In cooperative multi-agent reinforcement studying (MARL), as a consequence of its on-policy nature, coverage gradient (PG) strategies are usually believed to be much less pattern environment friendly than worth decomposition (VD) strategies, that are off-policy. Nonetheless, some latest empirical research reveal that with correct enter illustration and hyper-parameter tuning, multi-agent PG can obtain surprisingly robust efficiency in comparison with off-policy VD strategies.

Why might PG strategies work so properly? On this submit, we’ll current concrete evaluation to point out that in sure eventualities, e.g., environments with a extremely multi-modal reward panorama, VD might be problematic and result in undesired outcomes. In contrast, PG strategies with particular person insurance policies can converge to an optimum coverage in these circumstances. As well as, PG strategies with auto-regressive (AR) insurance policies can be taught multi-modal insurance policies.

Determine 1: completely different coverage illustration for the 4-player permutation sport.

CTDE in Cooperative MARL: VD and PG strategies

Centralized coaching and decentralized execution (CTDE) is a well-liked framework in cooperative MARL. It leverages world data for more practical coaching whereas conserving the illustration of particular person insurance policies for testing. CTDE might be carried out by way of worth decomposition (VD) or coverage gradient (PG), main to 2 various kinds of algorithms.

VD strategies be taught native Q networks and a mixing operate that mixes the native Q networks to a worldwide Q operate. The blending operate is normally enforced to fulfill the Particular person-World-Max (IGM) precept, which ensures the optimum joint motion might be computed by greedily selecting the optimum motion domestically for every agent.

In contrast, PG strategies straight apply coverage gradient to be taught a person coverage and a centralized worth operate for every agent. The worth operate takes as its enter the worldwide state (e.g., MAPPO) or the concatenation of all of the native observations (e.g., MADDPG), for an correct world worth estimate.

The permutation sport: a easy counterexample the place VD fails

We begin our evaluation by contemplating a stateless cooperative sport, particularly the permutation sport. In an $N$-player permutation sport, every agent can output $N$ actions ${ 1,ldots, N }$. Brokers obtain $+1$ reward if their actions are mutually completely different, i.e., the joint motion is a permutation over $1, ldots, N$; in any other case, they obtain $0$ reward. Be aware that there are $N!$ symmetric optimum methods on this sport.

Determine 2: the 4-player permutation sport.

Determine 3: high-level instinct on why VD fails within the 2-player permutation sport.

Allow us to deal with the 2-player permutation sport now and apply VD to the sport. On this stateless setting, we use $Q_1$ and $Q_2$ to indicate the native Q-functions, and use $Q_textrm{tot}$ to indicate the worldwide Q-function. The IGM precept requires that

[argmax_{a^1,a^2}Q_textrm{tot}(a^1,a^2)={argmax_{a^1}Q_1(a^1),argmax_{a^2}Q_2(a^2)}.]

We show that VD can’t signify the payoff of the 2-player permutation sport by contradiction. If VD strategies had been in a position to signify the payoff, we’d have

[Q_textrm{tot}(1, 2)=Q_textrm{tot}(2,1)=1quad text{and}quad Q_textrm{tot}(1, 1)=Q_textrm{tot}(2,2)=0.]

If both of those two brokers has completely different native Q values (e.g. $Q_1(1)> Q_1(2)$), we’ve $argmax_{a^1}Q_1(a^1)=1$. Then in accordance with the IGM precept, any optimum joint motion

[(a^{1star},a^{2star})=argmax_{a^1,a^2}Q_textrm{tot}(a^1,a^2)={argmax_{a^1}Q_1(a^1),argmax_{a^2}Q_2(a^2)}]

satisfies $a^{1star}=1$ and $a^{1star}neq 2$, so the joint motion $(a^1,a^2)=(2,1)$ is sub-optimal, i.e., $Q_textrm{tot}(2,1)<1$.

In any other case, if $Q_1(1)=Q_1(2)$ and $Q_2(1)=Q_2(2)$, then

[Q_textrm{tot}(1, 1)=Q_textrm{tot}(2,2)=Q_textrm{tot}(1, 2)=Q_textrm{tot}(2,1).]

Consequently, worth decomposition can’t signify the payoff matrix of the 2-player permutation sport.

What about PG strategies? Particular person insurance policies can certainly signify an optimum coverage for the permutation sport. Furthermore, stochastic gradient descent can assure PG to converge to certainly one of these optima underneath gentle assumptions. This means that, although PG strategies are much less well-liked in MARL in contrast with VD strategies, they are often preferable in sure circumstances which are frequent in real-world purposes, e.g., video games with a number of technique modalities.

We additionally comment that within the permutation sport, with the intention to signify an optimum joint coverage, every agent should select distinct actions. Consequently, a profitable implementation of PG should be certain that the insurance policies are agent-specific. This may be achieved by utilizing both particular person insurance policies with unshared parameters (known as PG-Ind in our paper), or an agent-ID conditioned coverage (PG-ID).

PG outperforms current VD strategies on well-liked MARL testbeds

Going past the easy illustrative instance of the permutation sport, we prolong our research to well-liked and extra sensible MARL benchmarks. Along with StarCraft Multi-Agent Problem (SMAC), the place the effectiveness of PG and agent-conditioned coverage enter has been verified, we present new ends in Google Analysis Soccer (GRF) and multi-player Hanabi Problem.

Determine 4: (left) profitable charges of PG strategies on GRF; (proper) greatest and common analysis scores on Hanabi-Full.

In GRF, PG strategies outperform the state-of-the-art VD baseline (CDS) in 5 eventualities. Apparently, we additionally discover that particular person insurance policies (PG-Ind) with out parameter sharing obtain comparable, typically even larger profitable charges, in comparison with agent-specific insurance policies (PG-ID) in all 5 eventualities. We consider PG-ID within the full-scale Hanabi sport with various numbers of gamers (2-5 gamers) and examine them to SAD, a robust off-policy Q-learning variant in Hanabi, and Worth Decomposition Networks (VDN). As demonstrated within the above desk, PG-ID is ready to produce outcomes akin to or higher than the most effective and common rewards achieved by SAD and VDN with various numbers of gamers utilizing the identical variety of atmosphere steps.

Past larger rewards: studying multi-modal conduct by way of auto-regressive coverage modeling

In addition to studying larger rewards, we additionally research the best way to be taught multi-modal insurance policies in cooperative MARL. Let’s return to the permutation sport. Though we’ve proved that PG can successfully be taught an optimum coverage, the technique mode that it lastly reaches can extremely rely on the coverage initialization. Thus, a pure query shall be:

Can we be taught a single coverage that may cowl all of the optimum modes?

Within the decentralized PG formulation, the factorized illustration of a joint coverage can solely signify one specific mode. Subsequently, we suggest an enhanced approach to parameterize the insurance policies for stronger expressiveness — the auto-regressive (AR) insurance policies.

Determine 5: comparability between particular person insurance policies (PG) and auto-regressive insurance policies (AR) within the 4-player permutation sport.

Formally, we factorize the joint coverage of $n$ brokers into the type of

[pi(mathbf{a} mid mathbf{o}) approx prod_{i=1}^n pi_{theta^{i}} left( a^{i}mid o^{i},a^{1},ldots,a^{i-1} right),]

the place the motion produced by agent $i$ relies upon by itself statement $o_i$ and all of the actions from earlier brokers $1,dots,i-1$. The auto-regressive factorization can signify any joint coverage in a centralized MDP. The solely modification to every agent’s coverage is the enter dimension, which is barely enlarged by together with earlier actions; and the output dimension of every agent’s coverage stays unchanged.

With such a minimal parameterization overhead, AR coverage considerably improves the illustration energy of PG strategies. We comment that PG with AR coverage (PG-AR) can concurrently signify all optimum coverage modes within the permutation sport.

Determine: the heatmaps of actions for insurance policies realized by PG-Ind (left) and PG-AR (center), and the heatmap for rewards (proper); whereas PG-Ind solely converge to a selected mode within the 4-player permutation sport, PG-AR efficiently discovers all of the optimum modes.

In additional complicated environments, together with SMAC and GRF, PG-AR can be taught fascinating emergent behaviors that require robust intra-agent coordination which will by no means be realized by PG-Ind.

Determine 6: (left) emergent conduct induced by PG-AR in SMAC and GRF. On the 2m_vs_1z map of SMAC, the marines maintain standing and assault alternately whereas guaranteeing there is just one attacking marine at every timestep; (proper) within the academy_3_vs_1_with_keeper state of affairs of GRF, brokers be taught a “Tiki-Taka” fashion conduct: every participant retains passing the ball to their teammates.

Discussions and Takeaways

On this submit, we offer a concrete evaluation of VD and PG strategies in cooperative MARL. First, we reveal the limitation on the expressiveness of well-liked VD strategies, displaying that they might not signify optimum insurance policies even in a easy permutation sport. In contrast, we present that PG strategies are provably extra expressive. We empirically confirm the expressiveness benefit of PG on well-liked MARL testbeds, together with SMAC, GRF, and Hanabi Problem. We hope the insights from this work may gain advantage the group in the direction of extra basic and extra highly effective cooperative MARL algorithms sooner or later.

This submit relies on our paper: Revisiting Some Widespread Practices in Cooperative Multi-Agent Reinforcement Studying (paper, web site).