Safe Mission-Level Path Planning for Exploration of Lunar Shadowed Regions by a Solar-Powered Rover

TL;DR

The paper tackles safe exploration of permanently shadowed lunar regions using a solar-powered rover by formulating a joint chance-constrained mission-level online planning problem. It combines offline mission-level path planning with online stochastic reachability to produce risk-bounded partial policy trees that ensure the rover visits as many science-worthy waypoints as possible without exceeding a specified failure probability. Key contributions include the risk-bounded AO-style optimization with artificial termination and backward search, the adaptation of TEMPEST to generate risk-bounded partial policy trees, and empirical validation on Cabeus Crater and LCROSS-adjacent regions demonstrating competitive rewards under tight risk bounds. This approach enables robust, energy-aware, long-range lunar exploration under stochastic faults, with practical implications for autonomous planning and human-operator-assisted missions.

Abstract

Exploration of the lunar south pole with a solar-powered rover is challenging due to the highly dynamic solar illumination conditions and the presence of permanently shadowed regions (PSRs). In turn, careful planning in space and time is essential. Mission-level path planning is a global, spatiotemporal paradigm that addresses this challenge, taking into account rover resources and mission requirements. However, existing approaches do not proactively account for random disturbances, such as recurring faults, that may temporarily delay rover traverse progress. In this paper, we formulate a chance-constrained mission-level planning problem for the exploration of PSRs by a solar-powered rover affected by random faults. The objective is to find a policy that visits as many waypoints of scientific interest as possible while respecting an upper bound on the probability of mission failure. Our approach assumes that faults occur randomly, but at a known, constant average rate. Each fault is resolved within a fixed time, simulating the recovery period of an autonomous system or the time required for a team of human operators to intervene. Unlike solutions based upon dynamic programming alone, our method breaks the chance-constrained optimization problem into smaller offline and online subtasks to make the problem computationally tractable. Specifically, our solution combines existing mission-level path planning techniques with a stochastic reachability analysis component. We find mission plans that remain within reach of safety throughout large state spaces. To empirically validate our algorithm, we simulate mission scenarios using orbital terrain and illumination maps of Cabeus Crater. Results from simulations of multi-day, long-range drives in the LCROSS impact region are also presented.

Figures (14)

• Figure 1: The importance of risk-aware planning when exploring PSRs with a solar-powered rover affected by recurring, random faults. Subfigure A shows a nominal mission plan visiting waypoints of interest in a given order. After a fault/delay occurs during the traverse (Subfigure B), risk-aware planning (blue dashed line) suggests an early PSR exit while risk-agnostic planning (white dashed line) finds an updated plan visiting the fourth waypoint, unaware of the danger of this traverse schedule. If a second fault occurs inside of the PSR (Subfigure C), a rover following the risk-agnostic plan misses a crucial solar charging period and battery energy depletion becomes unavoidable (Subfigure D). A rover following the risk-aware plan, on the other hand, is still capable of reaching the designated target region safely. Background image courtesy of NASA and Arizona State University. VIPER render courtesy of NASA.
• Figure 2: Conceptual view of the three possible state transitions and the associated probabilities for mobility action $a_i$ from originating state $\boldsymbol{x}_i$. For this state-action combination, the function $\boldsymbol{\Delta}_{nom}(\boldsymbol{x}_i, a_i)$ defines the change in state for a nominal transition. Only the spatial and temporal dimensions are visualized. The nominal transition is shown in blue, while the fault-induced transitions are shown in red.
• Figure 3: A partial policy tree (bold), grown from a given start state $\boldsymbol{x}_0$. OR nodes resulting from fault-related state transitions are artificially terminated (here, coloured in green). Without this mechanism, standard AO tree planners would return a complete policy, which would require many more expansion cycles (shown semi-transparent).
• Figure 4: Backward search to find a risk-bounded partial policy tree. In forward-time, the tree guides the rover from a start state towards region $w$, where a waypoint action can be taken. Stage A consists of discretizing the waypoint region into terminal nodes and initializing the search. Stage B conducts the search according to a user-provided cost function, back-propagating nominal actions and calculating the risk at newly-created nodes using their immediate forward-time successors. Once the search reaches the start state (Stage C), the best partial policy tree is the forward-time branch leading to the waypoint region. Node and state transition colors follow the same convention as in \ref{['fig:partial-tree']}.
• Figure 5: The iterative mechanism of risk-bounded TEMPEST. If no solution (feasible partial policy tree) is found for a given list of waypoints, the last waypoint is dropped and a new search is started with the updated waypoint list. Here, instances initialized in waypoint regions $w_3$ and $w_2$ fail to find feasible solutions in the search segment linking the start state to the first waypoint.