POMDP-Based Trajectory Planning for On-Ramp Highway Merging

TL;DR

The paper tackles trajectory planning for automated vehicle on-ramp merging under uncertainty by formulating the problem as a Partially Observable Markov Decision Process (POMDP) and solving it online with the Adaptive Belief Tree (ABT) algorithm. It introduces topology discretization to simplify the highway environment, a Frenet-frame based ego model that enables lateral lane changes, and IDM-based dynamics for other vehicles, all within a probabilistic observation framework. The approach is validated on three real-traffic scenarios drawn from the ExiD German highway dataset, showing safe, collision-free, and efficient merging trajectories across varying traffic conditions. This work demonstrates the versatility and potential real-time applicability of POMDP-based planning for complex automated driving tasks, while outlining directions to enhance real-time performance and robustness in denser traffic.

Abstract

This paper addresses the trajectory planning problem for automated vehicle on-ramp highway merging. To tackle this challenge, we extend our previous work on trajectory planning at unsignalized intersections using Partially Observable Markov Decision Processes (POMDPs). The method utilizes the Adaptive Belief Tree (ABT) algorithm, an approximate sampling-based approach to solve POMDPs efficiently. We outline the POMDP formulation process, beginning with discretizing the highway topology to reduce problem complexity. Additionally, we describe the dynamics and measurement models used to predict future states and establish the relationship between available noisy measurements and predictions. Building on our previous work, the dynamics model is expanded to account for lateral movements necessary for lane changes during the merging process. We also define the reward function, which serves as the primary mechanism for specifying the desired behavior of the automated vehicle, combining multiple goals such as avoiding collisions or maintaining appropriate velocity. Our simulation results, conducted on three scenarios based on real-life traffic data from German highways, demonstrate the method's ability to generate safe, collision-free, and efficient merging trajectories. This work shows the versatility of this POMDP-based approach in tackling various automated driving problems.

Figures (9)

• Figure 1: Picture of a highway exiDdataset (up) and discrete set of corresponding lanes (down).
• Figure 2: A diagram of a lane change illustrating certain parameters and states.
• Figure 3: Likelihood functions of features $f_1$(left), and $f_2$(right).
• Figure 4: Time-lapse of the 30 simulations for Scenario 1. Simulated ego vehicles are represented as semi-transparent blue rectangles, non-ego vehicles are shown as red rectangles. The cyan rectangle shows how a real vehicle we replaced with our ego vehicle performed the merge.
• Figure 5: Velocity and input values of 30 simulations for Scenario 1. The opacity of each line segment represents the total number of simulations sharing the values.