Overhead Crane and AGV Collaborative Dispatching System

When multiple overhead cranes and AGVs/RGVs operate collaboratively in the same production area, the core challenge for the dispatching system is safely and efficiently allocating material handling tasks within limited space and time. The quality of path arbitration decisions directly impacts overall line efficiency—without algorithmic dispatching, multi-equipment conflicts occur roughly 15–30 times per hour; after path arbitration optimization, this drops to 0–2 times per hour. This article provides a complete breakdown of the engineering implementation of Kelude's overhead crane and AGV/RGV multi-equipment path arbitration dispatching system, covering task queue sequencing, the path arbitration engine, conflict detection algorithms, and dynamic re-planning.

Task Queue Priority Sequencing for Multi-Equipment Scheduling

Material handling tasks issued by MES/WMS are stored in the scheduling PLC's task queue (ring buffer with a capacity of 128 entries). Each task contains the material ID, start coordinates, destination coordinates, priority (0–100, with 100 for urgent tasks), issue timestamp, and deadline requirements. Priority sequencing uses a three-weight formula: Total Priority = Task Priority × 0.4 + Wait Time Coefficient × 0.3 + Deadline Urgency Coefficient × 0.3. The wait time coefficient is calculated as (current time − issue time) / standard wait time (normalized to 0–100), while the deadline urgency coefficient is elapsed time / total deadline (the closer to 1, the more urgent). This three-weight design ensures urgent tasks can jump the queue without causing indefinite starvation—the longer a task waits, the higher it rises in the queue.

During task allocation, the system scans the top 20 tasks in the queue and evaluates three allocation conditions in sequence: ① whether an overhead crane is idle and positioned near the task's start point (Crane travel direction distance ≤10m); ② whether an AGV is idle and located at the start workstation; ③ whether the path is conflict-free (no overlap with paths of already-assigned tasks). Tasks meeting all three conditions are assigned immediately; otherwise, they return to the wait queue for re-evaluation in the next cycle. The system iterates every 100ms, and tasks that have waited more than 60 seconds are force-assigned (conflicting tasks are automatically downgraded and moved to the wait queue). Kelude's dispatching system achieves an average task wait time of ≤15 seconds, with urgent tasks averaging ≤5 seconds.

Conflict Detection and Dynamic Path Re-planning

The core difficulty in multi-equipment coordination lies in real-time conflict detection and the effectiveness of arbitration strategies. Conflict detection operates in two steps: zone overlap detection (checking whether two equipment work areas intersect) and time-window analysis (predicting whether two equipment trajectories will cross within the next 3–5 seconds). Detection runs every 100ms, synchronized with the dispatch cycle. When a conflict is detected, it is classified into one of five scenarios, each triggering a corresponding arbitration strategy:

Path Arbitration Engine: Algorithm Implementation

The path arbitration engine is the core decision-making component of the dispatching system. It runs inside the cyclic interrupt OB35 (100 ms cycle) of the dispatching PLC (S7-1500). The engine executes in three steps: task-to-equipment matching (covered in the previous section), followed by path planning, and finally conflict validation. For path planning, the A* heuristic search algorithm is recommended—starting from the equipment's current position and ending at the target workstation, with a cost function F = G + H (where G is the distance already traveled and H is the Manhattan distance to the destination). The search space is constrained to the reachable path network derived from the gridded factory map (node spacing 2 m; the S7-1500 can store up to 200 nodes). A single path search executes in 2–5 ms on the S7-1500 (depending on search depth, with a maximum of 50 nodes searched). For AGV paths, bidirectional travel and U-turn restrictions are considered, and a direction-change penalty is added to the path cost (each turn adds an extra 3 m to encourage long, straight paths).

Time-window conflict detection serves as the second line of defense in path arbitration. Once a path is assigned to each piece of equipment, the system decomposes it into a series of timestamped path segments (2 m per segment, each with an expected arrival time window of ±0.5 s). When a new task is assigned a path, the system checks every segment of that path against the time windows of already-assigned paths to detect overlaps (i.e., whether two pieces of equipment have intersecting time windows on the same path segment). The overlap condition is: |T_{Equipment A} – T_{Equipment B}| < safety interval (3 s for overhead crane-to-overhead crane, 2 s for overhead crane-to-AGV, and 1.5 s for AGV-to-AGV). If an overlap is detected, the path assignment is rejected, and the system searches for an alternative path or waits for the next cycle. The time-window array is stored in a DB block on the S7-1500 (each piece of equipment reserves 128 path segment records at 12 bytes per record; 12 pieces of equipment consume approximately 18 KB in total).

Final decision matrix for path arbitration: The dispatching system maintains an N×N mutual exclusion matrix (N = total number of equipment units). Matrix element M[i][j] = 0 indicates no path conflict between equipment i and j, while M[i][j] = 1 indicates a conflict (written by time-window detection). During arbitration, conflicting equipment pairs are traversed in descending priority order: the higher-priority equipment keeps its path unchanged, while the lower-priority equipment re-plans its route. If the lower-priority equipment has no alternative path (all reachable paths conflict), it enters a waiting state and the reason is logged. The matrix is cleared and recalculated every scheduling cycle (100 ms). In a 12-equipment scenario, Kelude's dispatching system completes path arbitration in ≤8 ms (including time-window detection), leaving 92 ms for communication and equipment control logic.

Multi-Equipment Collaborative Dispatching System: Architecture & Deployment

The dispatching system's hardware architecture is organized into three tiers: the field level (overhead crane PLCs, AGV onboard controllers, and RGV controllers connected to an industrial network switch via Profinet IRT), the dispatching level (a dispatching PLC S7-1500 or industrial PC running the path arbitration engine and task queue management), and the management level (MES/WMS servers exchanging work orders and task data with the dispatching level via OPC UA). Communication cycles between field devices and the dispatching PLC are: overhead cranes ≤50 ms (Profinet IRT, jitter ±1 μs), AGVs ≤100 ms (Profinet RT), and RGVs ≤50 ms. Communication between the dispatching PLC and MES depends on the task dispatch frequency, with a typical batch exchange every 100 ms (approximately 2 KB per exchange, including task queues and status reports).

Deployment case study — automotive final assembly workshop: 12 overhead cranes (5 LD Type, 3 QD Type, and 4 jib cranes) + 24 AGVs (16 under-cart type and 8 forklift type) + 8 RGVs, totaling 44 pieces of equipment operating collaboratively in a single plant area (approximately 200 m × 80 m). Before deployment, manual dispatching resulted in roughly 20 conflicts per hour (peaking at 40 per hour during high-demand periods), with an average task waiting time of 45 seconds. After deploying Kelude's dispatching system (dispatching PLC: S7-1500 CPU 1516-3 PN/DP, with a CP1543-1 module for OPC UA communication), the conflict rate dropped to 0–2 per hour, average task waiting time fell to 15 seconds, and dispatching efficiency improved by 42% (measured by the number of material handling tasks completed per hour). The project timeline from site survey to go-live was 8 weeks (including 3 weeks of software configuration, 2 weeks of on-site commissioning, and 1 week of integrated trial operation).

Scalability design: The dispatching system uses a modular architecture. Adding new equipment only requires adding a device configuration entry (approximately 50 bytes per unit) to the DB block of the dispatching PLC and configuring the device type and communication address on the HMI. A single S7-1500 can manage up to 12 pieces of equipment (limited by program scan cycle and communication resources). Beyond 12 units, a second dispatching PLC is stacked for zone-based dispatching—splitting by plant area (e.g., Zone A dispatching PLC manages 6 overhead cranes + 10 AGVs, while Zone B manages 8 RGVs + 14 AGVs). Boundary task data is exchanged between the two zones via a PN/PN Coupler or industrial Ethernet. Kelude's dispatching system supports expansion to up to 64 pieces of equipment in a single plant (with 4 dispatching PLCs cascaded).

Frequently Asked Questions