This coordination paradigm balances local urgency against public occupancy, reaching near-Oracle performance at minimal implementation cost is validated in simulation on real OpenStreetMap networks of twenty cities.
Abstract
In electric delivery fleets, mid-shift charging is non-trivial: each vehicle must decide when, where and how much to charge to finish on time with battery above a safety floor. The choices are coupled: queues build where too many vehicles pick the same station. Prior work resolves this coupling with central dispatching, precomputed schedules or reservations, machinery that charging infrastructure rarely supports. Instead, we use a family of learning agents under purely local control: every vehicle runs the same policy, deciding alone from its time budgets and broadcast station occupancies, leading to emergent coordination without central control or messaging. We validate this paradigm in simulation on real OpenStreetMap networks of twenty cities, each with a frozen scenario calibrated by an omniscient Oracle (99.5% of shifts completed on time), whereas a naive greedy rule (nearest station on low battery) completes just 73%. Agents trained with neuroevolution (NEAT) and policy gradients (PPO) on four cities and deployed zero-shot across all twenty, sixteen never seen in training, complete 96.8% and 98.6% of shifts, with the policy-gradient controllers proving more robust when demand or vehicle characteristics drift beyond the trained regime. In contrast, tuned threshold heuristics that read vehicle urgency alone fall short in contended cities (~80%). Through training, these learning agents rediscover partial charging and short opportunistic sessions, and route around busy stations, cutting per-session queue waits from about 45 minutes to under 2. In summary, this coordination paradigm balances local urgency against public occupancy, reaching near-Oracle performance at minimal implementation cost.
Dynamic routing is operationally fragile when autonomous electric vehicles differ in capacity and energy use, traffic changes mid-arc, and charging delays redispatch. We formulate a bi-objective problem minimizing operating cost and customer response without time windows. The model combines FIFO-consistent piecewise tr...
Kashif Bashir, R. Rafiq, M. A. Shah et al.· Scientific Journal of Engine...· 0 citations
Adopting battery electric vehicles (EVs) for vehicle fleets requires scheduling charging alongside day-to-day operations. This problem is complicated by complex utility cost structures, limited battery capacity, and competition for shared charging resources. Existing methods that simultaneously schedule routes and char...
Justin Whitaker, G. Droge, Mario Harper· Future Transportation· 0 citations
In this article, we consider an operator of an autonomous electric ride-hailing fleet who must decide, for every vehicle and period, whether to remain in place, travel empty to a rider, carry a rider, or charge at a station. The proposed mixed-integer programming model combines five operational features: deadhead trave...
Hamza Adeinat, Iman Adeinat· Applied System Innovation· 0 citations
Electric-vehicle fleets plan from digital-twin traffic, energy, charging, and station-availability state before dispatch. Communication loss, backend delay, and sensor noise make that state stale or incomplete. We study whether a conservative planning state, rather than update frequency alone, improves electric vehicle...
Abdelkader Kaddour, Lamri Sayad· International Journal of Aut...· 0 citations
Short-term congestion at public electric-vehicle charging stations depends on both near-future demand and how vehicles are distributed across the network. We forecast cumulative station-level charging initiations and use the forecasts in congestion-triggered station assignment. The analysis reuses 441077 publicly relea...
Oğuzkağan Alıç· Black Sea Journal of Enginee...· 0 citations
This paper treats multi-echelon resilience under electrified logistics as a closed-loop control problem rather than a scorecard. The plant is an inventory-and-fleet system whose material balances are coupled to usable pack energy, so a shipment is infeasible when every vehicle that could carry it violates an energy flo...
P. B· International Journal of Cre...· 0 citations
Related blog posts
Microsoft Research Blog· microsoft.comNov 16, 2026
AI coding agents are increasingly help developers fix bugs and build features largely on their own, but every interaction with their models consumes tokens. In each session, an agent sends instructions, tool descriptions, and session history to its model. Most efforts to reduce this cost focus on individual tasks. Yet developers return to the same […] The post TokenEconomizer: Learning from Experience to Make AI Coding Agents More Efficient appeared first on Microsoft Research.
MIT News · Artificial Intelligence· news.mit.eduNov 13, 2026