[Paper Review] Framework for Modeling and Optimization of On-Orbit Servicing Operations under Demand Uncertainties

Tristan Sarton du Jonchay, Hao Chen+2 more|arXiv (Cornell University)|Jun 16, 2022

Advanced Manufacturing and Logistics Optimization22 references27 citations

TL;DR

This paper proposes a Mixed-Integer Linear Programming (MILP) framework integrated with Rolling Horizon (RH) decision making to optimize on-orbit servicing (OOS) operations under uncertain demand. It models multi-servicer, multi-depot architectures for geostationary satellites, enabling both short-term scheduling and long-term strategic design under dynamic market conditions, with case studies demonstrating superior performance in cost-efficient mission planning and architecture trade-space analysis.

ABSTRACT

This paper develops a framework that models and optimizes the operations of complex on-orbit servicing infrastructures involving one or more servicers and orbital depots to provide multiple types of services to a fleet of geostationary satellites. The proposed method extends the state-of-the-art space logistics technique by addressing the unique challenges in on-orbit servicing applications and integrates it with the Rolling Horizon decision-making approach. The space logistics technique enables modeling of the on-orbit servicing logistical operations as a Mixed-Integer Linear Program whose optimal solutions can efficiently be found. The Rolling Horizon approach enables the assessment of the long-term value of an on-orbit servicing infrastructure by accounting for the uncertain service needs that arise over time among the geostationary satellites. Two case studies successfully demonstrate the effectiveness of the framework for 1) short-term operational scheduling and 2) long-term strategic decision making for on-orbit servicing architectures under diverse market conditions.

Motivation & Objective

• To address the lack of systematic optimization frameworks for sustainable many-to-many on-orbit servicing (OOS) infrastructures.
• To model complex OOS operations involving multiple servicers, orbital depots, and diverse service types as a logistics network.
• To integrate uncertainty in service demand over time through the Rolling Horizon (RH) decision-making approach.
• To enable both short-term operational scheduling and long-term strategic design of OOS architectures under varying market conditions.
• To extend traditional space logistics modeling with realistic constraints such as phasing maneuvers, propellant use, and service-tool mappings.

Proposed method

• Formulates OOS operations as a Mixed-Integer Linear Program (MILP) with commodity flow, service assignment, and dispatch variables.
• Models orbital transfers using phasing maneuvers with impulsive-thrust, enabling time-expanded network representation.
• Applies the rocket equation to compute propellant mass based on required delta-v, linking propulsion to vehicle mass and payload capacity.
• Introduces binary parameters (𝛽𝑠𝜏𝑡) to link servicer dispatch and service assignment, ensuring temporal and spatial consistency.
• Uses Rolling Horizon (RH) approach to handle demand uncertainty by re-optimizing at each time step with updated forecasts.
• Incorporates cost components for launch, depot operations, servicer operations, penalties for delays, and revenues from services into a unified objective function.

Experimental results

Research questions

• RQ1 How can on-orbit servicing operations be modeled as a logistics network that accounts for multiple servicers, depots, and service types?
• RQ2 What is the impact of demand uncertainty on long-term OOS infrastructure performance, and how can it be effectively managed?
• RQ3 How does the choice between a monolithic (single versatile servicer) and distributed (multiple specialized servicers) architecture affect operational efficiency and profitability?
• RQ4 What are the key cost drivers and trade-offs in designing scalable OOS architectures under dynamic market conditions?
• RQ5 Can the proposed framework efficiently support both short-term scheduling and long-term strategic planning for OOS systems?

Key findings

• The framework successfully optimizes short-term scheduling of existing OOS infrastructure over a single planning horizon, demonstrating computational efficiency and feasibility.
• In long-term simulations, the distributed architecture with four specialized servicers outperformed the monolithic design under high-demand conditions, achieving higher profitability and better service coverage.
• The monolithic architecture was more cost-effective under low-demand scenarios due to lower operational complexity and fewer vehicles.
• Sensitivity analyses revealed that demand uncertainty significantly affects optimal infrastructure design, with rolling horizon enabling adaptive responses to changing market conditions.
• The integration of phasing maneuver modeling and propellant mass calculation via the rocket equation enabled accurate representation of mission mass and delta-v constraints in the MILP formulation.
• The framework enables quantitative evaluation of trade-offs between vehicle versatility, depot deployment, and service delay penalties, supporting data-driven decision-making.