In the world of autonomous following vehicles, “accurate tracking” is only half of the challenge. The real test lies in ensuring the robot can follow smoothly while safely avoiding obstacles in complex environments. Behind this capability are the path planning and obstacle avoidance algorithms that form the brain of every intelligent mobile platform.
This article explores two of the most representative algorithms—A* and DWA (Dynamic Window Approach)—and analyzes which is more suitable for follow-me applications.
1. The Two Core Tasks of Path Planning
In a typical autonomous following system, path planning can be divided into two main tasks:
- Global Planning
Determines the overall route from the current position to the target, usually based on a static map. Examples include A*, Dijkstra, and D* Lite algorithms. - Local Planning
Continuously adjusts the path in real time according to sensor inputs, avoiding dynamic obstacles and maintaining smooth motion. Common examples are DWA, TEB, and APF algorithms.
For a follow-me robot, where the target keeps moving and the environment is constantly changing, local planning and real-time obstacle avoidance become the dominant factors.
2. A* Algorithm: The Reliable Foundation of Path Planning
A* (A-star) is the “veteran” of path planning algorithms.
It finds the least-cost path on a grid or graph using a heuristic search strategy, where the cost function is defined as:
f(n) = g(n) + h(n)f(n)
- g(n): Actual cost from the start to the current node
- h(n): Heuristic estimate from the current node to the goal
Advantages:
- Produces globally optimal and smooth paths
- Simple, stable, and well-understood
Limitations:
- Relies on a static map
- High computational cost
- Not ideal for real-time obstacle avoidance
In autonomous following vehicles, A* is typically used as a global path planner—it provides an ideal reference route, but cannot handle fast-changing environments on its own.
3. DWA Algorithm: The Dynamic Window Approach for Real-Time Navigation
The DWA (Dynamic Window Approach) is a cornerstone algorithm for local path planning and dynamic obstacle avoidance.
Unlike A*, which works in map space, DWA operates in velocity space—predicting and evaluating short-term trajectories based on the robot’s motion constraints.
The core steps of DWA are:
- Generate a set of possible velocity commands within the robot’s acceleration limits.
- Simulate short-term trajectories for each velocity pair (v, ω).
- Evaluate a cost function for every trajectory:
\text{Cost} = \alpha \cdot \text{Heading} + \beta \cdot \text{Velocity} + \gamma \cdot \text{Obstacle}- Choose the trajectory with the lowest cost.
Advantages:
- Excellent real-time performance
- Considers dynamic obstacles and kinematic limits
- Works well with LiDAR and vision sensors
Limitations:
- Sensitive to cost function tuning
- Can get stuck in local minima in tight spaces
For follow-me robots, DWA excels in dynamic environments such as shopping malls, parks, or industrial corridors, where real-time decision-making is essential.
4. Combining A* and DWA: The Best of Both Worlds
In practical autonomous following systems, the two algorithms are often combined:
- A* provides a global path over the map;
- DWA refines the trajectory locally to handle real-world obstacles.
This hybrid approach enables:
- Stable global navigation
- Real-time local avoidance
- Natural, human-like following behavior
At PSICV, we apply an A* + DWA + multi-sensor fusion strategy—where A* provides the reference path, DWA handles local dynamics, and visual or LiDAR perception ensures adaptive, safe motion.
5. Algorithm Selection Guide
5. Algorithm Selection Guide
| Application Scenario | Recommended Algorithm | Key Features |
| Indoor static environments (malls, warehouses) | A* + DWA | Clear global route with flexible local avoidance |
| Outdoor open areas (parks, campuses) | DWA + Vision | High adaptability and real-time performance |
| Dynamic human-following (strollers, mobility aids) | DWA + UWB/Vision Fusion | Balances target tracking and obstacle safety |
Conclusion
In the evolution of autonomous following vehicles, the key question is not “Which algorithm is best?” but “Which algorithm combination fits the task?”
- A* provides the sense of direction.
- DWA delivers the sense of motion.
- Sensor fusion (UWB, vision, LiDAR) brings the sense of perception.
The future will likely see further integration of deep reinforcement learning and graph-based neural planners, making path planning for follow-me robots ever more adaptive, intelligent, and human-aware.