Products/fuzzyTECH Editions/Benchmarks

All digital computers have been developed on the basis of Boolean 0/1 type Logic. Thus computing the fuzzy logic algorithm is quite hard for standard microcontrollers and initial implementations of the fuzzy logic algorithm in software were relatively slow and used quite substantial ROM and RAM memory.

We have worked very hard over the past 8 years to optimized the fuzzy logic algorithm in a way that even low-end microcontrollers that may cost less than $1 in quantities can run fuzzy logic systems at high speeds. About a decade ago, a simple fuzzy logic controller took about 1 second computation time and required up to 20KB ROM space on the standard 8051 microcontroller at 12 MHz. Our optimized fuzzy logic solution now lets you compute the same fuzzy logic system within 1 millisecond requireing less than 500 bytes of ROM. Fast 16-bit microcontrollers, such as the 80C166, do compute even very large fuzzy logic systems within less than 0.5 milliseconds. If your application requires even faster response times, fuzzy coprocessors, such as the benchmarked 81C99x deliver even higher performance.

For a detailed comparison of the computing requirements of fuzzy logic systems on different hardware platform, we have created a set of 5 different standardized fuzzy logic systems as benchmark systems. These fuzzy logic systems (Bench 0 to Bench 4) stem from industrial fuzzy logic applications and represent typical configurations:

• Bench 0: Fuzzy Logic Position Controller (7 Rules, 2 Inputs with 3 terms each, 1 Output with 5 terms, rules weights = 1)
  
• Bench 1: Fuzzy PI-Controller (20 Rules, 2 Inputs with 5 terms each, 1 Output with 5 terms, rules weights = 1)
  
• Bench 2: Fuzzy PD-Controller (20 Rules, 2 Inputs with 5 terms each, 1 Output with 5 terms, various rules weights)
  
• Bench 3: Parameter Estimator for ABS (80 Rules, 3 Inputs with 3/4/6 terms, 1 Output with 5 terms, various rules weights)
  
• Bench 4: Intervehicle Dynamic Controller (500 Rules, 8 Inputs with 7 terms, 4 Outputs with 7 terms, various rules weights)

Bench 4 stems from a research application and is the largest known embedded fuzzy logic system.

The following table shows the benchmark results for typical hardware platforms. Notice that for the PC platform, we have selected an Intel processor with no cash memory and no floating point unit, so that you can calculate the performance of faster processors more easily.

All computing times are given as average / maximum (average case / worst case)