Fleet Orchestration Engineer - Neuraverse (human)

Your Mission & Challenges

• Multi-robot task allocation algorithms: auction-based (CBBA, market-based), MILP solvers for offline planning, Hungarian for simple bipartite matching, robot-capability matching, deadline-aware scheduling, priority queuing
• Mission decomposition: facility-level goal ("deliver 50 trays from kitchen to floor 3 by 18:00") per-robot behaviour-tree skill bundles + cross-robot dependency graph + handoff point definitions
• Fleet traffic management beyond static geofencing: dynamic corridor allocation, intersection arbitration, shared resource pools (charging docks, narrow passages, elevators, doors), preemption rules
• Cooperative-task primitives: synchronised place-pick handoffs (Robot A places, Robot B picks at coordinated time), coordinated bimanual across two physical robots holding the same object, leader/follower formations, multi-robot peg-in-hole
• Fleet-level performance management: throughput dashboards, robot utilisation balancing, SLA tracking, predictive task rerouting when one robot slows down
• VDA 5050 master mode: NEURA fleet orchestrator dispatches to NEURA + third-party robots in NEURA-managed humanoid/service facilities (hospitality, healthcare, mixed-fleet demos)
• VDA 5050 client interop testing: confirm NEURA robots integrate cleanly with third-party fleet management systems (Open-RMF, Otto, MiR, Symovo) for industrial deployments — protocol layer owned by Robot Connectivity Engineer
• Expose atomic skill primitives (MoveArmTo, Grasp, Release, Insert, etc.) as typed NodeGraph nodes — co-reviewed with platform Motion Planning Engineers
• Skill primitive API surface design: port schemas, error types, pre/postcondition contracts for third-party developers
• Pre-deployment behaviour-tree simulation in the Cognitive Twin: simulating full skill trees before on-robot deployment — Digital Twin & Multi-Sim Engineer provides the simulation infrastructure
• NodeGraph node templates for cooperative-task primitives (the orchestrator's execution vocabulary)

What we can look forward to

• Multi-robot systems background as the primary axis — has shipped task allocation, fleet coordination, or multi-robot orchestration in production. Looks like Open-RMF contributors, AMR fleet engineers (Locus Robotics, Otto Motors, MiR fleet, Symbotic, Geek+, Boston Dynamics Spot coordinator), or industrial automation FMS engineers (Dematic, Vanderlande, Swisslog)
• Multi-robot task allocation algorithm fluency: auction-based, MILP, Hungarian, market-based; understanding of NP-hardness and approximation trade-offs in MRTA problems
• Distributed systems engineering: consensus, leader election, eventual consistency, fault recovery — the orchestrator is a distributed system at heart
• VDA 5050 (v2.0+) protocol-level fluency: order, instantAction, state, visualization messages; master and client mode
• BehaviorTree.CPP v4 and Groot2 proficiency for the NodeGraph SDK side of the role
• Python and C++; comfort with optimisation libraries (OR-tools, Gurobi, CPLEX) for algorithm prototyping

Nice to Have

• Academic background in multi-robot systems: RSS/ICRA/IROS publications in MRTA, multi-robot motion planning, swarm robotics, or fleet coordination
• Open-RMF maintainer or significant contributor — fleet_adapter authoring, dispatcher implementation, RMF Demos
• Cooperative manipulation experience: bimanual across two physical robots, multi-robot assembly, leader/follower
• Per-customer FMS adapter experience: writing integrations to plug NEURA robots into existing third-party fleet managers (Otto, MiR, Locus, Symovo)
• Motion planning familiarity (MoveIt2, OMPL, or equivalent) sufficient to evaluate planner behaviour in pre-deployment simulation
• Formal verification or model checking for behaviour tree skill trees
• ROS 2 (Jazzy or Humble) — useful for ROS 2-based fleet