Max Pressure Control Research Articles

Max-pressure traffic controller based on travel times: An experimental analysis

The traffic control of an arbitrary network of signalized intersections is considered. This work presents a new version of the recently proposed max-pressure controller, also known as back-pressure. The most remarkable features of the max-pressure algorithm for traffic signal control are: scalability, stability, and distribution. The modified version presented in this paper improves the practical applicability of the max-pressure controller by considering as input travel times instead of queue lengths. The two main practical advantages of this new version are: (i) travel times are easier to estimate than queue lengths, and (ii) max-pressure controller based on travel times has an inherent capacity-aware property, i.e., it takes into account the finite capacity of each link. Travel time tends to diverge when the queue length is close to its capacity. It should be noted that previous max-pressure algorithms rely exclusively on queue length measurements, which may be difficult to accomplish in practice. Moreover, these previous algorithms generally assume queues with unbounded capacity. This may be problematic because a model with unbounded capacity links is not able to reproduce spillbacks, which are one of the most critical phenomena that a traffic signal controller should avoid. After presenting the new version of the max-pressure controller, it is analyzed and compared with existing control policies in a microscopic traffic simulator. Moreover, results of a real implementation of the developed algorithm to a signalized intersection, located at an urban arterial in Jerusalem, are shown and analyzed. To the best of the authors’ knowledge, this experiment is the first real implementation of a max-pressure controller at a signalized intersection.

Max-pressure control of dynamic lane reversal and autonomous intersection management

Autonomous intersection management (AIM) (which coordinates intersection movements to avoid signal phases) and dynamic lane reversal (DLR) (which frequently changes lane directions in response to time-varying demand) have previously been proposed for connected autonomous vehicles. A major open question for both is finding the optimal control policy. This paper develops a decentralized max-pressure policy that controls both AIM and DLR based on queue lengths on adjacent links. Using a stochastic queueing model, we prove that the max-pressure policy is also throughput-optimal; any demand that can be stabilized (queue lengths remain bounded) will be stabilized by the max-pressure policy. We show numerically that DLR significantly increases the stability region, particularly for asymmetric demand. Since the stochastic queueing model excludes some realistic aspects of traffic flow, we adapt the max-pressure control for simulation-based dynamic traffic assignment. Results on a city network show significant improvements from max-pressure AIM with and without DLR.

Design and Analysis of Delay-Tolerant Intelligent Intersection Management

The rapid development of vehicular network and autonomous driving technologies provides opportunities to significantly improve transportation safety and efficiency. One promising application is centralized intelligent intersection management, where an intersection manager accepts requests from approaching vehicles (via vehicle-to-infrastructure communication messages) and schedules the order for those vehicles to safely crossing the intersection. However, communication delays and packet losses may occur due to the unreliable nature of wireless communication or malicious security attacks (e.g., jamming and flooding), and could cause deadlocks and unsafe situations. In our previous work, we considered these issues and proposed a delay-tolerant intersection management protocol for intersections with a single lane in each direction. In this work, we address key challenges in efficiency and deadlock when there are multiple lanes from each direction, and propose a delay-tolerant protocol for general multi-lane intersection management. We prove that this protocol is deadlock free, safe, and satisfies the liveness property. Furthermore, we extend the traffic simulation suite SUMO with communication modules, implement our protocol in the extended simulator, and quantitatively analyze its performance with the consideration of communication delays. Finally, we also model systems that use smart traffic lights with various back-pressure scheduling methods in SUMO, including the basic back-pressure control, the capacity-aware back-pressure control, and the adaptive max-pressure control. We then compare our delay-tolerant intelligent intersection protocol with smart traffic lights that use the three back-pressure scheduling methods, in the case of a network of interconnected intersections. Simulation results demonstrate that our approach significant outperforms the smart traffic lights under normal operation (i.e., when the communication delay is not too large).

A Simulation Study on Max Pressure Control of Signalized Intersections

In this paper, a decentralized traffic signal control strategy named max pressure control is reviewed and examined. This control strategy aims at optimizing overall network throughputs, but applies a distributed approach that only requires local information to generate timing plans for each intersection. A Vissim simulation study is conducted to compare existing max pressure schemes. The results show that a recently proposed cyclic-based approach performs more poorly than the original non-cyclic approach. Further, to address two issues that hinder existing schemes: frequent changes of phase and queue spillover/blockage, two modifications are suggested. The simulation reveals that network performance can be improved after modifications.

Stability and implementation of a cycle-based max pressure controller for signalized traffic networks

Intelligent use of network capacity via responsive signal control will become increasingly essential as congestion increases on urban roadways. Existing adaptive control systems require lengthy location-specific tuning procedures or expensive central communications infrastructure. Previous theoretical work proposed the application of a max pressure controller to maximize network throughput in a distributed manner with minimal calibration. Yet this algorithm as originally formulated has unpractical hardware and safety constraints. We fundamentally alter the formulation of the max pressure controller to a setting where the actuation can only update once per multiple steps of the modeled dynamics. This is motivated by the case of a traffic signal that can only update green splits based on observed link-counts once per time of 60-120 seconds. Furthermore, we extend the domain of allowable actuations from a single signal phase to any convex combination of available signal phases to model intra-cycle signal changes dictated by pre-selected cycle green splits. We show that this extended max pressure controller will stabilize a vertical queueing network given restrictions on admissible demand flows that are slightly stronger than those suggested in the original formulation of max pressure. We ultimately apply our cycle-based extension of max pressure to a simulation of an existing arterial network and provide comparison to the control policy that is currently deployed at the modeled location.

Adaptive Max Pressure Control of Network of Signalized Intersections

Abstract: The control of a road network of signalised intersections is considered. Variants of a decentralised universal feedback control policy (UFC) are evaluated and compared with an open loop traffic regulator periodically distributing the same actuation duration to each stage (set of permitted movements). At the beginning of each cycle and for a particular intersection, the Max-Pressure (MP) controller selects a stage to actuate as a function of the adjacent queues to that intersection. Vehicle movements are modelled as a store and forward (SF) queuing network. A controller stabilises a demand if the mean queue length remains bounded. The current MP version requires no knowledge of the demand while estimates turn probabilities. A section near LA comprised of 16 intersections and 76 links is employed. The network performance is appraised when regular traffic conditions, congestions or unforeseen feasible demand fluctuations take place. The conducted study assesses the potential benefits of MP and confirms its adaptability. Each control strategy is evaluated in terms of queue lengths, delays, travel times etc. by discrete event simulations confirming the provided theoretical guaranteed (queue) bounds of MP.

Performance Analysis of Centralized and Distributed Systems for Urban Traffic Control

In this paper, the performance of systems for urban traffic control is analyzed with consideration of real-time variations and uncertainties in traffic flow. The control systems range from centralized to semicentralized to distributed. The centralized system is represented by the global brute force and genetic optimization algorithms; the semicentralized system is represented by a linear quadratic regulator formulation; and the recently proposed max-pressure controller represents the distributed system. The control systems were applied to networks with different topology and demand levels, and the network performances were evaluated through a Monte Carlo stochastic simulation framework that incorporates various sources of variations that mimic the traffic variability observed in the real world. As expected, the brute force solutions outperformed the others, although the differences between the performances of different systems were not as significant as expected. It was found that the primary benefit gained from the brute force solution came from the setting of signal offsets. This result suggests that calculation of offsets plays an important role in bridging the gap between the computationally demanding centralized systems and the parsimonious distributed systems.

Max pressure control of a network of signalized intersections

The control of a network of signalized intersections is considered. Vehicles arrive in iid (independent, identically distributed) streams at entry links, independently make turns at intersections with fixed probabilities or turn ratios, and leave the network upon reaching an exit link. There is a separate queue for each turn movement at each intersection. These are point queues with no limit on storage capacity. At each time the control selects a ‘stage’, which actuates a set of simultaneous vehicle movements at given iid saturation flow rates. Network evolution is modeled as a controlled store-and-forward (SF) queuing network. The control can be a function of the state, which is the vector of all the queue lengths. A set of demands is said to be feasible if there is a control that stabilizes the queues, that is the time-average of every mean queue length is bounded. The set of feasible demands D is a convex set defined by a collection of linear inequalities involving only the mean values of the demands, turn ratios and saturation rates. If the demands are in the interior Do of D, there is a fixed-time control that stabilizes the queues. The max pressure (MP) control is introduced. At each intersection, MP selects a stage that depends only on the queues adjacent to the intersection. The MP control does not require knowledge of the mean demands. MP stabilizes the network if the demand is in Do. Thus MP maximizes network throughput. MP does require knowledge of mean turn ratios and saturation rates, but an adaptive version of MP will have the same performance, if turn movements and saturation rates can be measured. The advantage of MP over other SF network control formulations is that it (1) only requires local information at each intersection and (2) provably maximizes throughput. Examples show that other local controllers, including priority service and fully actuated control, may not be stabilizing. Several modifications of MP are offered including one that guarantees minimum green for each approach and another that considers weighted queues; also discussed is the effect of finite storage capacity.