Technical Summary: The Maximum Initial Mass Problem in Low-Thrust Trajectory Design

Problem Definition

The paper addresses a gap in low-thrust optimal control theory: the lack of explicit formalization and analytical treatment of the Maximum Initial Mass (MIM) problem. While the field is dominated by minimum-propellant (maximum-final-mass) and minimum-time transfer formulations, the MIM problem asks a distinct question: for a fixed transfer geometry, fixed final state, and fixed transfer time, what is the largest initial mass from which a low-thrust spacecraft can successfully accomplish the transfer?

Conceptually, this inverts the standard "given initial mass, minimize time/propellant" paradigm to "given time of flight, maximize initial mass." Although this perspective has been utilized informally in competitions (e.g., GTOCs) and for asteroid mining feasibility studies, it has not been rigorously established as a standalone optimal control formulation with its own necessary conditions and structural properties.

Methodology

The authors employ Pontryagin's Minimum Principle (PMP) to derive the first-order necessary conditions for the MIM problem and compare them structurally with the classical minimum-time problem.

1. Formulation:

   • Dynamics: The system is modeled using standard low-thrust dynamics where the state vector $\mathbf{x}$ (e.g., modified equinoctial elements) and mass $m$ evolve under a throttle $u \in [0,1]$ and thrust direction $\hat{\mathbf{i}}$.
   • Objective: Maximize initial mass $m(t_0)$ subject to fixed transfer time $t_f - t_0$ and boundary conditions $\mathbf{x}(t_0) = \mathbf{x}_0, \mathbf{x}(t_f) = \mathbf{x}_f$. This is converted to a minimization of $J = -m(t_0)$.
   • Hamiltonian Analysis: The authors derive the Hamiltonian and the associated switching function. A key finding is that the optimal throttle is always $u^*=1$ (full thrust) in the non-degenerate case, and the thrust direction follows a primer-vector law ($\hat{\mathbf{i}}^* \propto -\mathbf{B}(\mathbf{x})^T\boldsymbol{\lambda}$). This results in a continuous control structure, avoiding the discontinuous bang-bang profiles often seen in other formulations, which facilitates rapid convergence for both direct and indirect numerical methods.

2. Transversality and Gauge Conditions:

   • The paper critically re-examines the transversality condition for free-time problems. It argues that the common assumption that the Hamiltonian must be zero at the final time ($H(t_f) + \boldsymbol{\lambda}(t_f)^T\dot{\mathbf{x}}_f(t_f) = 0$) for minimum-time problems is a gauge choice rather than a necessary condition derived from PMP for boundary optima. The true condition is a one-sided inequality ($H(t_f) + \boldsymbol{\lambda}(t_f)^T\dot{\mathbf{x}}_f(t_f) \le 0$).
   • This insight allows for the normalization of the costate vector $\lambda_m(t_0) = 1$ without violating necessary conditions.

3. Extremal Correspondence:

   • Using the derived conditions, the authors prove Theorem 1: There is a one-to-one correspondence between extremals of the minimum-time problem (with fixed initial mass) and the maximum-initial-mass problem (with fixed transfer time). Specifically, a trajectory that is time-optimal for a given mass is also mass-optimal for the resulting flight time, and vice versa.

4. Homotopy Application:

   • The authors apply the MIM formulation to the multiple-revolution minimum-time problem. They construct a homotopy path by varying the number of revolutions ($N_{rev}$) while tracking the maximum initial mass surface $M(t_f, N_{rev})$. This approach is contrasted with traditional thrust-continuation homotopies, which often suffer from ill-conditioning in multi-revolution scenarios.

Key Results

• Numerical Properties: The MIM problem yields a well-posed boundary value problem with seven equations and seven unknowns (for fixed endpoints), avoiding the scaling indeterminacy issues common in minimum-time indirect shooting methods. The optimal control is strictly full-thrust, ensuring smooth numerical convergence.
• Theoretical Clarification: The paper clarifies that the terminal Hamiltonian equality condition is not a necessary condition for time-optimality but a normalization choice. This resolves a "common misconception" in the literature and validates the use of alternative gauge conditions to improve numerical stability.
• Solution Landscape Visualization: In a benchmark transfer from Geostationary Transfer Orbit (GTO) to Geostationary Orbit (GEO), the MIM homotopy successfully mapped the solution landscape. The method revealed a smooth, differentiable surface of maximum initial mass.
  • It identified multiple extremal trajectories, including local minima and maxima of the final time along iso-mass curves.
  • It successfully located the global time-optimal solution (1500 kg initial mass) and demonstrated that some previously reported solutions (e.g., solution E in the benchmark) correspond to local maxima of flight time, not minima.
• Homotopy Efficiency: The continuation path in the MIM framework (varying $N_{rev}$ to track iso-mass curves) was observed to be smooth and well-conditioned, offering a more robust alternative to thrust-continuation methods for navigating complex multi-revolution landscapes.

Significance and Claims

The paper argues that the Maximum Initial Mass problem deserves to be elevated to a "first-class" optimal control formulation in low-thrust trajectory design. Its significance lies in three main areas:

1. Complementary Design Primitive: While structurally related to the minimum-time problem, the MIM formulation offers a distinct perspective that fixes time and frees mass. This provides a complementary tool for mission design, particularly for preliminary analysis where large sets of transfers must be evaluated.
2. Analytical Rigor: By formally deriving the necessary conditions and correcting the understanding of transversality conditions in time-optimal problems, the paper provides a more rigorous theoretical foundation for low-thrust optimization.
3. Algorithmic Utility: The MIM formulation enables the construction of smooth homotopy paths that are less prone to the ill-conditioning found in traditional thrust-continuation methods. This allows for the systematic exploration of complex solution spaces (such as multi-revolution transfers) and the generation of large databases of optimal trajectories for surrogate modeling and reachability analysis.

The authors conclude that adopting the MIM formulation as a standard design primitive can improve the efficiency and robustness of low-thrust mission design, particularly in data-driven approaches and preliminary mission analysis.