Improving correctness of finite-state machine synthesis from multiple partial input/output sequences

Abstract

Our previous work focused on the synthesis of sequential circuits based on a partial input/output sequence. As the behavioural description of the target circuit is not known the correctness of the result can not be verified. This paper proposes a method which increases the correctness percentage of the finite-state machine (FSM) synthesis using multiple partial input/output sequences. The synthesizer is based on Genetic Algorithm. The experimental results show that the correctness percentage can be increased to 100% by increasing of the number of input/output sequences.

Improvements in FSM Evolutions from Partial Input/Output Sequences

“…These output sequences are compared with the correct (desired) output sequences (the outputs of the TSs) and a fitness value is assigned to each candidate solution. Unlike other approaches [3,4], the proposed methodology includes a heuristic to help the algorithm to evade from local optima. One of the main causes of premature convergence is loss of genetic diversity due to selection pressure [13].…”

“…4) and an estimated value of five for S, leading to a 168-input/output TS (see subsection 3.1). Moreover, it is desirable to have multiple TSs to improve the performance of the learning system (see [4]). In fact, eighteen TSs (N = 18) were used, each with 32 bits in length, which corresponds to more than three 168-length sequences.…”

“…More recent attempts have been successful in synthesizing sequential systems with the aid of Genetic Algorithms (GAs) [2,3,4]. In [1] an approach to synthesize synchronous sequential logic circuits from partial I/O sequences by using technologybased representation (a netlist of gates and flip-flops) is presented.…”

“…The method is able to synthesize a variety of small FSMs, such as serial adders and 4-bit sequence detectors. With the same purpose other works [3,4] used technology-independent state-based representation in which the next-state and the output, corresponding to each current-state/input pair of the state-transition table, are coded in a binary string defined as the chromosome. Because, in this case, the number of states of the finitestate automaton is unknown, a large number of states must be used at the start.…”

A Scenario-Based Approach to Protocol Design Using Evolutionary Techniques

“…Lacking the crossover genetic operator [10] the approach has shown, however, poor performances. Recent works have been successful in synthesizing sequential systems with the aid of Genetic Algorithms (GAs) [6,7,8]. An approach to synthesize synchronous sequential logic circuits from partial input/output (I/O) sequences is presented in [6].…”