Causal Expectations for Decentralised Coordination in Multi-Agent Systems

John R. E. Mills, Peter R. Lewis

Keywords: Expectation, Coordination, Contention, Beliefs, Popperian Expectations, Expectation Event Calculus, Multi-agent Systems, Causal Expectations, Causal Learning

TL;DR: Using causal expectations for coordination in MAS

Abstract: Coordination in multi-agent systems (MAS) is commonly seen as a problem of distributed decision-making under shared constraints. However, coordination within decentralised MAS environments also requires agents to anticipate the causal consequences of their own actions and those of others. This paper explores whether decentralised coordination can emerge through the creation and reuse of explicit causal expectations without communication, centralised control, or hardcoded coordination rules. We implement Popperian Expectations in autonomous agents operating in an intersection environment without coordination mechanisms. We represent causal expectations using the Expectation Event Calculus (EEC), enabling agents to form them internally and revise them when they are falsified. Over 2000 crossing attempts, agents achieved a 98.6\% task completion rate with 733,116 total simulations, compared to the baseline model which achieved 99.9\% task completion but with a total of 6,530,568 total simulations. The results show that agents converge to a stable set of causal expectation rules while significantly reducing the number of simulations needed, with minimal reduction in safety. The results suggest that explicit causal expectations can provide a sufficient mechanism for decentralised coordination in MAS.

Paper Type: Full (minimum of 10 pages and a maximum of 16 excluding references)

Poster Opt In: Yes, I'm open to having my submission accepted as a poster (leave blank if you are submitting a poster, or if you DON'T want your submission to be accepted as a poster instead of a full or short paper)

Email Sharing: We authorize the sharing of all author emails with Program Chairs.

Data Release: We authorize the release of our submission and author names to the public in the event of acceptance.

Submission Number: 9