Autonomous Intersection Management over Continuous Space: A Microscopic and Precise Solution via Computational Optimal Control

Lane-Free Crossing of CAVs through Intersections as a Minimum-Time Optimal Control Problem

“…The first benchmark is the conflict-point-reservation approach in [7], where each CAV calculates its own trajectory by jointly minimising the travelling time and energy consumption and stores its reservation time for each conflict point in a coordinator. The second benchmark is the lane-free method proposed by [5]. This work calculates the control inputs for a fixed final time, however, there is no reservation of conflict points and the CAVs can freely use all the space of the junction, as long as there is no collision.…”

“…The results in Table 2 show that the proposed strategy improves the crossing time of CAVs by, respectively, up to 65% (and an average improvement of 52% for different number of CAVs), and 54% as compared to the reservation-based approach in [7] and lane-free method in [5]. Moreover, the average speed of CAVs by the proposed algorithm is larger (for larger number of crossing CAVs) and the standard deviation of speeds are smaller than the ones by the reservation-based method.…”

“…The reservationbased methods restrain the motion of CAVs to orderly pass the critical points which results in a higher crossing time and lower average speed values than the proposed strategy. Furthermore, the lane-free method in [5] fixes the crossing time to make the calculation tractable, however, it lowers average speeds and therefore increases crossing time. Fig.…”

“…The lane-free approach allows CAVs to utilise the whole space of the intersection which can significantly improve the throughput of the intersection. Prior studies formulated the lane-free crossing of CAVs as an OCP [5,6]. To avoid collisions, the Euclidean distance between any pair of CAVs are constrained to be greater than a safe margin.…”

“…The approach is interesting, however, solving the offline non-convex problem for all possible crossing scenarios takes an impractical time. Alternatively, [5] fixed the crossing time to a constant value and converted the minimum-time optimal control problem to a feasibility problem to solve online. However, finding a feasible (collision-free) solution to the problem of the intersection crossing does not fully exploit the CAVs' advantages to minimise the crossing time.…”