Low-speed shared spaces such as parking lots, campuses, plazas, and pick-up zones are where automated mobility comes closest to people. There are no lanes and few rules, visibility is poor, and vehicles, pedestrians, shopping carts, and micro-mobility systems move through the same space at close range. Roughly one in five collisions in the United States are parking related, and nearly half of drivers in cities rate parking as the most stressful part of a trip. Today’s automated parking features handle the last few meters but still depend on a supervising driver.

Our research aims at highly automated systems that can operate in these spaces without supervision. The initial target is a vehicle that enters a lot, finds a spot, negotiates with everyone around it, and parks while its driver walks away. The broader goal is vehicles and micro-mobility systems that move safely and considerately wherever people and machines share the ground. The work is built on Honda’s Cooperative Intelligence (CI), AI that understands the intentions of the people around it and acts cooperatively with them. Our mission is to save people time and give them a stress-free mobility experience. Our research currently spans the following areas.

Strategy and exploration under uncertainty

Where should the system go, and when should it commit, when the environment is only partially observed and constantly changing? In a parking lot this means deciding, in real time, whether to park in the spot ahead, wait for one that is about to open, or keep exploring.

Intent prediction and negotiation

We want a system that can interpret the intentions of other agents and act in a cooperative manner. In a parking lot this means inferring where other drivers intend to go from their motion and choosing spots and maneuvers that are intention-informed and socially balanced.

Interaction-aware navigation

Sharing space with pedestrians, carts, and other vehicles calls for yielding, passing, and waiting the way an attentive human would, rather than treating everything as an obstacle to avoid. In a parking lot this means moving through a crowded aisle while cars pull out, pedestrians cross, and carts drift.

Reasoning under occlusion

Safety is one of our core pillars, and much of a shared space is hidden behind parked vehicles, corners, and structures. We are therefore exploring planning methods that reason about what could emerge from unseen regions, including reachability-based approaches with formal safety guarantees, without making the system overly cautious.

Precise maneuvering in tight spaces

The last few meters demand centimeter-level precision, and other agents do not stop moving while the vehicle parks. We are investigating smooth and flexible parking maneuvers that handle interaction during the maneuver and work for every type of spot found in real lots: perpendicular, angled, parallel, in dead-end aisles and in cluttered or non-standard spaces.

Learning-based planning

Hand-designed rules do not cover the variety of real shared spaces. We are exploring learned policies that generalize across geometries and remain safe outside their training experience, along with learned models that predict intent and make classical planners faster and more human-like.

We evaluate these ideas in simulation and on a range of four-wheeled vehicle and micro-mobility platforms, and we improve them using real-world data from our own tests. Together they aim at mobility systems that understand and negotiate with people in the unstructured spaces.