Speed Distribution Based Approach for Shockwave Detection in a Connected Driving Environment

One of the key advantages of connectivity in highway environments is the possibility of shockwave detection at the onset of formation, which can provide more flexibility in mitigating congestion. Past research used data from radar guns and loop detectors to show that a rapid increase in speed variance could be an indicator of shockwave formation. This paper investigates the possibility of utilizing connected vehicles data and vehicle trajectory data to determine if any increase in speed variance over distances could be an indicator of shockwave formation. Moreover, the effects of limited information in a connected driving environment on shockwave detection based on speed variance were explored. Two datasets were evaluated: I-5 Connected Vehicles dataset and NGSIM US 101 dataset. Six segments analyzed in the I-5 dataset showed that a jump in speed variance could detect congestion earlier than looking at average speed alone. The NGSIM US 101 scenarios of 100, 50 and 10 percent market penetration rates (MPRs) were analyzed assuming 100, 80, and 50 percent of speed data were received at each time step. For MPRs of 100 and 50 percent, speed variance was able to identify the six shockwaves in the dataset. The RMSE, calculated for various MPRs, showed an inverse relationship to MPR. The impact of misinformation from potential cyberattacks or equipment malfunctions was also tested on the US 101 dataset. Speed variance was more robust than average speed when speeds were reported either higher or lower than actual speeds. When speeds were falsely reported as a combination of higher and lower than actual speeds, variance continually increased, though a jump in variance was still an indication of shockwave formation. When incorrect speeds were reported for only a high variance interval by 1-5 mph and 5-10 mph, speed variance remained a strong indicator of congestion formation. Analyzing the US 101 dataset with larger distance intervals, by individual lanes, and by different lane aggregations improved variance based shockwave detection reliability. Shockwaves detected earlier and more reliably can delay shockwave propagation and further reduce negative impacts on safety, performance, and emissions