Abstract:Multi-agent pathfinding (MAPF) is the problem of finding a set of conflict-free paths for a set of agents. Typically, the agents' moves are limited to a pre-defined graph of possible locations and allowed transitions between them, e.g. a 4-neighborhood grid. We explore how to solve MAPF problems when each agent can move between any pair of possible locations as long as traversing the line segment connecting them does not lead to a collision with the obstacles. This is known as any-angle pathfinding. We present the first optimal any-angle multi-agent pathfinding algorithm. Our planner is based on the Continuous Conflict-based Search (CCBS) algorithm and an optimal any-angle variant of the Safe Interval Path Planning (TO-AA-SIPP). The straightforward combination of those, however, scales poorly since any-angle path finding induces search trees with a very large branching factor. To mitigate this, we adapt two techniques from classical MAPF to the any-angle setting, namely Disjoint Splitting and Multi-Constraints. Experimental results on different combinations of these techniques show they enable solving over 30% more problems than the vanilla combination of CCBS and TO-AA-SIPP. In addition, we present a bounded-suboptimal variant of our algorithm, that enables trading runtime for solution cost in a controlled manner.
What problem does this paper attempt to address?
This paper attempts to solve a specific challenge in the multi - agent pathfinding (MAPF) problem, that is, to find a set of optimal conflict - free paths in any - angle environments. Traditional MAPF research usually assumes that agents can only move along predefined graph structures (such as 4 - neighborhood grids), and each movement only takes one time unit. However, these assumptions are not applicable to many practical robot application scenarios.
### Specific Problem Description
1. **Any - Angle Pathfinding**:
- Traditional MAPF algorithms limit agents to move only along predefined discrete points and edges, while this paper explores that agents can move directly between any two positions as long as the line segment between them does not intersect with obstacles.
2. **Optimization of Multi - agent Pathfinding**:
- Researchers proposed a new algorithm AA - CCBS (Any - Angle Continuous Conflict - based Search), which combines continuous conflict - based search (CCBS) and the optimal any - angle single - agent pathfinding algorithm TO - AA - SIPP. This new algorithm can handle the any - angle multi - agent pathfinding problem while ensuring optimality.
3. **Scalability and Efficiency Improvement**:
- Directly combining CCBS and TO - AA - SIPP will result in a very large branching factor of the search tree, thus affecting the scalability of the algorithm. To solve this problem, researchers introduced two techniques: disjoint splitting and multi - constraints. These techniques significantly improve the algorithm's ability to solve problems in complex environments.
4. **Generation of Sub - optimal Solutions**:
- In addition, researchers also proposed a bounded - suboptimal version of the AA - CCBS algorithm, allowing a trade - off between running time and solution quality. In this way, near - optimal solutions can be found in a shorter time.
### Summary
The main contribution of this paper is to propose the first optimal algorithm AA - CCBS that can handle any - angle multi - agent pathfinding, and significantly improve the algorithm's scalability and solution efficiency by introducing techniques such as disjoint splitting and multi - constraints. In addition, a bounded - suboptimal version of the algorithm is also provided to meet the requirements of different application scenarios.
### Key Formulas
- **Cost Function**:
\[
SOC=\sum_{i = 1}^{n}c(\pi_i)
\]
where \(c(\pi_i)\) represents the path cost of the \(i\)-th agent, and \(SOC\) represents the sum of the path costs of all agents.
- **Multi - Constraints (MC)**:
\[
MC =\{(a_i,[t,t'])\}
\]
where \(a_i\) is an action, and \([t,t']\) is the time interval during which the action cannot be executed safely.
- **Disjoint Splitting (DS)**:
\[
\text{DS}((a_i,t_i),(a_j,t_j))=\left\{\begin{array}{ll}(a_i,[t,t']),&\text{Apply negative constraint to the first agent}\\(a_i,[t,t']),(a_j,[t,t']),&\text{Apply positive constraint to the first agent and negative constraint to the second agent}\end{array}\right.
\]
Through these improvements, AA - CCBS can solve complex any - angle multi - agent pathfinding problems more efficiently.