The research constituting this thesis originates from a Federal Highway Administration's Graduate Research Fellow- ship entitled "Calibration of RFLO Traffic Simulation Program." The RFLO Simulation Program provides a macroscopic description of vehicular traffic flow. The majority of the research was performed at the FHWA's Turner-Fairbank Highway Research Center (McLean, Va.) during the time from June 15 to September 15, 1989. The actual calibration procedure centered around RFLO's Relaxation Constant parameter, a partial differential equation - containing the parameter - representing the acceleration/deceleration of macroscopic traffic, and the microscopic traffic data sets that were available at the research facility.
Initially, the microscopic traffic data needed trans- formed into a macroscopic traffic description in order to facilitate a legitimate comparison to RFLO. While becoming acquainted with the data sets several types of errors and misconceptions were identified and subsequently streamlined from the recorded traffic descriptions. The transformation process would create quantities of Volume (Q), Density (k), and Speed (v) describing a facility's traffic behavior de- pendent upon a distance region (DX) and time interval (DT). In order to justify these averaging regions a DX-DT Aggrega- tion Analysis resulted. This analysis used carefully selected DX-DT combinations and the Rational Subgroupping Concept (Montgomery, 1985) to arrive at the conclusion, for the Roscoe Boulevard data set, of using a DX of 200 feet and a DT equal to 3 seconds. By performing the transformation to this aggregation level an "optimal" overall macroscopic description of the Roscoe Boulevard's traffic behavior was attained.
The calibration of RFLO's Relaxation Constant was expected to provide additional agreement with Prigogine and Herman (1971) whom suggest that the (RFLO) "forcing func- tion" is a disastrous simplification of the manner and abil- ity for which macroscopic traffic reacts to changing condi- tions within the transportation medium. The calibration procedure isolated the "forcing function," from various constraints within the model, by selecting a traffic re- sponse that would most effectively represent a very quick and unconstrained acceleration to the free flow speed of the facility. After estimating the required variables necessary to RFLO, for facility definition, the calibration procedures were performed. These procedures used a wide range of Relaxation Constant values and a three-dimensional calibra- tion methodology while attempting to minimize the difference between the actual behavior data and that produced by the RFLO model. The results of the calibration indicate that RFLO does react more efficiently (quickly), than the actual traffic behavior, to such an extent that the largest relaxation time could not retard this behavior.
The last examination of this thesis pertains to non- equilibrium behavior, an optimal aggregation level for mac- roscopic traffic representation, and a concept towards manlpulating traffic data by means of a three-dimensional visual model. Previous research has produced two (Wagner & May, 1963) and three-dimensional (Makigami et al, 1985) visual traffic representations which have provided a better understanding of the actual behavior occurring. However, with the aforementioned "optimal" macroscopic aggregation level and several manipulative concepts it may be possible to enable nonequilibrium traffic phenomenon to be detected from visual models of various traffic behavior character- istics. The results of this examination provide several suggestions. However, the representations of vehicular speed standard deviations, fully composing each and every macroscopic speed, and a coefficient of variation model provide the best outcomes for the final investigation of this thesis. Although these visual models are suggested in the thesis the three-dimensional graphing capabilities, offered at the Turner-Fairbank Highway Research Center, limited the extent for proving their actual value.