Skip to Main Content
 

Global Search Box

 
 
 
 

ETD Abstract Container

Abstract Header

EVALUATION OF PERFORMANCE MEASURES EXHIBITED BY DRIVING SIMULATION PARTICIPANTS

Koneti, Sai Sandeep

Abstract Details

2016, Master of Science in Civil Engineering, Cleveland State University, Washkewicz College of Engineering.
The topic of this thesis is the use of steering measures to evaluate the improvement in performance exhibited by those participating in driving simulation experiments. If the steering measure improves over time, it may be an indicator that the participant is learning to drive the simulator. Understanding which steering measure best shows improvement is therefore valuable to future driving simulation experiments where identifying whether or not participants have improved their performance is crucial to the validity of the simulation experimental results. In the literature, a variety of measures describing steering and speed control have been used to evaluate driving performance on several maneuvers and/or tasks. Learning has been described as the reduction of steering variability, a decrease in the distance of the vehicle from the center of the travel lane, an increase in average speed, and a decrease in travel time. Various researchers have reported observing an improvement over time in the case of lane keeping and over trials of specific maneuvers such as lane changing. To examine whether a steering input measure or a lane position measure is superior in identifying the performance improvement of participants, data from a previously conducted driving simulation experiment were used. The experimental scenario was of a 2-lane rural road with alternating straight road segments and horizontal curves. There were 120 straight segments, each 400 m long and 120 horizontal curves, each subtending 90 degrees with a radius of 200 m. The posted speed limit was 50 mph (80 km/h). The scenario had been driven by twenty participants, 7 females and 13 males. Each participant drove the same scenario in either the RS-100 or the RS-600 driving simulator. The data for each of the 20 participants for the first 20 horizontal curves and straight segments were reduced into 6 performance measures: average lane position (AVE(LP)), standard deviation of lane position (SD(LP)), root mean square of lane position (RMS(LP)), average steering input (AVE(SI)), standard deviation of steering input (SD(SI)), and root mean square of steering input (RMS(SI)) . To distinguish between relatively good behavior and poor behavior, the number of times the simulator vehicle was out of the intended travel lane was counted for each horizontal curve and straight segment. The purpose of this thesis was to examine whether a steering input measure or a lane position measure is superior in identifying the performance improvement of participants. Based on the findings from McGehee et al (2004), the variability of steering input is expected to decrease over time. Additionally, the findings from Sahami and Sayed (2010, 2013) suggest that the standard deviation of lane position will decrease over time. What is not clear is whether the steering input or the standard deviation of lane position is the better, or more sensitive measure. A visual analysis of graphical plots is conducted on the data for each of the performance measures. A measure which shows large pertubations is preferred over a measure which shows small pertubations because large pertubations can be easily identified. These large pertubations are expected to decrease over time when learning has occurred. Based on the instabilities observed in the various performance measure data, SD(LP) was found to be the better measure. The number of large pertubations decreased thus implying that learning had occurred. The improvement was more apparent for SD(LP) than the other performance measures. The scaling of the axes on the performance plots are hiding some of the interesting changes in performance. Therefore, a quantitative analysis of the percentage change in performance measure values between successive horizontal curves and successive straight segments was conducted. For horizontal curves, lane position measures AVE(LP) and SD(LP) showed the greatest percentage change most frequently. For the straigh segments, the lane position measure AVE(LP) and steering input measure (AVE(SI) showed the greatest percentage change most often. Comparison of means tests were conducted to test whether the good and poor driving were stastically different. Each of the performance measures were examined. For horizontal curves, SD(LP) and AVE(LP) were found to be significantly less for good driving than poor driving. Similarly, for straight segments, AVE(LP) was found to be significnatly less. Based on the three different analyses that were performed, AVE(LP) is the superior measure. It has shown to distinguish good from poor driving in the comparison of means test. It is also the measure most often having the greatest percentage change in successive horizontal curves and successive straight segments.
Jacqueline M. Jenkins, Ph.D. (Advisor)
Lutful Khan I., Ph.D. (Committee Member)
Mehdi Jalalpour, Ph.D. (Committee Member)
104 p.

Recommended Citations

Citations

  • Koneti, S. S. (2016). EVALUATION OF PERFORMANCE MEASURES EXHIBITED BY DRIVING SIMULATION PARTICIPANTS [Master's thesis, Cleveland State University]. OhioLINK Electronic Theses and Dissertations Center. http://rave.ohiolink.edu/etdc/view?acc_num=csu147566932339873

    APA Style (7th edition)

  • Koneti, Sai Sandeep. EVALUATION OF PERFORMANCE MEASURES EXHIBITED BY DRIVING SIMULATION PARTICIPANTS. 2016. Cleveland State University, Master's thesis. OhioLINK Electronic Theses and Dissertations Center, http://rave.ohiolink.edu/etdc/view?acc_num=csu147566932339873.

    MLA Style (8th edition)

  • Koneti, Sai Sandeep. "EVALUATION OF PERFORMANCE MEASURES EXHIBITED BY DRIVING SIMULATION PARTICIPANTS." Master's thesis, Cleveland State University, 2016. http://rave.ohiolink.edu/etdc/view?acc_num=csu147566932339873

    Chicago Manual of Style (17th edition)