Skye Mceowen,Daniel J. Calderone,Aman Tiwary,Jason S. K. Zhou,Taewan Kim,Purnanand Elango,Behcet Acikmese
Abstract:This paper presents auto-tuned primal-dual successive convexification (Auto-SCvx), an algorithm designed to reliably achieve dynamically-feasible trajectory solutions for constrained hypersonic reentry optimal control problems across a large mission parameter space. In Auto-SCvx, we solve a sequence of convex subproblems until convergence to a solution of the original nonconvex problem. This method iteratively optimizes dual variables in closed-form in order to update the penalty hyperparameters used in the primal variable updates. A benefit of this method is that it is auto-tuning, and requires no hand-tuning by the user with respect to the constraint penalty weights. Several example hypersonic reentry problems are posed and solved using this method, and comparative studies are conducted against current methods. In these numerical studies, our algorithm demonstrates equal and often improved performance while not requiring hand-tuning of penalty hyperparameters.
What problem does this paper attempt to address?
The problem that this paper attempts to solve is how to reliably generate dynamically feasible trajectories that satisfy various constraints in the trajectory optimization of hypersonic re - entry vehicles. Specifically, the authors propose a new algorithm named Auto - Tuning Primal - Dual Successive Convexification (Auto - SCvx) to solve the hypersonic re - entry optimal control problem with a wide task parameter space.
### Main Problems and Challenges
1. **Non - convexity**: The hypersonic re - entry trajectory optimization problem is inherently non - convex, and it is difficult for traditional methods to guarantee the global optimal solution.
2. **Complex Constraints**: It is necessary to consider a variety of complex constraints, such as heat flux rate, dynamic pressure, normal load, etc., to ensure that the vehicle does not exceed the safe range during re - entry.
3. **Real - time Performance**: In order to meet the needs of online adjustment, the algorithm needs to have real - time performance and be able to generate reliable trajectories in a short time.
4. **Manual Parameter Tuning**: The traditional SCP (Sequential Convex Programming) method usually requires manual adjustment of penalty weights, which is not only time - consuming but also error - prone.
### Solutions
To solve the above problems, the paper proposes the Auto - SCvx algorithm, whose main features include:
- **Auto - Tuning**: Automatically update the penalty hyper - parameters by optimizing the dual variables without the need for user - manual adjustment.
- **Efficient Solution**: Solve a series of convex sub - problems iteratively and finally converge to an approximate optimal solution of the original non - convex problem.
- **High Precision**: Combine the inverse - free exact discretization technique to ensure that the dynamic equations are accurately satisfied throughout the time range.
- **Flexibility**: Allow the use of sparse time grids and can handle various different control parameterizations and path constraints.
### Mathematical Models and Formulas
1. **Dimensionless State and Control Variables**:
\[
\text{Time}\,t=\frac{\tau}{\sqrt{\frac{R_{\oplus}}{g_{\oplus}}}}, \quad \text{Distance}\,r = \frac{r}{R_{\oplus}}, \quad \text{Speed}\,v=\frac{v}{\sqrt{g_{\oplus}R_{\oplus}}}, \quad \text{Acceleration}\,a=\frac{a}{g_{\oplus}}
\]
where \(g_{\oplus}=9.81\,\text{m/s}^2\) is the gravitational acceleration on the earth's surface, and \(R_{\oplus}=6378\times 10^{3}\,\text{m}\) is the radius of the earth.
2. **Dynamic Equations**:
\[
\dot{\mathbf{x}}=f(\mathbf{x},\mathbf{u})
\]
where \(\mathbf{x}\) is the state vector and \(\mathbf{x}\) is the control input vector.
3. **Lift and Drag Accelerations**:
\[
C_{L}=\frac{L}{\frac{1}{2}\rho v^{2}S_{ref}}, \quad C_{D}=\frac{D}{\frac{1}{2}\rho v^{2}S_{ref}}
\]
where \(L\) and \(D\) are the lift and drag respectively, \(\rho\) is the atmospheric density, and \(S_{ref}\) is the reference area.
4. **Path Constraints**:
\[
c(\mathbf{x},\mathbf{u})\leq0
\]
Ensure that the heat flux rate, dynamic pressure, normal load, etc. do not exceed the allowable range.
### Summary
By introducing the Auto - SCvx algorithm, this paper solves the non - convexity and complex constraint problems in hypersonic re - entry trajectory optimization, and at the same time realizes automatic tuning and efficient real - time performance. This method has shown superior performance in multiple numerical experiments and can reliably generate dynamically feasible trajectories that satisfy various constraints.