Global Multi-Vehicle Trajectory Planning Challenge

Intelligent vehicles and wheeled robots are revolutionizing transportation and logistics, offering promising advancements in efficiency and safety. A critical function of these autonomous systems is trajectory planning—the process of generating a safe, comfortable, and optimized spatio-temporal path for a vehicle or robot to follow. While traditional autonomous driving focuses on individual vehicles, cooperative trajectory planning involves simultaneously generating trajectories for an entire fleet of intelligent vehicles or robots, introducing a higher level of complexity.

The aim of this competition is to test the capabilities of state-of-the-art cooperative trajectory planning methods and foster innovation among new researchers in this research field. Cooperative trajectory planning presents unique challenges, particularly when the number of vehicles in a group increases. As the swarm grows, the complexity of collision-avoidance constraints rises geometrically, making it difficult to scale solutions effectively. Additionally, when multiple vehicles navigate around each other, numerous topological scenarios, known as homotopy classes, would arise. Selecting an inefficient homotopy class leads to local optima, resulting in suboptimal and low-efficiency solutions. The intricacies of cooperative trajectory planning, especially in cluttered environments, present an open research problem that demands further exploration. This competition aims to push the boundaries of what is possible and inspire researchers to tackle these challenges head-on, advancing the state of the art in cooperative multi-vehicle systems.

The primary purpose of this competition is to provide researchers with an accessible opportunity to compare their cooperative trajectory planners against others using a set of benchmark cases. We aim to foster broad participation within the trajectory planning community, encouraging researchers to develop and share innovative solutions to challenging benchmark problems.

To achieve this goal, we focus exclusively on cooperative trajectory planning, without involving the complexities of perception, localization, communication, or tracking control. By narrowing the scope, we make this competition more approachable for researchers dedicated solely to planning, regardless of expertise in other system components. We want to ensure that participating in this competition is straightforward and inclusive. Therefore, participants are not required to submit the source code of their planners; instead, they only need to submit their solution data. This approach removes any dependencies on specific coding platforms or languages, promoting accessibility and fairness. In terms of solution submissions, our goal is to capture the spatio-temporal motion behavior of the vehicle swarm using minimal, non-redundant data. This simplification not only streamlines the submission process but also enhances fairness, allowing the focus to remain on the quality of the planning solutions themselves rather than on implementation details.