Dataset: Speed Trajectory Data from Adaptive Eco-driving Applications

The eco-approach and departure (EAD) application for signalized intersections has been proved to be environmentally efficient in a Connected and Automated Vehicles (CAVs) system. In the real-world traffic, the traffic-related information received from sensing or communication devices is highly uncertain due to the limited sensing range and varying driving behaviors of other vehicles. This uncertainty increases the difficulty to predict the actual queue length of the downstream intersection. It further brings great challenge to derive an energy efficient speed profile for vehicles to follow. This research proposes an adaptive strategy for connected eco-driving towards a signalized intersection under real world conditions including uncertain traffic condition. A graph-based model is created with nodes representing dynamic states of the host vehicle (distance to intersection and current speed) and indicator of queue status and directed edges with weight representing expected energy consumption between two connected states. Then a dynamic programing approach is applied to identify the optimal speed for each vehicle-queue-signal state iteratively from downstream to the upstream. The uncertainty can be addressed by formulating stochastic models when describing the transition of queue-signal state. For uncertain traffic conditions, numerical simulation results show an average energy saving of 9%. It also indicates that energy consumption of a vehicle equipped with adaptive EAD strategy and a 100m-range sensor is equivalent to a vehicle with conventional EAD strategy and a 190m-range sensor. To some extent, the proposed strategy could double the effective detection range in eco-driving.


The trajectory data was collected from numerical simulation using three types of methods including the proposed method in this research, the ideal method and other baseline EAD methods.  The proposed method corresponds to the adaptive strategy for connected eco-driving with known historical queue distribution. The ideal trajectory for absolute minimum energy consumption can be derived when the actual queue length is known (i.e. perfect information) at the beginning of the simulation. This strategy can only be achieved if all vehicles are connected to share their positions to the study vehicle.  Besides the ideal method, couple of baseline EAD methods (Baselinek) are setup for comparison: Assuming the queue length to be Qk, the vehicle first follows the ideal trajectory of the assumed Qk length, then change to the corresponding strategy after detecting the real queue length. These baselines are the methods given the same information as the proposed method except the historical queue distribution is missing. Note that if k is 0, Baseline0 corresponds to the scenario when the vehicle follows the existing EAD strategy with no-queue assumption until the sensor detects preceding traffic.