Autonomous Intersection Management concepts

With the future promises of fully automated driving new opportunities arise for managing the traffic on intersections, this is called Autonomous Intersection Management (AIM). Where current signalled intersections are only able to communicate decisions to groups of traffic participants, AIM uses the possibility to do this to individuals. When observing classical intersections and AIM from a distance eventually all comes down to a space time reservation. A group reservation for classical intersections, an individual reservation for automated vehicles.

These individual reservations prove a method to greatly improve the effective capacity of an intersection [1]. They enable very tight safety margins and intersection crossings in a ‘woven’ manner. Anyone who has seen Minority Report has had a glimpse of what may be possible. Optimization methods will make it possible to approach the highest theoretical capacity of an intersection: 100% car coverage of the intersection at the highest speed possible.

Currently there are two major approaches for the interaction with CAV’s. The first is from Texas University focussing on a multi-agent solution and a (first come first serve) reservation system (still named the original AIM)[2],[3]. The second is from MIT focussing on slot reservations using a cloud based solution (Light Traffic) [4].

AIM is aimed at fully autonomous intersections. However, the number of autonomous vehicles is currently very low and is expected to increase slightly during the next years. The transition towards fully autonomous intersections will start the coming years and the general expectation is that the transition period will take at least a decade. There are studies providing solutions for the transition path as well. One of the examples is the Texas paper on Mixed Autonomous and Human-Operated Vehicles at Intersections [5].

Transition towards Automated Intersection Management

Automated Intersection Management is only possible in a fully automated environment. Each user of the intersection needs to be connected to the controller, and received instructions by the controller needs to be followed exactly. As long as not all vehicles are connected and automated and vulnerable road users are still crossing the intersection, Automated Intersection Management is not possible.

New technologies offer possibilities to realize future benefits of Automated Intersection Management already. Incorporating these new communication, sensor and processing technologies in traditional traffic light controllers, both the road users and road authorities can benefit of an increased traffic flow and reduced pollutions resulting in an increased liveability in cities.

In this paper we distinguish four phases in the transition towards Automated Intersection Management:

  1. Conventional traffic light control
  2. Connected traffic light control
  3. Hybrid Intersection Management
  4. Automated Intersection Management

Phases 1 consists of current well known technologies for traffic light control. The second phase describes the current state-of-the-art in the field of traffic light control. In this paper we introduce the new phase 3 required for the transition towards Automated Intersection Management (phase 4).

Conventional traffic light control

Conventional traffic light control is either fixed timed, vehicle actuated or adaptive [6]. Fixed timed controllers use a predetermined cycle in which directions are served in a fixed sequence with fixed green, amber and red times. Each cycle of the traffic light controller is the same, the traffic lights are not adjusted to the actual traffic state at the intersection. Fixed timed controllers are extremely predictive but are not responsive to the actual traffic demand at the intersection.

In response the fixed timed control, vehicle actuated traffic light controllers were introduced. Using loop detectors in the road surface, the green time of each direction is extended until a large enough gap between two successive vehicles is detected to change the light safely to amber. The durations of the green phases are adjusted to the actual traffic demand, however the predictability of the traffic light controller is very low.

Adaptive traffic light controllers, like TRANSYT, UTOPIA and SCOOT [7, 8, 9] use an internal traffic model to predict future traffic flows at intersections. The sequence in which directions are served is fixed, however the duration of the green phases are optimised based on the expected traffic flows. Vehicle actuated control may still be used for optimal switching between phases. Adaptive traffic light controllers are also responsive to the actual traffic demand and are more predictable compared to vehicle actuated controllers.

Connected traffic light control

Conventional traffic light controller use indirect communication channels with the road users. Road side sensors, e.g. induction loops, are used to detected vehicles approaching the intersection. Traffic signals are used to communicate the state of the intersection (green, amber, red) to the drivers. With the recent introduction of connected traffic lights direct communication with traffic becomes possible. Both ITS-G5 and cellular technologies make it possible to transmit data between the traffic light controller and nearby vehicles.

The communication between the traffic light controller and vehicles is two-way. While arriving, vehicles continuously broadcast their current position and speed to the traffic light controller using the CAM and DEMN message protocols [10]. This enables the traffic light controller to make reliable predictions of the expected arrival times of vehicles and to efficiently schedule its green phases [11].

The scheduled green phases are transmitted from the traffic light controller to all arriving vehicles using the SPaT/MAP protocol [12]. The Signal Phase and Timing (SPaT) message describes for each lane of the controlled intersection the expected state switches of the traffic light for the near future. Together with the MAP message, describing the topology of the intersection, the SPaT message is broadcasted to all approaching vehicles. Inside the vehicle the SPaT is translated into time-to-green and/or time-to-red information, informing the driver about the remaining waiting time or green time of the traffic light.

SPaT/MAP messages are often related to a third use case: green light optimal speed advice (GLOSA). The driver gets an individual speed advice enabling him to pass the intersection without stopping. Unfortunately a reliable speed advice only based on the SPaT-message is not possible, additional information of the surrounding traffic is required in order to calculate an appropriate and reliable speed advice.

Hybrid Intersection Management

Connected traffic light controllers receive information from individual vehicles, CAM and DEMN messages, and use this information for efficiently scheduling green phases. However the information send back to the approaching vehicles is collective and for all the vehicles the same. It contains for example the start- and end times of scheduled green phases. This information alone is not enough to calculate a reliable speed advice for approaching vehicles: also the amount of downstream vehicles is required to calculate a reliable speed advice.

With the introduction of Hybrid Intersection Management we propose a new message-format: the individual Space & Time message (iSPaT) The iSPaT message contains for each individual vehicle a unique space-time reservation in which it may cross the intersection. In the hybrid situation, these reservations are off course within a green phase of the related traffic light for safety reasons. The big advantage of the iSPaT is that arriving vehicles receive the time they are expected to arrive at the intersection, taken into account the number of downstream vehicles. This makes it possible to optimise the arrival of vehicles at the intersection, with a positive impact on the driver comfort, the air quality and the traffic flows.

The iSPaT messages are, comparable to the SPaT/MAP messages, publicly published and available to all road users at the intersection. This has two main advantages. Firstly, no peer-to-peer direct connection between the traffic light controller and an approaching vehicle needs to be established. Secondly, other vehicles can benefit from knowing the time-space reservations from other vehicles. Knowing what to expect at the intersection can further optimize the crossing of vehicles.

The iSPaT information can be used to inform the human driver in an approaching vehicle. The received information will be in-vehicle translated into a personal speed/lane advice for the driver. This advice is comparable to GLOSA with a higher reliability and extended with a lane-advice. For autonomous vehicles it is also possible to connect the incoming instructions directly to the motor management system. This way the autonomous vehicle will be ‘guided’ over the intersection in the most efficient way.

Automated Intersection Management

The transition from the hybrid intersection to the automated intersection is relatively small. When the penetration rate of autonomous vehicles increases it becomes possible to schedule autonomous-only phases in the hybrid intersection or to shift to a full autonomous controller: Automatic Intersection Management. In this case, all vehicles are guided over the intersection in a ‘woven’ manner. As previous studies showed, the effective capacity of intersections will be maximized [1].

In contrast to the hybrid intersection, the automated intersection does not schedule green phases anymore. All vehicles are served individually with tight safety margins. This is possible with the iSPaT-messages, which instruct vehicles about their individual time-space reservation. Connecting the iSPaT to the motor management system makes it possible for a vehicle to cross the intersection in its own timeslot.

Autonomous vehicles are expected to rely heavily on their own sensors. If a vehicle detects a (moving) obstacle within its safety range, it will automatically take actions to increase the safety distance. Possible actions are slowing down or adjusting its driving direction, both resulting in releasing the reserved timeslot. This may cause a chain of reactions of other vehicles, ultimately leading to a chaotic situation at the controlled intersection.

Autonomous vehicles continuously broadcast a Cooperative Awareness Message (CAM) to the controlled intersection. By sharing the CAM-messages together with the iSPaT-messages between all vehicles, it becomes possible for a single vehicle to check if nearby vehicles are following their reserved timeslot. Knowing that the timeslots are not conflicting, it can safely follow its own timeslot as long as the other vehicles are within their reserved timeslot. This concept makes ‘woven’ traffic at intersections possible with respect for road safety.

By distinguishing autonomous and non-autonomous phases by the automated controller, vulnerable road users are still able to pass the intersection in a safe manner. Traditional traffic lights indicate when pedestrians and cyclists may cross the intersection, which can be considered as a combined times-space reservation for these users.