The lessons learned from this project include Return on Investment (ROI) for CAWS/AEB, testing procedures, the importance of bus original equipment manufacturer (OEM) involvement, transit research challenges, retrofitting buses with new tech, and future research data needs for CAWS/AEB.

• Incorporate driver actions into CAWS/AEB reaction algorithms to reduce unnecessary warnings. This project revealed that for in many instances where warnings or braking could be triggered, drivers were already aware of or taking action to prevent collisions when the CAWS/AEB sensor was activated. For instance, a driver already decelerating due to anticipated collision risks might find a CAWS warning redundant and annoying.
• Consider additional sensor augmentation or fusion to reduce false positives when using LIDAR in sensing vulnerable road users. This study found that LIDAR with sub-optimal sensor placements led to false positives, particularly with highly reflective objects such as road signs. Adjusting sensor mounting or improving detection algorithms, along with sensor augmentation/fusion or computer vision, could mitigate these issues.
• Develop noise filtering algorithms for data processing streams. This project found that real-time bus motion data analysis was challenged by signal noise from multiple sensors. Having noise filtering algorithms in the data processing stream would have been beneficial.
• Develop a local area testing plan in consultation with local roadway authorities and the CAWS/AEB vendor. Unlike the level urban intersection simulation on the Smart Road track, Pierce Transit buses navigated steep grades in real conditions, necessitating adjustments to the PASS system.
• Utilize existing data collection systems to evaluate alterations in vehicle operation due to CAWS/AEB. Many transit agencies already have on-board data collection systems that can be valuable tools for evaluating the impacts of CAWS/AEB. . In this project, an existing system of video cameras and recorders were used to measure CAWS accuracy. Historical video clips can help estimate past collisions and gauge CAWS/AEB's potential prevention impact. These systems could also potentially integrate into the CAWS detection system.
• Allow ample time for contract negotiations. It is especially important to confirm early in the proposal stage that all partners agree on data sharing and integration of components.
• Be creative and flexible when retrofitting the technology for buses. For example, in this project, it was observed that folding bicycle racks occupied prime sensor space on the bus fronts. To address this, the vendor mounted the sensors on a bracket beneath the bicycle rack.
• Design retrofit equipment to be as compact as possible and explore options to rearrange current equipment to fit new components. Given the limited space in a pre-existing electrical cabinet housing various equipment, there was a need to reconfigure and optimize space.
• Employ ruggedized automotive grade power regulators and develop robust operating systems capable of automatic reboot and restoration. This project observed significant fluctuations in the DC voltage on test buses, and power from the direct battery circuit could be unexpectedly interrupted during data and software uploads and downloads.