Boolean Network Robotics as an Intermediate Step in the Synthesis of Finite State Machines for Robot Control

We propose an approach to the automatic synthesis of robot control software based on the finite state machine (FSM) formalism. In our previous research, we have introduced Boolean network robotics as a novel approach to the automatic design of robot control software. In this paper, we show that it is possible to leverage automatically designed Boolean networks to synthesize FSMs for robot control. Boolean network robotics exhibits a number of interesting properties. Firstly, notwithstanding the large size of the state space of a Boolean network and its ability to display complex and rich dynamics, the automatic design is able to produce networks whose trajectories are confined in small volumes of the state space. Secondly, the automatic design produces networks in which one can identify clusters of states associated with functional behavioral units of the robots. It is our contention that the automatic design of a Boolean network controller can be a convenient intermediate step in the synthesis of a FSM, which offers the advantage of being a compact, readable, and modifiable representation. In this paper, we show that clusters of states traversed by network trajectories can be mapped to states of a FSM. We illustrate the viability of our proposal in two notable robotic tasks, namely collision avoidance and sequence recognition. The first task can be achieved by a memoryless control program, whilst in the second the robots need memory.