General Axis engineers autonomous vehicle platforms and AI control systems for cargo movement, berth-side operations, and terminal logistics, with safety, uptime, and throughput as core design constraints.
Target platform availability for mission-critical terminal operations.
Control-loop responsiveness for precision navigation and docking.
Continuous operations model with fleet monitoring and failover.
Repeatable docking precision at transfer bays and service stations.
Ports move the world's goods, yet much of that work is still dangerous, repetitive, and constrained by labor. We build autonomy that makes terminals safer and more productive.
General Axis is an autonomy company focused exclusively on the port and wharf environment. We combine automotive-grade sensing, real-time control, and fleet-level orchestration into vehicles that operate around the clock in some of the most demanding industrial settings in the world.
Our approach is safety-first and deployment-driven. Every system is engineered to fail safe, validated under real operating conditions, and rolled out in stages alongside the people who run the terminal, not as a replacement for their expertise, but as a force multiplier for it.
Read our full mission →Four tightly-coupled layers turn raw terminal sensing into coordinated, safe, around-the-clock vehicle operation.
Sensor-fusion stack for obstacle classification, pedestrian awareness, and dynamic risk handling across mixed-operability zones.
Mission orchestration that assigns routes and priorities to autonomous assets based on berth operations and turnaround targets.
Sub-20ms control loops for precision maneuvering near vessel edges, crane zones, and centimeter-level docking alignment.
Supervisory telemetry tracking fleet state, route adherence, utilization, and predictive maintenance signals in live operations.
A staged path that de-risks deployment and keeps humans in the loop until the system has earned trust.
Survey wharf lanes, staging zones, and handoff points; build a high-fidelity digital map of the terminal.
Mount sensors, calibrate the autonomy stack, and connect to existing yard and berth scheduling systems.
Run staged rollouts with human oversight, validating safety envelopes against real operating conditions.
Hand over to continuous autonomous operation with 24/7 fleet monitoring and failover readiness.
A terminal runs on many vehicles acting as one. Our research focuses on how robots sense, decide, and stay coordinated, even when the network is imperfect.
We treat the wireless link itself as a sensing modality. Signal strength, latency, and connectivity between vehicles reveal relative position, network topology, and coverage gaps. The fleet's radio becomes another way to perceive the terminal.
Fleets reason over time, not just the next move. We model task allocation, routing, and berth handoffs as sequential decisions under uncertainty, so dozens of vehicles plan cooperatively toward turnaround targets instead of competing for the same lanes and bays.
Coordination has to survive dropped links, sensor faults, and degraded vehicles. Our methods keep the fleet safely productive when communication is lost or assets fail, so the system slows down gracefully instead of stalling.
Every vehicle carries overlapping vision, LiDAR, radar, and localization sensors so no single failure or weather condition can blind it. Each modality covers the others' weaknesses.
Primary semantic vision: worker and pedestrian detection, ISO container-code OCR, lane and edge markings, and signal recognition across the full ring around the vehicle. HDR holds detail through the harsh sun-to-shadow transitions of quay-side operation.
Early detection of vehicles, cranes-in-motion, and obstructions far down a transit corridor, giving the planner the lookahead needed for smooth speed control at lane speed.
Passive metric depth for close-quarters maneuvers, verifying clearance during reversing, bay approach, and alignment where centimeter accuracy matters.
The backbone of 3D perception: a dense, metrically-accurate point cloud for obstacle detection, free-space estimation, and LiDAR localization that is fully independent of lighting, with identical performance day or night.
High-resolution forward sensing for long-range, small-object detection in the direction of travel, with automotive-grade durability against the vibration and shock of terminal surfaces.
Eliminates the low blind zone around the chassis, detecting curbs, dropped lashing gear, pallets, and crouching workers that fall beneath the main sensor planes.
All-weather redundancy and direct velocity measurement. Where camera and LiDAR degrade in sea fog or blowing dust, radar keeps tracking and supplies the instantaneous closing-speed used for collision avoidance.
Centimeter-level global position and heading fused with inertial dead-reckoning. Under cranes and between high stacks where satellites drop out, the stack falls back to LiDAR SLAM against the HD map with no loss of pose continuity.
The last line: ultrasonic ranging for final-centimeter dock alignment, plus physical contact edges that trigger an immediate safe stop independent of the compute path.
An interactive, top-down model of the autonomy stack at work. Drive the AGV or let it run on autopilot, roll in sea fog, and watch each sensor modality pick up workers, traffic, and dropped gear in real time.
Raw sensor data becomes safe motion through eight tightly-coupled stages, running on-vehicle every control cycle. Each stage feeds the next with bounded, deterministic latency.
Every sensor is hardware time-stamped over PTP and spatially calibrated to a common vehicle frame. Point clouds are motion-compensated so a 20 Hz LiDAR sweep and a 60 fps camera frame describe the same instant in the world.
Camera, LiDAR, and radar fuse into one coherent scene, combining the semantic richness of vision, the metric precision of LiDAR, and the velocity and all-weather reliability of radar into a single occupancy and object model.
Deep networks detect, classify, and segment workers, vehicles, containers, and infrastructure, then track each object across time with stable IDs and motion state, running on-device under TensorRT for deterministic latency.
The vehicle places itself on a centimeter-accurate HD map of the terminal, fusing RTK-GNSS, inertial odometry, and LiDAR map-matching so pose stays continuous even when satellites are occluded by cranes and container stacks.
For every tracked agent (a turning reach-stacker, a crossing worker, a merging tractor), the stack forecasts likely future trajectories, so the planner reasons about where the world will be, not just where it is.
A three-tier planner turns a dispatched mission into a safe, comfortable trajectory: mission planning selects the route, behavior planning decides to yield, merge, or dock, and motion planning produces a kinematically-feasible path.
A model-predictive controller closes the loop at sub-20-millisecond cadence, commanding steering, throttle, and braking through redundant drive-by-wire actuation for centimeter-accurate path tracking and docking.
A separate, independently-powered safety channel validates every command against geofences and proximity limits. On any fault it commands a minimal-risk condition (a controlled stop), backed by a hardware e-stop that bypasses main compute entirely.
From automotive-grade edge compute up to cloud fleet orchestration. Every layer is containerized, independently updatable, and built on open, proven robotics foundations.
Autonomy in a working terminal only earns its place if it is demonstrably safe. We treat functional safety as a first-class engineering requirement at every layer of the stack.
Independent sensing, compute, power, and braking paths so no single fault can remove the vehicle's ability to stop safely.
A separate, isolated channel validates every motion command against geofences and proximity limits, outside the main compute path.
On any detected fault the vehicle executes a controlled stop to a known-safe state, backed by a hardware emergency stop.
Staged deployment keeps trained operators supervising until the system has demonstrably earned trust in the real environment.
Vehicles operate only within mapped, approved operational design domains, with speed and behavior bounded per zone.
Every mission is logged with full sensor and decision telemetry for traceability, incident review, and continuous improvement.
One autonomy platform, adapted to the full range of port and wharf operations, from the quay edge to the yard and the gate.
Autonomous transport between quay cranes and the yard, with precise berth-edge maneuvering inside active crane envelopes.
Coordinated stacking, retrieval, and horizontal transport across high-density storage blocks with minimal idle time.
Retrofit and purpose-built autonomous tractors and automated guided vehicles for repetitive intra-terminal haulage.
Low-speed precision movement for roll-on/roll-off operations where clearances are tight and traffic is mixed.
Rail-to-yard and gate-to-stack continuity that keeps containers moving across modal handoff points.
Self-scheduled charging and return-to-service so fleet availability stays high without manual intervention.
Each concept below maps to a specific capability area in autonomous port and wharf operations.
→ High-level architecture for autonomous vehicle coordination across wharf lanes, staging zones, and berth-side handoff points.
→ Sequenced movement model for loading, unloading, and transfer cycles between quay cranes, transport vehicles, and yard nodes.
→ Operational zoning strategy that separates manned and autonomous paths to reduce conflict points in high-density areas.
→ Task orchestration concept for assigning routes and priorities to autonomous assets based on berth operations and turnaround targets.
→ Precision guidance concept for safe maneuvering near vessel edges, crane zones, and restricted infrastructure boundaries.
→ Continuous control approach linking storage yard dispatch with wharf-side execution for reduced idle time and queue buildup.
→ System telemetry concept for tracking fleet state, route adherence, and predictive maintenance signals in live operations.
→ Fallback and recovery framework for obstruction events, schedule drift, and mission reassignment under changing port conditions.
→ Energy management concept for charge scheduling, dock alignment, and return-to-service without disrupting mission throughput.
→ Sensor fusion model for obstacle classification, pedestrian awareness, and dynamic risk handling in mixed-operability zones.
→ Docking control concept for repeatable alignment at transfer bays and service stations with centimeter-level stopping accuracy.
→ Supervisory interface concept showing mission progress, utilization, alerts, and dispatch balance across vehicle groups.
→ Macro-level systems view linking berth activity, transit corridors, and yard throughput into one autonomous command picture.
→ Proximity safety concept for directional sensing, blind-spot mitigation, and safe braking margins during reverse and forward motion.
Straight answers on how the platform is built, deployed, and operated safely.
Contact
For collaboration, pilot deployments, and integration planning across port and wharf environments, reach out to our team. We typically respond within two business days.
Prefer email? Write to contact@generalaxisrobotics.com.