The behavior of an autonomous vehicle can be impacted by various internal factors like onboard system failure, sensor failure, etc. or by external factors like risky maneuvers by immediate neighbors threatening a collision, sudden change in road conditions, etc. This can result in a failure of the coordination maneuver like multi-vehicle intersection clearance. In such situations when conditions dynamically change and the nominal operational condition is violated by internal or external influences, an autonomous vehicle must have the capability to reach the minimal risk condition. Bringing the vehicle to a halt is one of the ways to achieve minimal risk condition.
This thesis introduces a safe stop algorithm which generates controls for multiple autonomous vehicles considering the presence of legacy manually driven vehicles on the road. A Model Predictive Control based algorithm is proposed which is robust to errors in communication, localization, control implementation, and model mismatch. Collisions avoided and discomfort faced by the driver are two evaluation parameters. Simulations show that the robust controller under the influence of errors can perform as well as the non-robust controller in the absence of these errors.