Almost all multi-agent reinforcement learning algorithms without communication follow the principle of centralized training with decentralized execution. During the centralized training, agents can be guided by the same signals, such as the global state. However, agents lack the shared signal and choose actions given local observations during execution. Inspired by viewpoint invariance and contrastive learning, we propose consensus learning for cooperative multi-agent reinforcement learning in this study. Although based on local observations, different agents can infer the same consensus in discrete spaces without communication. We feed the inferred one-hot consensus to the network of agents as an explicit input in a decentralized way, thereby fostering their cooperative spirit. With minor model modifications, our suggested framework can be extended to a variety of multi-agent reinforcement learning algorithms. Moreover, we carry out these variants on some fully cooperative tasks and get convincing results.

Multi-UAV path planning presents a critical challenge in Unmanned Aerial Vehicle (UAV) applications, particularly in environments with various obstacles and restrictions. These conditions transform multi-UAV path planning into a complex optimization problem with multiple constraints, significantly reducing the number of feasible solutions and complicating the generation of optimal flight trajectories. Although the slime mold algorithm (SMA) has proven effective in optimization missions, it still suffers from limitations such as inadequate exploration capacity, premature convergence, and a propensity to become stuck in local optima. These drawbacks degrade its performance in intricate multi-UAV scenarios. This study proposes a self-adaptive improved slime mold algorithm called AI-SMA to address these issues. Firstly, AI-SMA incorporates a novel search mechanism to balance exploration and exploitation by integrating ranking-based differential evolution (rank-DE). Then, a self-adaptive switch operator is introduced to increase population diversity in later iterations and avoid premature convergence. Finally, a self-adaptive perturbation strategy is implemented to provide an effective escape mechanism, facilitating faster convergence. Extensive experiments were conducted on the CEC 2017 benchmark test suite and multi-UAV path-planning scenarios. The results show that AI-SMA improves the quality of optimal fitness by approximately $7.83\%$ over the original SMA while demonstrating superior robustness and effectiveness in generating collision-free trajectories.

Almost all multi-agent reinforcement learning algorithms without communication follow the principle of centralized training with decentralized execution. During centralized training, agents can be guided by the same signals, such as the global state. During decentralized execution, however, agents lack the shared signal. Inspired by viewpoint invariance and contrastive learning, we propose consensus learning for cooperative multi-agent reinforcement learning in this paper. Although based on local observations, different agents can infer the same consensus in discrete space. During decentralized execution, we feed the inferred consensus as an explicit input to the network of agents, thereby developing their spirit of cooperation. Our proposed method can be extended to various multi-agent reinforcement learning algorithms with small model changes. Moreover, we carry out them on some fully cooperative tasks and get convincing results.