AGV Path Planning: A* vs Dynamic Obstacle Avoidance
November 15, 2026
Why Path Planning Architecture Matters for AGV Buyers
When specifying an AGV, the navigation stack determines throughput, safety, and integration effort. Two core components are global path planning (often A*) and local dynamic obstacle avoidance. Buyers must understand how these interact to select the right sensors, compute modules, and software licenses. This article compares A* and dynamic avoidance from a sourcing and integration perspective, helping you specify components that meet cycle-time and safety requirements without overpaying.
A* Global Planning: Strengths and Limitations
A* is a graph-search algorithm that finds the shortest path on a known map using a heuristic. It is deterministic, computationally light, and well-suited for static environments like warehouses with fixed racks. For AGVs, A* typically runs on the fleet manager or onboard industrial PC, using a grid or topological map.
Key strengths: optimal paths, low CPU load, easy to validate. Limitations: it assumes a static map and cannot react to unexpected obstacles. In mixed traffic, pure A* leads to deadlocks or collisions unless paired with a local avoidance layer. When specifying, ensure the AGV controller or IPC has enough headroom for map resolution and replanning frequency. For large maps, A* can become memory-intensive; consider variants like D* Lite if dynamic replanning is needed.
Dynamic Obstacle Avoidance: Local Reactivity
Dynamic avoidance handles moving people, forklifts, and other AGVs. Common approaches include dynamic window approach (DWA), time-elastic bands (TEB), and velocity obstacles. These algorithms use real-time sensor data from 2D/3D LiDAR, depth cameras, or ultrasonic sensors to adjust velocity and steering.
Strengths: reacts to unknown obstacles, improves safety in mixed environments. Limitations: can be computationally heavy, may cause oscillatory motion, and requires careful tuning. Sensor choice is critical: 2D LiDAR is cost-effective for planar detection but misses low obstacles; 3D LiDAR or depth cameras provide richer data at higher cost. For AGVs operating near people, safety-rated laser scanners are often mandatory to achieve PL d or PL e per ISO 13849.
Integration: How A* and Avoidance Work Together
Most production AGVs use a layered approach: a global planner (A*) generates a route, and a local planner (DWA/TEB) executes it while avoiding obstacles. The local planner may deviate from the global path and then rejoin it. This architecture requires a map server, a localization module (e.g., SLAM or reflector-based), and a motion controller.
When sourcing, specify the interface between layers: does the local planner have authority to replan globally? What is the recovery behavior if the robot is blocked? For fleet operations, the traffic manager must coordinate right-of-way to prevent livelock. Ask suppliers for simulation logs or test data showing performance in crowded aisles.
What to Specify Before Ordering
- Map size and resolution: Determines memory and CPU for A*. For maps over 10,000 mΒ², confirm the controller's capacity.
- Sensor suite: 2D LiDAR for basic avoidance; 3D LiDAR or depth cameras for complex environments. Ensure safety scanners meet required PL.
- Compute platform: Industrial PC or embedded controller with sufficient cores for planning and perception. Check real-time OS support if needed.
- Software licensing: Clarify if path planning software is included or requires per-robot licenses. Open-source stacks (ROS/ROS2) can reduce cost but demand integration expertise.
- Safety certification: For collaborative applications, ensure avoidance functions are certified or can be validated.
Cost Considerations
Costs split into hardware (sensors, compute) and software (licenses, integration). A* itself is inexpensive to implement; the cost lies in map creation and maintenance. Dynamic avoidance drives sensor and compute costs. A common mistake is overspending on 3D LiDAR when a 2D scanner plus bumpers suffices for controlled zones. Conversely, skimping on safety scanners can lead to costly retrofits.
From Chinese factories, you can source LiDAR, industrial PCs, and motion controllers at competitive prices. However, software quality varies. Request a demo or sample unit to test path planning in your environment. For volume orders, negotiate firmware customization and long-term support.
Sourcing Tips from Chinese Factories
Many Chinese suppliers offer turnkey AGV kits or components. When evaluating:
- Ask for the planning algorithm's response time and replanning frequency under load.
- Verify sensor specifications with datasheets and third-party test reports.
- Check if the software stack is open for customization or locked.
- Request references from similar applications (e.g., e-commerce warehouses).
- Clarify warranty and spare parts availability for sensors and controllers.
For OEMs, consider separating hardware and software procurement: buy sensors and compute from cost-effective Chinese suppliers, but license planning software from a proven vendor if your team lacks robotics expertise.
Bottom Line
A* and dynamic obstacle avoidance are complementary, not competing. Specify a layered architecture with clear interfaces, right-size your sensor suite to the environment, and validate performance with real-world tests. Source hardware from Chinese factories for cost savings, but invest in software quality and safety certification. This balance delivers reliable AGV navigation without unnecessary expense.
π More Articles
OEE Tracking for Spare Parts: Reduce Robot Downtime
Nov 18, 2026Vacuum Gripper Leak Test: Acceptable Leakage Rates
Nov 17, 2026π’ Verify Chinese Suppliers
Reviewed & ranked robot parts manufacturers for global buyers
Find Suppliers β