaining convergence.