Trace Generator for Plan Recognition with One or More Agents

Multi-agent Blocks World
TraceGen is a python implementation of a trace generator for plan recognition with one or more agents. It uses a collection of partially ordered plan graphs (in text format) as the input plan library, and produces the activity traces of a given number of agents (n) for a given number of steps (T). If n>1, then TraceGen will randomly partition the agents into (dynamic) teams and make each team execute a randomly selected plan from the library. You may also make each team (or the agent, if n=1) interleave the execution of multiple plans with the help of an input parameter that controls the probability of interleaving.

TraceGen was created for developing and testing multi-agent plan recognition algorithms. If you find it useful, we would appreciate if you let us know.

Documentation

Download

Publication that uses TraceGen

Contact