Seminar at Google: Andreas Malikopoulos
Dr. Andreas Malikopoulos delivered a talk entitled “Optimal Path Planning and Coordination for Connected and Automated Vehicles” at Google on January 27, 2020.
Connected and automated vehicles (CAVs) provide the most intriguing opportunity for enabling users to better monitor transportation network conditions and make better operating decisions to improve safety and reduce pollution, energy consumption, and travel delays. CAVs are typical cyber-physical systems where the cyber component (e.g., data and shared information through vehicle-to-vehicle and vehicle-to-infrastructure communication) can aim at optimally controlling the physical entities (e.g., CAVs, non-CAVs). The cyber-physical nature of such systems is associated with significant control challenges and gives rise to a new level of complexity in modeling and control. As we move to increasingly complex emerging mobility systems, new control approaches are needed to optimize the impact on system behavior of the interplay between vehicles at different traffic scenarios.
In this talk, Dr. Malikopoulos presented a decentralized control framework for coordination of CAVs in different traffic scenarios, e.g., merging at roadways and roundabouts, crossing unsignalized intersections, cruising in congested traffic, passing through speed reduction zones, and lane-merging or passing maneuvers. The framework included: (1) an upper-level optimization that yielded for each CAV its optimal path, including the time and lane, to pass through a given traffic scenario by alleviating congestion; and (2) a low-level optimization that yielded for each CAV its optimal control input (acceleration/deceleration) to achieve the optimal path and time derived in the upper-level. Dr. Malikopoulos presented a geometric duality framework using hyperplanes to prove strong duality of the upper-level optimization problem and the condition under which the optimal solution always existed. He showed that strong duality implied that the optimal path for each CAV did not activate any of the state, control, and safety constraints of the low level optimization, thus allowing for online implementation.