Safety is the number one issue in the deployment of any vehicle technology. This project addresses this important topic of safety along an evolutionary deployment path where vehicles will eventually become autonomous and connected with each other and infrastructure. Along this path vehicles with different levels of automation and connectivity will have to co exist and safety needs to be well understood and impact of safety measures on traffic flow need to be analyzed. In today's driving environment drivers perform vehicle maneuvers competing for space and time in order to improve their travel time. This practice often puts them in a risky situation for a short period of time. During that short time period a neighboring vehicle maneuver or lack of it may lead into collision. The severity of such collision depends very much on the scenario and respective vehicle characteristics. Autonomous vehicles cannot be designed to take such obvious risks due to liability issues. The question is how this risk is assessed, what are the dilemmas faced, and how can they be overcome? What would be the impact on traffic as the penetration of autonomous vehicles on traffic flow given their conservative safety gaps? We are considering several different scenarios that involve both connected and non-connected autonomous vehicles. In a non-connected environment, vehicles whether manually driven or autonomous rely on their own assessment to make a particular maneuver. Typical maneuvers may include overtaking, changing lanes, merging, stopping. In a connected environment vehicle can negotiate such maneuvers and safety spacings which may work well in a fully connected environment but it will face challenges in a mixed environment of connected and no connected vehicles.

We propose a collision risk index (CPI) that depends on how much the vehicle violates the minimum safety spacing for any given situation and for how long. The minimum safety spacing is calculated as the minimum spacing required for the vehicle to stop without collision if a vehicle in the same lane executes emergency braking to a full stop. In addition to CPI other measures of risk assessment will also be considered and compared. The focus on intervehicle spacing in vehicle following, merging and changing lanes has a direct impact on traffic flow and therefore the trade-offs between safety and traffic flow capacity clearly appear.

Once the methods for assessing safety are established, we plan to evaluate how humans drive using publicly available data and microscopic traffic simulations and quantify the level of risk using CPI and other measures. We then plan to introduce autonomous vehicles that are designed not to take any risks, namely not violate the collision risk index under any conditions and evaluate the impact on traffic flow. For example, if an autonomous vehicle does not find an empty spacing that is greater than the required safety spacing during merging or changing lanes in a congested environment will it create a queue and slow down traffic? Or will it take a risk and forced itself into the traffic stream like drivers do? Or will it wait for reasonable time in an effort to minimize the risk of collision? What will be the impact on queues and traffic flow in each case? How is the traffic will be affected as the percentage of autonomous vehicles increases in a mixed traffic with human driven vehicles? These issues will be raised and analyzed in this project. One possible solution is for the autonomous vehicle to look first for a spacing that leads to zero CPI and if over a certain period of time it cannot find one then it can raise the acceptable CPI from zero to that of a fraction of the CPI for human drivers. Such decisions will bring up the trade-off between safety and mobility. An approach that will solve some of these problems is connectivity that will allow autonomous vehicles to communicate with each other. Under such an environment safety spacings can be negotiated in a way that CPI will be reduced considerably as the penetration of connected autonomous vehicles is increased. We plan to use results from past projects regarding safety spacing and communication protocols where we assume all vehicles are connected. In this project we assume mixed traffic where connected vehicles operate with non-connected vehicles. This is a far more complicated situation yet more practical as connected vehicles will have to co-exist with non-connected ones till full penetration takes place.