Project spotlight
FleetForge
by FleetRL
Train, evaluate, and deploy intelligence for orchestrating autonomous fleets.
Demo Video
About This Project
FleetForge is built for the next bottleneck in autonomy: fleet-level decision making. In the next 5 years, as robotaxis, delivery robots, drones, and other autonomous systems move from demos to deployment, the challenge is no longer just whether one vehicle can move effectively. The harder question is how to decide where each vehicle should be, when it should move or wait, and how operators can maximize ROI under demand spikes, traffic, disruptions, and limited supply.
We built FleetForge, a HUD-compatible RL and evaluation environment for AI fleet orchestrators. The agent operates inside a simulated mobility network where it observes demand, traffic, vehicles, requests, and business metrics. It then makes high-level fleet decisions like matching demand, repositioning idle supply, and adapting to live events before touching real-world fleets.
FleetForge is also extensible by design; operators can add new variables, constraints, and objectives (e.g. battery state for robotaxis, payload limits for drones, or aisle congestion for warehouse robots) as long as those signals can be observed + verified.