Traversing Mars: Cooperative Informative Path Planning to Efficiently Navigate Unknown Scenes

TL;DR

The paper tackles efficient navigation in unknown environments by a two-robot system where a scouting aerial agent reveals a minimum-cost follower path for a ground rover. It introduces a follower-path-aware cooperative IPP with a two-stage scouting strategy, combining feasible-path feasibility checks and optimistic lower-bounds to drive exploration toward the optimal follower path, while providing termination guarantees. Theoretical results prove completeness and optimality conditions, and empirical experiments in Mars-like terrains show significant improvements in time to find the optimal path and to terminate, across varying cost gradients and obstacle densities. This work enables safer, faster, and wear-minimizing multi-robot exploration suitable for planetary missions and similar search-and-rescue or inspection tasks.

Abstract

The ability to traverse an unknown environment is crucial for autonomous robot operations. However, due to the limited sensing capabilities and system constraints, approaching this problem with a single robot agent can be slow, costly, and unsafe. For example, in planetary exploration missions, the wear on the wheels of a rover from abrasive terrain should be minimized at all costs as reparations are infeasible. On the other hand, utilizing a scouting robot such as a micro aerial vehicle (MAV) has the potential to reduce wear and time costs and increasing safety of a follower robot. This work proposes a novel cooperative IPP framework that allows a scout (e.g., an MAV) to efficiently explore the minimum-cost-path for a follower (e.g., a rover) to reach the goal. We derive theoretic guarantees for our algorithm, and prove that the algorithm always terminates, always finds the optimal path if it exists, and terminates early when the found path is shown to be optimal or infeasible. We show in thorough experimental evaluation that the guarantees hold in practice, and that our algorithm is 22.5% quicker to find the optimal path and 15% quicker to terminate compared to existing methods.

Figures (9)

• Figure 1: Illustration of our problem and approach. A follower robot (e.g. a rover, bottom) has to traverse through unknown space to reach a goal (top) while minimizing traversal cost (soil color, from red=high to green=low). The goal is for a scouting robot (e.g. an MAV) to explore the scene such that the follower path is optimal. Our approach generates an optimistic follower path (blue) guiding the scouting-IPP. For illustration, sparsified IPP-samples are shown colored by information gain (colored from red=low to green=high), with the past (teal) and currently planned path (green) of the scout.
• Figure 2: Overview of the proposed approach. The scout adds observations to the map. The follower planner computes feasible and optimistic paths on the given map. These paths are used to guide future exploration by the scout. Once the optimal follower path has been established, the scout terminates and the path can be executed. (Images: NASA/JPL)
• Figure 3: Different environments used for evaluation. Variable follower cost fields are shown in color from red (high) to green (low). The optimal collision-free path is shown in blue.
• Figure 4: Normalized follower path cost $Q_F(t)/Q_F^\star$ and scene coverage after $Q_S(t) = t$ seconds of scouting for different methods.
• Figure 5: Scouting performance for varying cost gradients in a scene.

Theorems & Definitions (13)

• Definition 4.1: Feasible Path
• Definition 4.2: Optimistic Map and Optimistic Path
• Proposition 4.1: Feasible Problem
• proof
• Proposition 4.2: Infeasible Problem
• proof
• Proposition 4.3: Cost Upper Bound
• proof
• Proposition 4.4: Cost Lower Bound
• proof