Multi-agent controller

Presentation of our tool

Our tool is an implementation, based on python and PPLPY, of a three-layer method:

SWA-SMT solver: verifying reachability and compute a trace of a successful run.
This method is based on Stopwatches Timed Automata synchronized with channels and SMT solver.

RL: get a real-time policies for a controller for a multi-agent system.
This method is based on Reinforcement Learning and using our previously computed dataset of SWA-SMT solutions.

Here is few episodes played by a neural network policy trained with TD3BC:

Multi-agent controller

An implementation of a controller based on SMT, Timed Automata and Reinforcement Learning.

Presentation of our tool