Decentralized optimal control of Connected Automated Vehicles at signal-free intersections including comfort-constrained turns and safety guarantees

Abstract

We extend earlier work for optimally controlling Connected Automated Vehicles (CAVs) crossing a signal-free intersection by including all possible turns taken so as to optimize a passenger comfort metric along with energy and travel time minimization. We show that it is possible to achieve this goal in a decentralized manner with each CAV solving an optimal control problem, and derive explicit solutions that guarantee collision avoidance and safe distance constraints within a control zone. We investigate the associated tradeoffs between minimizing energy and vehicle travel time, as well as the passenger comfort metric and include extensive simulations to illustrate this framework.

Optimal Path Planning for Connected and Automated Vehicles at Urban Intersections

“…The solution allows the CAVs to coordinate and pass through these traffic scenarios without creating congestion and under the hard safety constraint of collision avoidance. Similar control approaches have considered passengers' comfort in addition to alleviating congestion [35], [36].…”

“…The proposed framework advances the state of the art in the following ways. First, in contrast to other efforts where either the safety constraint was not considered [27], [29]- [32], [35], or considered using a constant safety distance [36], in our framework, the low-level analytical solution considers the safety distance between the CAVs to be a function of speed leading to a complicated, yet very interesting, analysis. Moreover, we augment the double integrator model representing a CAV with an additional state corresponding to the distance from its preceding CAV, thus we are able to address the lateral colision constraint in the low-level optimization.…”

“…Moreover, we augment the double integrator model representing a CAV with an additional state corresponding to the distance from its preceding CAV, thus we are able to address the lateral colision constraint in the low-level optimization. Second, in several efforts reported in the literature to date, the upper-level optimization either (a) was implemented with centralized approaches [14], [15], [17]- [19]; or (b) was considered given [35], [38]; or (c) was implemented using a strict first-in-first-out queueing structure [27], [29], [32], [36]. In our proposed framework, the upper-level optimization yields, in a decentralized fashion, the optimal time for each CAV to pass a given traffic scenario along with the appropriate lane that needs to occupy.…”

Joint Time and Energy-Optimal Control of Connected Automated Vehicles at Signal-Free Intersections with Speed-Dependent Safety Guarantees

“…First, instead of solving a throughput maximization problem followed by an energy minimization problem for each CAV, here we formulate a problem in which each CAV seeks to jointly minimize both its travel time through a specified Control Zone (CZ) and Merging Zone (MZ) and its energy consumption. This allows us to readily quantify the tradeoff between these two criteria (see also [20] where left/right turns are included along with a passenger comfort metric). Second, unlike [19], [20] where we first resolve possible collisions in the MZ and then apply optimal control over the CZ, here we relax the constant speed assumption inside the MZ and handle lateral collision avoidance as additional state constraints; this provides flexibility in controlling CAVs within the MZ.…”

“…This allows us to readily quantify the tradeoff between these two criteria (see also [20] where left/right turns are included along with a passenger comfort metric). Second, unlike [19], [20] where we first resolve possible collisions in the MZ and then apply optimal control over the CZ, here we relax the constant speed assumption inside the MZ and handle lateral collision avoidance as additional state constraints; this provides flexibility in controlling CAVs within the MZ. Third, unlike [19], [20] where we limit ourselves to a distance-dependent rear-end safety constraint, here we include a speed-dependent rearend safety constraint, which better captures the relationship between two consecutive vehicles traveling on the same road.…”