Intermittent Strategic Cooperation-Based Path Planning studies how self-interested agents move through a graph when cooperation is possible at selected locations. Each agent has its own start and target, and each agent seeks to minimize its own path time. Cooperation can reduce execution time at specific nodes, but it may also require waiting, synchronization, or a detour from an individually shortest path.
The problem combines multi-agent path planning with strategic decision-making: agents must reason not only about graph distances, but also about whether a temporary cooperative interaction is individually worthwhile.
In classical path planning, each agent can often be planned independently or coordinated under a shared team objective. Intermittent strategic cooperation occupies a different space. Agents are selfish, but the environment contains local opportunities where cooperation can benefit both agents.
A cooperation opportunity is useful only when the agents can meet at the relevant node at compatible times. An agent may therefore prefer to:
The central question is when cooperation emerges from individual incentives and how those incentives shape the resulting paths.
An instance is defined over a graph $G = (V, E)$, a set of agents $A$, start and target vertices for each agent, and node execution costs.
For each node $v \in V$:
For cooperation nodes, $\tau_2(v) < \tau_1(v)$.
Agents choose paths and timing strategies that minimize their own completion times. A cooperative execution can occur only when the relevant agents are co-located at a cooperation node at compatible times.
Intermittent strategic cooperation raises questions at the intersection of multi-agent planning, algorithms, and game theory:
Together, these repositories support both the algorithmic study of the problem and empirical exploration through simulation and visualization.
The ISCPP model is relevant to domains where autonomous or human-controlled agents are primarily self-interested but can occasionally benefit from local coordination:
Potential extensions include:
Researchers interested in multi-agent planning, algorithmic game theory, incentive-aware coordination, and cooperative decision-making are welcome to contribute ideas, algorithms, experiments, and extensions.
Please direct inquiries and collaboration proposals to:
i.shedlezki@gmail.comagmon@cs.biu.ac.ilIf you use the ISCPP model, associated algorithms, simulator, or datasets in academic work, please cite the relevant repository and accompanying publication.