Astronautics — A Journal of Space Engineering

Permanent lunar habitation will require robotic systems that can maintain infrastructure, recover from local failures, and acquire new operational capabilities under limited Earth supervision. Existing planetary robots are largely fixed-function specialists, while end-to-end foundation-model policies remain difficult to validate and extend for safety-critical surface operations. We present SELENE (Self-Evolving Lunar Embodied ageNt Ecosystem), an architectural proposal for contract-governed lunar robotic autonomy centered on a shared Atomic Action Library $\mathbb{A}$. The key abstraction is the Atomic Action Contract: a typed skill interface that specifies parameters, preconditions, goal predicates, execution bindings, safety envelopes, runtime reports, and validation metadata. Through this contract, a VLM-driven Cognitive Agent plans over executable skills, a multi-modal Execution Agent realizes them through optimization-based controllers, Vision–Language–Action (VLA) policies, Vision–Language–Navigation (VLN) policies, or reinforcement-learned policies, and an offline Evolutionary Agentic Framework synthesizes and registers new candidate contracts without modifying the planner or the execution interface. This paper presents an architecture-level validation of that contract mechanism. We instantiate SELENE across two heterogeneous pathways on LunarBot and its simulation counterpart, with optimization-based control supported as a third execution modality. A pre-trained VLA policy adapted from 100 teleoperated demonstrations achieves 29/30 task success (96.7 percent) in in-domain trials on the physical LunarBot. A curriculum–RL policy instantiates the traversal pathway in simulated lunar-gravity terrain. Together, these results show that the Atomic Action Contract can serve as a common registration and dispatch interface across heterogeneous control modalities. The same contract layer also defines the path toward runtime gap-triggered self-evolution, mission-grade admission, and lunar-environment validation in subsequent system-level studies.

The thermal control subsystem is one of the critical subsystems of a satellite. Optimization of the design of this subsystem could benefit a satellite’s power and mass budgets. In previous optimization studies regarding the thermal management of low Earth orbit satellites, solstices are typically assumed to be the thermally extreme cases. However, this could not always be accurate. In addition, previous studies often evaluated specific parts of the thermal system, while overlooking the dynamics of energy conversion within the satellite. In this study, thermally extreme cases are identified using a lumped analysis approach. Moreover, heat dissipation of certain components is modeled based on the relationship between the generation and the consumption of power. The process begins with an orbital analysis, which provides inputs for the lumped analysis. According to the results of the lumped analysis, finite element analyses are performed for the worst hot and the worst cold cases. The aim is to minimize the need for heater energy consumption by adjusting the coordinates of the corners of the radiators, where constraints are set by the safe temperature intervals of specific components. This computationally demanding optimization problem is solved by Bayesian optimization, also a first for a satellite design study. The optimized system requires a reduced amount of heater energy while satisfying the temperature constraints.

It is challenging to detect continuous low-thrust maneuver parameters of non-cooperative satellites because the signals are weak over limited observation arcs and are readily masked by measurement noise and orbit-determination errors. This paper proposes a conditional diffusion model for detecting and estimating continuous low-thrust maneuver parameters from relative-orbit observations. The method uses relative-orbit observations of the non-cooperative target to construct conditional inputs that incorporate orbital dynamical priors. Single-step differencing and dimensionless processing are then used to strengthen weak maneuver signatures. The conditional diffusion model learns the evolution of maneuver parameters under noisy conditions and estimates three-axis continuous low-thrust acceleration sequences. Based on simulations considering the Gaussian noise of relative positions and velocities, the proposed method achieved 85.2% maneuver detection accuracy, while that of the batch least-squares benchmark method was 67.8%. The proposed method is simulated and verified based on Sentinel-6A. Results show that the continuous low-thrust maneuver can be robustly identified under low signal-to-noise ratios and the temporal parameter evolution can be recovered. The method provides a practical route for analyzing non-cooperative satellite maneuver and supporting on-orbit space situational awareness.