We present a heuristic-guided multi-robot motion planning framework that solves the problem of <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">n</i> dynamical agents visiting <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">m</i> unlabeled targets in a partially known environment for planetary surface missions without solving the two-point boundary value problem (BVP). The framework design is motivated by typical planetary surface mission constraints of limited power, limited computation, and limited communication. The framework maintains a centralized, dynamically updated probabilistic roadmap (PRM) that incorporates new obstacle updates as the agents move in the environment. The dynamic roadmap captures the changing obstacle topology and provides updated cost-to-go heuristics to accelerate each agent's independent single-query motion-planning process. The agents use a feasible sampling-based motion planner without computing the BVP while leveraging the roadmap heuristics to quickly plan and visit their assigned target. The agents handle robot-robot and robot-obstacle collision avoidance in a decentralized fashion. We conduct multiple simulation experiments using robots with non-linear dynamics to show our planner performs better in overall planning time and mission time than approaches not using the roadmap heuristic. We also field our algorithm on prototype rovers and demonstrate the viability of implementing our algorithm on real-world hardware platforms.