Autonomous AI-Assisted Spacecraft Maneuver Planning and Control

Currently developing a modular framework that selects, sequences, and evaluates classical GNC and astrodynamics tools to generate feasible spacecraft maneuver plans under defined objectives and constraints.

The architecture is established, and development of the initial tool library is underway.

Autonomous Planning AI GNC Astrodynamics Space Teams Pro

Mission objectives and constraints are converted into an executable sequence of classical GNC tools.

Research problem

Reducing the manual burden of maneuver planning

Spacecraft missions can require engineers to design and validate multiple maneuver phases while accounting for fuel, timing, safety, and operational constraints. This process is often manual, iterative, and difficult to scale as mission complexity increases.

This research investigates whether portions of that process can be automated without replacing the classical physics-based methods used to generate and control spacecraft trajectories.

System architecture

Maneuver Planning Architecture

The framework connects four components through standardized interfaces, allowing both engineers and the autonomous planner to use the same library of GNC and astrodynamics tools.

Mission Definition

Defines the initial state, target state, mission objective, success criteria, and operational constraints.

GNC Tool Library

Provides modular targeting, transfer, guidance, evaluation, and constraint-checking tools through standardized interfaces.

Autonomous Planner

Selects candidate tools, generates and evaluates possible maneuvers, rejects infeasible options, and assembles the final maneuver sequence.

Simulation / Evaluation

Executes and visualizes the generated sequence in Space Teams Pro, then evaluates feasibility, constraint satisfaction, and mission performance.

Technical approach

AI-Assisted Planning Built on Classical GNC

Classical GNC Foundation

Classical GNC and astrodynamics methods are mature, interpretable, and effective within their intended applications. These tools generate and evaluate the candidate spacecraft maneuvers used by the planner. Depending on the selected scenario, the library may include Lambert targeting, Clohessy-Wiltshire targeting, transfer methods, and proximity-guidance tools.

Transparent Planning Baseline

The initial planner will use rule-based logic and heuristic scoring to compare candidate maneuvers based on feasibility, constraint satisfaction, fuel usage, and time of flight. Its explicit decision logic will provide an interpretable baseline against which the AI-assisted planner can be evaluated.

AI-Assisted Planner

An AI-assisted planner will use the same tool library, mission objectives, and constraints to select and rank candidate tools and maneuvers. Its performance will be compared with the rule-based baseline to determine whether AI improves planning efficiency, solution quality, or adaptability. The AI changes how maneuvers are selected and sequenced; the classical GNC tools still generate and evaluate the maneuvers.

Development status

Current work and planned development

Established

Architecture

The mission-definition, tool-library, planning, execution, and evaluation layers have been outlined.

In development

Initial Tool Library

The proposal is being technically refined while the first modular tool in the GNC library is developed.

Planned

Planner and STP Integration

Next steps include developing the rule-based planner, integrating the framework with Space Teams Pro, implementing the AI-assisted planner, and evaluation.

What I am building

A Modular Framework for Autonomous Maneuver Planning

This thesis will produce a modular framework that defines spacecraft maneuver problems, provides standardized access to classical GNC tools, and uses autonomous planners to assemble feasible maneuver sequences.

The rule-based and AI-assisted planners will be evaluated in Space Teams Pro using feasibility, constraint satisfaction, Δv and fuel usage, time of flight, and planning time. This comparison will determine whether AI improves the tool-selection and sequencing process while retaining the classical GNC foundation.

Return to all projects