CHAPTER 2

Lighting Automation and the ITS Framework

Research is needed to address several gaps in the application of SSL in roadway lighting and to consider the impact of emerging technologies, including advanced driver assistance systems (ADAS) and CAVs. While ADAS provide driver assistance, CAVs introduce communication capabilities, including vehicle-to-vehicle (V2V), vehicle-to-infrastructure (V2I), infrastructure-to-vehicle (I2V), vehicle-to-everything (V2X) communications.

This effort considered these issues, with the first section detailing an on-road evaluation of ADAS under varying LED lighting and environmental conditions to establish the impact of lighting condition on these systems. The second section of the report deals with the current state of connected vehicle (CV) technology and its implications for LED roadway lighting.

On-Road Evaluation of ADAS Performance

Based on the review of research of sensors used in automated vehicles (AVs) and ADAS, the following challenges and gaps in research were identified:

1. Sensor performance in poor weather conditions like rain, fog, and snow inhibit the use of AVs and ADAS. While some sensors, such as radar, are not majorly affected by poor weather, these sensors alone cannot provide enough information for AVs or ADAS to operate safely.
2. Sensor performance in low or no light conditions is another major issue that could inhibit the use of AVs or ADAS. Lidar systems could be used in no light conditions, but these sensors are expensive and the noise from the lidar data at night could also hinder AV performance (Jo et al., 2017).
3. Using multi-sensor fusion or V2I/I2V communication could cover some of these sensor performance issues. At night or in low-light conditions, AVs could communicate with streetlights, triggering them to brighten so that camera-based sensors could perform object detection tasks more accurately. However, more research is required on light levels required to maintain sensor performance at night or in low-light conditions.

In this first portion of the research, the effects of LED roadway lighting on ADAS were evaluated. The team conducted an experiment in which the performance of a vehicle equipped with an ADAS was measured under varying lighting and pavement conditions. The study specifically focused on the performance of the automatic emergency braking (AEB) system. The goal of the study was to evaluate the performance of the AEB system under varying lighting and pavement conditions to verify if the existing lighting standards for LEDs are adequate for the AEB system to perform optimally.

Research Questions

This research will allow state departments of transportation (DOTs) and municipalities to future-proof lighting systems under consideration for installation. The expected service life of newly installed LED luminaires will likely fall within the timeframe of emerging automated and autonomous vehicle technologies. From a lighting perspective, the key elements for consideration include:

• What is the possible impact of the roadway lighting system’s spectral content on ADAS?
• What is the impact of lighting levels on ADAS?
• What should be installed now to accommodate any future changes?
• What are the effects of varying pavement conditions (i.e., wet and dry) and the interactions between roadway lighting systems and ADAS technologies during these pavement conditions?
• If issues are discovered, a review of what changes in the lighting system could be considered to help mitigate the impacts of pavement conditions and other environmental factors.

Experimental Approach

In response to the gaps developed above, an initial research effort was undertaken to consider vehicles equipped with ADAS features, specifically, AEB. In this case, the investigation included the testing and evaluation of ADAS performance in a lighted environment. The results of this investigation are used to provide recommendations for any modifications that need to be made to lighting recommendations for these systems.

This experiment was conducted on the Virginia Smart Roads where an adult sized mannequin (Figure 1) was used as a potential hazard and an experimental vehicle was driven toward the hazard. For this study, the research team evaluated the performance of the test vehicle’s AEB system by recording whether the vehicle’s AEB subsystem was activated or not, the rate at which the vehicle slowed down under each of the experimental conditions, and the distance at which the vehicle started to brake.

The experiment included light levels and color temperatures consistent with existing practices for roadway lighting. Test conditions also included dry and wet pavement. The ADAS was evaluated at 20 and 30 mph. The proposed experimental design is shown in Table 1.

The research team used a 2019 Subaru Outback with Eyesight. The Eyesight system relies on stereovision systems. Stereovision systems use two or more cameras in conjunction to provide better depth of field (Rasshofer and Gresser, 2005) and additional color and texture information from the environment (Sivaraman and Trivedi, 2013).

Table 1. Independent variables and their experimental levels.

Results

The results were analyzed first as a logistical regression considering AEB activation and then by deceleration and distance for activation.

Repeated Measures Logistic Regression – Odds of AEB Activation

For most of the light type and level conditions, the odds of AEB activation were higher in the lighted conditions than in the no fixed roadway lighting condition. For two of the light type and level combinations (2200 K LED High and 3000 K LED Med) the odds of AEB activation were lower in the lighted condition than in the no fixed roadway lighting condition. The odds of AEB activation were lower (by 64%) when the mannequin was oriented perpendicular compared to parallel. The odds of AEB activation were lower at the lower vertical illuminance compared to the higher vertical illuminance (by 14%). In wet pavement conditions, the odds of AEB activation were lower by 28% compared to clear conditions. The odds of AEB activation were 72% lower in asphalt pavement compared to concrete pavement. Finally, the odds of AEB activation at 30 mph were 90% lower than the odds of AEB activation at 20 mph.

Deceleration

The main effects of mannequin orientation, pavement condition, pavement surface type and speed were significant. Post hoc pairwise comparisons of the deceleration under two different mannequin orientations showed that deceleration when the mannequin was parallel was higher than when the mannequin was perpendicular. Post hoc pairwise comparisons of deceleration under pavement types showed that deceleration was significantly higher on asphalt than on concrete pavement. Deceleration was statistically higher in wet pavement conditions than in the dry pavement condition. Finally, deceleration was significantly higher in the 30-mph condition than in the 20-mph condition.

Distance to Braking Initialization

The main effects of light type and level, vertical illuminance on mannequin, pavement type, and speed were significant. Post hoc pairwise comparisons of the distance to braking showed that in most of the cases, increase in light level resulted in an increase in the distance at which braking was initialized. The exception was the 2200 K LED High condition, where an increase in vertical illuminance on the mannequin resulted in an increase in the braking distance. The influence of pavement type on distance to braking initialization showed that the braking distance for concrete pavement was significantly higher than for asphalt pavement. Finally, the effect of speed on distance to braking initialization revealed that at 20 mph the distance at which braking initialized was statistically longer than the distance to braking initialization at 30 mph.

Discussion

These results have important implications for the use of AEB systems in detecting pedestrians on roadways at night. By identifying the light type and light level, and other environmental conditions that affect the accuracy of the system, these findings can inform the development and improvement of such systems to enhance pedestrian safety. For example, increasing the vertical illuminance of the pedestrian could improve the accuracy of the system, as could developing algorithms that can detect pedestrians in non-standard orientations. Interestingly, these results also indicate the correlated color temperature of the light source did not have a major effect on the performance of the AEB system. Additionally, the findings suggest that using the system on well-lit roads (greater than 0.3 cd/m²), dry pavement, and on concrete pavement could result in more accurate AEB activations.

This study has some limitations. First, only one vehicle’s AEB system was evaluated in varying lighting and pavement conditions. This approach was taken as other vehicles’ systems could not be integrated with the data acquisition system (DAS) and key performance measures could not be collected. Second, there was no additional traffic on the Virginia Smart Roads during testing. Third, the mannequin which was used as a simulated pedestrian did not move and was stationary, whereas in reality, pedestrians are never stationary. Future research should focus on addressing these limitations to add to the body of knowledge on the performance of Level 2 ADAS in varying lighting and pavement conditions.

Finally, as machine vision technologies rely on visible light for AEB, they are also significantly affected by the same limitations as human eyes. For optimal performance at night, an AEB system should be able to detect pedestrians and track them on the vehicle’s approach to avoid collisions. Absence of road lighting or not enough of it could provide insufficient visual information to the machine vision system and it might become extremely difficult for the system to track pedestrians and other hazards, thus adversely affecting system performance. For AEBs to perform optimally at night, system designers should thus consider using alternative approaches, such as measuring color contrast, information from secondary sensors (e.g., radar), or infrared, in addition to using machine vision.

Conclusion

In conclusion, the findings from the three analysis approaches provide important insights into the factors that affect the performance of AEB systems. The results suggest that lighting conditions, mannequin orientation, vertical illuminance, pavement surface type, pavement condition, and speed are all significant factors that affect the performance of AEB systems. It is noteworthy that these systems show the same limitations on performance as human eyes. Therefore, it is expected that the same lighting configuration that is used for humans is effective for these systems. These findings have important implications for the design and implementation of AEB systems and can help improve the safety of automated and autonomous vehicles and reduce the risk of accidents on the road.

Recommendations

Based on the results of the study, the following recommendations can be made to increase the accuracy of ADAS, particularly regarding AEBs under roadway lighting conditions. Readers should recall that these recommendations are based on data from a vehicle traveling at speeds of 25 mph and 30 mph.

• Roadways should be illuminated to average luminance of greater than 0.3 cd/m². Based on the testing conducted in this study, vertical illuminance of 20 lux is recommended for pedestrians for optimal performance of AEB systems. The current American Association of State Highway and Transportation Officials (AASHTO) Roadway Lighting Design Guide recommends horizontal illuminance of between 14 and 24 lux in areas of high pedestrian use; the Guide indicates that some lighting guides recommend equal horizontal and vertical luminance (AASHTO, 2018).
• On roadway surfaces that are darker, like asphalt, a higher light level is recommended to increase the accuracy of ADAS that rely on machine vision.
• On roadways with speeds higher than 20 mph, it is important to augment the AEB system with additional sensors to improve their accuracy and performance.

Connected Vehicle Technology and its Implications for Solid-State Lighting

This second part of this activity considered the current state of CV/infrastructure technology. As highlighted in the experimental effort in the previous section, there is significant variability in the performance of even one vehicle’s use of an AEB system under various lighting conditions. To enhance this discussion, consideration should be given to the potential impact of CV technological and roadway lighting.

A CV is a vehicle equipped with some sort of wireless communication device that allows it to share information with other vehicles and objects on the roadway. CV technologies enable vehicles to communicate with infrastructure (V2I), between vehicles (V2V), and with other objects on the roadway such as bicycles, pedestrians, or obstacles (V2X).

There are many potential mediums by which connectivity can be enabled. Satellite, cellular, Wi-Fi and other short-range communications all represent methods by which vehicles today are already connected, and the vehicles of tomorrow will become increasingly connected. DSRC, a Wi-Fi-dedicated short-range communication method that has been developed for high-speed low-latency situations to specifically enable safety applications is one such medium. C-V2X, which is cellular-based, is another.

Emerging CAVs and ADAS-equipped vehicles receive information from various sensors and devices and use this information to inform decisions on the vehicle’s movement. While many sensors, such as GPS and odometry, are not sensitive to light conditions, others such as visual cameras, Light Detection and Ranging (lidar), and radar, may be impacted by light in certain conditions, such as dawn and dusk, full darkness, glare in bright sunlight, and/or rapid changes in lighting conditions when entering and exiting settings such as tunnels and underpasses. However, research on the impact of these conditions is limited.

Other devices and sensors that ADAS-equipped vehicles use are not sensitive to light or weather conditions. These sensors and devices can be used to provide redundancy for when the other sensors are limited in their abilities.

Conclusion and Future Directions

The Federal Communications Commission’s (FCC’s) next step will be to release a second Report & Order that will better define how the 30 MHz communication range can be utilized. They may also begin responding to waiver requests for state DOTs and automakers to begin using C-V2X. While C-V2X equipment is not yet available commercially off-the-shelf in the United States, it does have more widespread use in other countries and use in the U.S. is expected to increase soon, as industry uncertainties are resolved. The technology debate is over, and industry is moving from DSRC to C-V2X, as the desire for an interoperable system trumps any preference for DSRC over C-V2X that may have existed prior to the FCC’s actions being confirmed by the courts.

State DOTs and other industry leaders are working on developing a path forward to deployment for V2X technology and the USDOT may ensure consistency and lessen uncertainty. These new procedures will take time to unfold, and new devices will take time to mature. Federal leadership, vision, and direction will be essential to success going forward, and agencies are working with industry to develop a national vision for an interoperable and cyber-secure CV program.

Recommendations

As this report indicates, the uncertainty in the future usage of the bandwidth and the CV2X environment leads to uncertainty in the recommendations for the lighting systems and the future links between lighting and vehicles. The current recommendations are:

1. Install controls-ready luminaires that can be converted to control systems when the link to vehicles through CV2X or other methodologies are desired.
2. If a control system is installed, ensure that the system has link or API subroutine calls that can be utilized to the connect the systems to CV2X implementations.