Design of YOLOv8-SLAM Robot Path Planning Method for Outdoor Complex Environments

Imagine a robot trying to walk through a busy city park. To a human, the path is obvious: step around the person sitting on the bench, wait for the cyclist to pass, and keep moving toward the fountain. To a machine, this scene is a chaotic storm of data. It must instantly recognize that a moving shape is a person and not a tree, understand that a parked car is a solid wall it cannot pass, and calculate a safe route without bumping into anyone. This is the core challenge of autonomous navigation in the real world. While robots have become quite good at moving in empty rooms or on factory floors, the outdoors is a different beast. The light changes, shadows stretch, and people and vehicles move unpredictably. If a robot cannot see a small child running across the path or misjudge the distance to a moving car, the result is a collision or a robot that gets stuck, unable to decide what to do next.

For years, engineers have tried to solve this by building separate systems for seeing, knowing where they are, and planning where to go. One system looks at the camera feed to find objects, another uses those images to build a map of the surroundings, and a third decides the best route. The problem is that these systems often work in isolation. If the "seeing" system misses a small sign or a person in the distance, the "planning" system might send the robot right into them. If the "mapping" system gets confused by a moving car, the robot might think the road is blocked when it is actually clear. Researchers at Chuzhou Polytechnic in China have proposed a new way to handle this complexity. They developed a single, connected system that ties seeing, mapping, and planning together so that a mistake in one area can be corrected by the others, allowing a robot to navigate safely through crowded, outdoor spaces.

The researchers started by improving how the robot sees the world. They used a powerful image-recognition tool called YOLOv8, which is like a very fast and sharp pair of eyes, but they made it smarter for difficult situations. In a busy outdoor scene, small objects like traffic cones, distant pedestrians, or road signs are often hard to spot because they are tiny or hidden behind other things. The team added special attention mechanisms to the software, teaching it to focus intensely on these small details and to ignore confusing background noise like swaying tree branches or shifting shadows. They also gave the robot the ability to understand the scene not just as a collection of shapes, but as a map of different types of ground and obstacles, labeling every pixel of the image as either a road, a sidewalk, a person, or a vehicle. This detailed understanding is crucial because it tells the robot not just where things are, but what they are and whether they are safe to move near.

Once the robot can see clearly, the next challenge is knowing exactly where it is. Traditional mapping systems often assume the world is still, like a museum. But in a park or a street, people and cars are constantly moving. If a robot tries to build a map while a person walks past, it might get confused and think the person is part of the building, causing the robot to lose its way. The new system solves this by using the detailed "eyes" it just built to identify moving objects and temporarily ignore them while it figures out its position. It checks if an object is moving by comparing its position over several frames of video and looking at how deep it is in the scene. If the system confirms something is moving, it filters that information out before building the map. This allows the robot to create a stable, accurate map of the static world, even while people and cars are rushing around it.

With a clear view and a reliable map, the robot must then decide how to move. The researchers created a two-step planning strategy that works like a human thinking ahead. First, the robot looks at the big picture to find a general route from its starting point to its destination. It does this by considering not just the shortest distance, but also the "risk" of the path. A path that is slightly longer but goes through a quiet, open area is chosen over a shorter path that cuts through a crowd of people. Then, as the robot moves, a second, faster system constantly adjusts its steps to avoid sudden obstacles. If a pedestrian steps out unexpectedly, this local system instantly calculates a new, safe trajectory to dodge them without stopping the whole journey. The two systems talk to each other constantly, ensuring that the robot's long-term goal and its immediate safety are always in sync.

The team tested this new approach in a variety of ways, from computer simulations to real-world trials on a campus. In simulations filled with moving obstacles, the new system successfully reached its goal 95.31% of the time, significantly outperforming older methods that struggled with collisions or getting lost. When they took the robot out onto real campus roads with actual pedestrians, parked cars, and winding paths, the results were even more impressive. The robot completed 57 out of 60 navigation runs without hitting anything or getting stuck, a success rate of 95%. It also managed to travel a shorter distance on average and made fewer near-misses than the other systems tested. The robot was able to spot small objects like traffic signs and bicycles from far away, ignore the confusion of moving people, and find a smooth, safe path through the chaos.

This work suggests that the key to reliable outdoor robots is not just making each part of the system better, but making them work together as a single, intelligent unit. By connecting the ability to see small details, the ability to filter out moving distractions, and the ability to plan safe routes, the researchers have created a system that handles the unpredictability of the real world much better than before. While the system still faces challenges in extreme weather or with very dense crowds, the results show a clear path forward. The robot is no longer just following a pre-programmed line; it is understanding its environment, predicting what might happen next, and making smart decisions to get where it needs to go.