Multi-agent hybrid system simulation

In this paper a technique is presented for simulating a set of hybrid systems (agents) whose continuous dynamics are decoupled; however coupling is introduced in the form of constraints (inequalities whose violation signifies the occurrence of a discrete event). We introduce a step size selection algorithm, motivated by the control theoretic concept of input/output linearization, which allows two or more agents to be integrated asynchronously using different step sizes when the state is away from the constraint. When the state approaches a constraint, the algorithm is able to bring the two local time clocks into synchronization and localize the time of constraint violation. The algorithm is guaranteed to never miss or overshoot the constraints, eliminating the need for simulation roll-back.