Trajectory Planning in a Crossroads for a Fleet of Driverless Vehicles

Abstract

In a transportation system with fully autonomous vehicles (or cybercars), a lot of tasks have to be executed to answer the demand of a customer to travel between two points. The framework of this paper is an approach consisting in decomposing the planning into three levels, each of which using only a relevant subset of the information, thus reducing the complexity: the macroscopic level, e.g. a city or a region; the mesoscopic level, e.g. a city quarter; the microscopic level, i.e. the surroundings of the cybercar.At the macroscopic level, a path is computed. By path, we mean a succession of edges (road segments) in the graph description of the road network.At the mesoscopic level, paths are transmitted by the upper level and turned into trajectories. A trajectory is a precise space-time curve that the cybercar has to follow. Typical precisions are 10 cm and 1/10 s. The goal of this level is to produce trajectories for all the cybercars circulating in a given area. These trajectories have to be safe, most notably collision-free, but also efficient, i.e. deadlock-free.At the microscopic level, the cybercar's control moves along the trajectory and ensures that none of these collisions which couldn't have been foreseen at the higher levels occur.It should also be stressed that this framework is not top-down only. Each lower level can raise warnings to higher levels ; e.g. stopping before an unforeseen children raises a warning from the microscopic level, where it is detected, to the mesoscopic level, where new trajectories are calculated.In this paper, we focus on the mesoscopic level. More precisely, we present the study on a simple crossroads (X junction) of the concepts developed to deal in a general context. Several trajectory-generating algorithms are described and compared in terms of efficiency. One key element of these algorithms is the respect of the controllability constraints of the vehicles, for the actual vehicles to be able to follow the generated trajectories. A second key element is the management of the shared resources, i.e. the places on the junction where trajectories intersect.

State-Driven Priority Scheduling Mechanisms for Driverless Vehicles Approaching Intersections

“…Ismail et al [18] proposed an iCACC tool to manage intersection and vehicle trajectory adaptively by using CACC, thus optimizing vehicle speed profiles to minimize delay and prevent crashes. Olivier et al [19] presented centralized supervised reservation systems for crossroads with a proposed algorithm to determine the trajectories and speeds of all driverless vehicles approaching intersections. Li et al [20] studied a cooperative driving mechanism at blind crossings with a proposed concept of safety driving patterns.…”

“…For example, corresponding to Fig. 7(b), the priority scheduling sequence without Renew messages of vehicles υ 18 to υ 25 is (υ 18 , υ 20 , υ 22 , υ 24 , υ 25 , υ 19 , υ 21 , υ 23 ), whereas when the Renew mechanism is activated, the sequence will be (υ 20 , υ 22 , υ 24 , υ 25 , υ 18 , υ 19 , υ 21 , υ 23 ) because the arriving sequence in Q + is updated dynamically along with the change in P T W s. From the experiment results, we make sure that although the full scheduling sequence has changed, the sequence of vehicles(H) remains the same, and the scheduling performance is guaranteed.…”

Cooperative driving at isolated intersections based on the optimal minimization of the maximum exit time

“…As the authors pointed out, the time complexity exponentially increases with the number of vehicle and lanes. Dresner and Stone (2004), Mehani and de La Fortelle (2007) or David et al (2012) introduced a simple control policy, First Come First Serve (FCFS), with low computational cost. Matteo and Sascha (2009) proposed a more complicated control policy, such as the longest-in-system.…”

“…In addition, the method of distributing the right-of-way is also very important to guarantee the safety at the conflict zone. Two approaches of negotiating the right-of-way are conducted: the first one is a decentralized approach (Grünewald et al, 2006), where vehicles communicate with each other to obtain the right-of-way; the second one is centralized, where the intersection manager receives the requests of vehicles and sends the right-of-way accordingly (Li and Wang, 2006;Mehani and de La Fortelle, 2007;Dresner and Stone, 2008). In the first approach, the main burden of caring for the safety falls on the performance of the communication system.…”

“…The most basic application of cooperative driving at intersections is to provide information to the driver about the surrounding driving environment. A more advanced application is to elaborate the driving behavior in order to direct autonomous vehicles through an intersection (Arora et al, 2012;Grünewald et al, 2006;Li and Wang, 2006;Mehani and de La Fortelle, 2007;Dresner and Stone, 774 J. Wu et al…”