A Time-dependent Risk-aware distributed Multi-Agent Path Finder based on A*

TL;DR

The paper tackles dynamic multi-agent path finding by introducing a distributed, time-dependent risk-aware planner, $A^*_{+}T$, that anticipates other agents and moving obstacles using velocity-based forecasts. It integrates a comprehensive risk framework with components for occupancy, proximity, dynamic-object, time, and distance, and augments A* with a time dimension and shared trajectories, facilitated by the DTAA for obstacle prediction. The authors validate the approach in Gazebo against established MAPF methods ($CBS$, $ECBS$, $SIPP$) and demonstrate real-world viability through single-robot experiments with dynamic obstacles, showing safer and more navigable paths at the cost of sometimes longer routes. These results highlight the practicality of distributed, risk-aware planning for real-world robotics, where dynamic agents and moving objects are common and coordination is essential.

Abstract

Multi-Agent Path-Finding (MAPF) focuses on the collaborative planning of paths for multiple agents within shared spaces, aiming for collision-free navigation. Conventional planning methods often overlook the presence of other agents, which can result in conflicts. In response, this article introduces the A$^*_+$T algorithm, a distributed approach that improves coordination among agents by anticipating their positions based on their movement speeds. The algorithm also considers dynamic obstacles, assessing potential collisions with respect to observed speeds and trajectories, thereby facilitating collision-free path planning in environments populated by other agents and moving objects. It incorporates a risk layer surrounding both dynamic and static entities, enhancing its utility in real-world applications. Each agent functions autonomously while being mindful of the paths chosen by others, effectively addressing the complexities inherent in multi-agent situations. The performance of A$^*_+$T has been rigorously tested in the Gazebo simulation environment and benchmarked against established approaches such as CBS, ECBS, and SIPP. Furthermore, the algorithm has shown competence in single-agent experiments, with results demonstrating its effectiveness in managing dynamic obstacles and affirming its practical relevance across various scenarios.

Figures (10)

• Figure 1: A snapshot from lab experiments where a Spot robot plans a path through a gap in a wall. This is the same experiment as in Figure \ref{['fig:risk:layer']}
• Figure 2: A block diagram of the system, It shows how A$^*_+$T shares its path with the other agents, as well how it interact and depend on other modules. The block map and Odometry can be changed and different options was utilized in this article.
• Figure 3: An instance where the A$^*_+$T algorithm determines a longer but safer path (green line), opting for a larger opening over a smaller, more hazardous route.
• Figure 4: When an agent generates a path $\Psi$, it transmits its $\Psi$ to all other agents for consideration during their path planning. Conversely, when any other agent computes its own $\Psi_i$, it also shares $\Psi_i$ with all other agents, including the initial agent that transmitted a $\Psi$.
• Figure 5: Three agents plan their trajectory in a simulated house using A$^*_+$T. The map visulized is with the risk layer $R_p$.