Multi-Robot Coordination for Planning under Context Uncertainty

Abstract

Real-world robots often operate in settings where objective priorities depend on the underlying context of operation. When the underlying context is unknown apriori, multiple robots may have to coordinate to gather informative observations to infer the context, since acting based on an incorrect context can lead to misaligned and unsafe behavior. Once the underlying true context is inferred, the robots optimize their task-specific objectives in the preference order induced by the context. We formalize this problem as a Multi-Robot Context-Uncertain Stochastic Shortest Path (MR-CUSSP), which captures context-relevant information at landmark states through joint observations. Our two-stage solution approach is composed of: (1) CIMOP (Coordinated Inference for Multi-Objective Planning) to compute plans that guide robots toward informative landmarks to efficiently infer the true context, and (2) LCBS (Lexicographic Conflict-Based Search) for collision-free multi-robot path planning with lexicographic objective preferences, induced by the context. We evaluate the algorithms using three simulated domains and demonstrate its practical applicability using five mobile robots in the salp domain setup.

Figures (6)

• Figure 1: Illustration of multiple GTernal robots wilson2020robotarium with a shared belief over true context. Accurate context-relevant observations are available only when robots are in a required configuration at a landmark (e.g., chain at $\ell_1$ and ring at $\ell_2$). After context inference, robots compute plans aligned with the context-induced objective preferences to reach their goal.
• Figure 2: Solution approach overview. A most-likely-outcome determinization is first applied to obtain a discrete graph representation of the stochastic domain, enabling the use of graph-based planning methods. CIMOP prioritizes visiting landmark states that minimize the belief entropy and assigns robots accordingly, based on the current shared belief which is updated based on joint observations at landmark states. Once the context is inferred ($c_g$), the induced lexicographic ordering $\Theta(c_g)$ and a discrete graph representation of the environment, along with a heuristic, are used for task planning. LCBS uses lexicographic $A^*$ to compute preference-aligned paths, detects conflicts in the joint plan, and iteratively adds constraints using binary branching ren2023binary until a conflict-free solution is obtained.
• Figure 3: Summary of experimental results across three domains. The first column shows the domain maps. The second and third columns show entropy-of-belief trends for $5$ and $35$ robots, respectively. The fourth column shows path cost with increasing percentage of redundant landmarks for analysis in cluttered environments. The fifth column shows plan computation time with increasing numbers of robots. All results are averaged over five instances with randomized robot starting locations.
• Figure 4: Success rate on 15 instances (five from each of the three domains) under constrained planning time limit with five robots and three objectives.
• Figure 5: Execution sequence with five GTernal robots wilson2020robotarium: (1) start with uniform belief over three contexts, (2) form a chain to observe the cave landmark and update belief, (3) four robots form a ring to observe the crevice landmark and collapse belief, after which (4) the robots independently plan prioritizing minimizing coral damage over energy consumption and time to goal as imposed by the inferred context $c_2$.

Theorems & Definitions (1)

• Definition 1